Category: 3. Business

Alvioli, M., Marchesini, I., Reichenbach, P., Rossi, M., Ardizzone, F., Fiorucci, F., and Guzzetti, F.: Automatic delineation of geomorphological slope units with r.slopeunits v1.0 and their optimization for landslide susceptibility modeling, Geosci. Model Dev., 9, 3975–3991, https://doi.org/10.5194/gmd-9-3975-2016, 2016.  

ARPA Piemonte: Eventi idrometeorologici dal 19 al 24 ottobre 2019. Rapporto evento, Agenzia Regionale per la Protezione Ambientale – Regione Piemonte, Torino, https://www.arpa.piemonte.it/pubblicazione/eventi-alluvionali-piemonte-evento-19-24-ottobre-2019 (last access: 16 December 2023), 2019. 

ARPA Piemonte: https://www.arpa.piemonte.it, last access: 15 February 2024. 

Bandis, S. C., Del Monaco, G., Margottini, C., Serafini, S., Trocciola, A., Dutto, F., and Mortara, G.: Landslide phenomena during the extreme meteorological event of 4–6 November 1994 in Piemonte Region, N. Italy, in: Proceedings of the 7th International Symposium on Landslides, Trondheim, Norway, 159–164, 1996. 

Baum, R. L., Savage, W. Z., and Godt, J. W.: TRIGRS – A Fortran Program for Transient Rainfall Infiltration and Grid-Based Regional Slope-Stability Analysis, Version 2.0, U.S. Geol. Surv. Open-File Rep., 75, https://pubs.usgs.gov/of/2008/1159/ (last access: 31 October 2023), 2008. 

Beguería, S.: Validation and Evaluation of Predictive Models in Hazard Assessment and Risk Management, Nat. Hazards, 37, 315–329, https://doi.org/10.1007/s11069-005-5182-6, 2006. 

Bernardie, S., Vandromme, R., Mariotti, A., Houet, T., Grémont, M., Grandjean, G., Bouroullec, I., and Thiery, Y.: Estimation of Landslides Activities Evolution Due to Land–Use and Climate Change in a Pyrenean Valley, in: Advancing Culture of Living with Landslides, Springer International Publishing, Cham, 859–867, https://doi.org/10.1007/978-3-319-53498-5_98, 2017. 

Bogaard, T. and Greco, R.: Invited perspectives: Hydrological perspectives on precipitation intensity-duration thresholds for landslide initiation: proposing hydro-meteorological thresholds, Nat. Hazards Earth Syst. Sci., 18, 31–39, https://doi.org/10.5194/nhess-18-31-2018, 2018. 

Bordoni, M., Vivaldi, V., Lucchelli, L., Ciabatta, L., Brocca, L., Galve, J. P., and Meisina, C.: Development of a data-driven model for spatial and temporal shallow landslide probability of occurrence at catchment scale, Landslides, 18, 1209–1229, https://doi.org/10.1007/s10346-020-01592-3, 2021. 

Brenning, A.: Spatial prediction models for landslide hazards: review, comparison and evaluation, Nat. Hazards Earth Syst. Sci., 5, 853–862, https://doi.org/10.5194/nhess-5-853-2005, 2005. 

Caine, N.: The Rainfall Intensity: Duration Control of Shallow Landslides and Debris Flows, Geogr. Ann. A, 62, 23–27, 1980. 

Camera, C. A. S., Bajni, G., Corno, I., Raffa, M., Stevenazzi, S., and Apuani, T.: Introducing intense rainfall and snowmelt variables to implement a process-related non-stationary shallow landslide susceptibility analysis, Sci. Total Environ., 786, 147360, https://doi.org/10.1016/j.scitotenv.2021.147360, 2021. 

Campbell, R. H.: Soil Slips, Debris Flows, and Rainstorms in the Santa Monica Mountains and Vicinity, Southern California, U.S. Geol. Surv. Prof. Pap. 851, 51 pp., https://pubs.usgs.gov/pp/0851/report.pdf (last access: 31 October 2023), 1975. 

Cantarino, I., Carrion, M. A., Goerlich, F., and Martinez Ibañez, V.: A ROC analysis-based classification method for landslide susceptibility maps, Landslides, 16, 265–282, https://doi.org/10.1007/s10346-018-1063-4, 2019. 

Carrara, A.: Multivariate models for landslide hazard evaluation, J. Int. Ass. Math. Geol., 15, 403–426, https://doi.org/10.1007/BF01031290, 1983. 

Carrara, A., Cardinali, M., Detti, R., Guzzetti, F., Pasqui, V., and Reichenbach, P.: GIS techniques and statistical models in evaluating landslide hazard, Earth Surf. Proc. Land., 16, 427–445, https://doi.org/10.1002/esp.3290160505, 1991. 

Carrara, A., Crosta, G., and Frattini, P.: Comparing models of debris-flow susceptibility in the alpine environment, Geomorphology, 94, 353–378, https://doi.org/10.1016/j.geomorph.2006.10.033, 2008. 

Coletti, G., Basso, D., Frixa, A., and Corselli, C.: Transported Rhodoliths Witness the lost carbonate factory: A case history from the miocene pietra da cantoni limestone (Nw Italy), Riv. Ital. Paleontol. S., 121, 345–368, https://doi.org/10.13130/2039-4942/6522, 2015. 

Crosta, G.: Regionalization of rainfall thresholds: An aid to landslide hazard evaluation, Environ. Geol., 35, 131–145, https://doi.org/10.1007/s002540050300, 1998. 

Crosta, G. B. and Frattini, P.: Distributed modelling of shallow landslides triggered by intense rainfall, Nat. Hazards Earth Syst. Sci., 3, 81–93, https://doi.org/10.5194/nhess-3-81-2003, 2003. 

Crozier, M. J.: Prediction of rainfall-triggered landslides: A test of the antecedent water status model, Earth Surf. Proc. Land., 24, 825–833, 1999. 

Cruden, D. M. and Varnes, D. J.: Landslide types and processes. Landslides: Investigation and Mitigation, Special Report 247, Transportation Research Board, Washington, 36–75, 1996. 

Davis, J. and Goadrich, M.: The relationship between Precision-Recall and ROC curves, in: Proceedings of the 23rd international conference on Machine learning – ICML ’06. Association for Computing Machinery, New York, NY, USA, 233–240, https://doi.org/10.1145/1143844.1143874, 2006. 

Drummond, C. and Holte, R. C.: Cost curves: An improved method for visualizing classifier performance, Mach. Learn., 65, 95–130, https://doi.org/10.1007/s10994-006-8199-5, 2006. 

Fioravanti, G., Fraschetti, P., Lena, F., Perconti, W., and Emanuela, P. (ISPRA): I normali climatici 1991–2020 di temperatura e precipitazione in Italia, Stato dell’ambiente, 9, https://www.isprambiente.gov.it /files2022/pubblicazioni/stato-ambiente/i-normali-climatici-1991-2020-di-temperatura-e-precipitazione-in-italia_19082022.pdf (last access: 9 January 2024), 2022. 

Flach, P. A. and Kull, M.: Precision-Recall-Gain curves: PR analysis done right, Adv. Neural Inf. Process. Syst., 2015 January, 838–846, 2015. 

Frattini, P. and Crosta, G. B.: The role of material properties and landscape morphology on landslide size distributions, Earth Planet. Sc. Lett., 361, 310–319, https://doi.org/10.1016/j.epsl.2012.10.029, 2013. 

Frattini, P., Crosta, G., and Sosio, R.: Approaches for defining thresholds and return periods for rainfall-triggered shallow landslides, Hydrol. Process., 23, 1444–1460, https://doi.org/10.1002/hyp.7269, 2009. 

Frattini, P., Crosta, G., and Carrara, A.: Techniques for evaluating the performance of landslide susceptibility models, Eng. Geol., 111, 62–72, https://doi.org/10.1016/j.enggeo.2009.12.004, 2010. 

Fredlund, D. G., Morgenstern, N. R., and Widger, R. A.: Shear Strength of Unsaturated Soils, Can. Geotech. J., 15, 313–321, https://doi.org/10.1139/t78-029, 1978. 

Gilbert, G. K.: Finley’s Tornado Predictions, Am. Meteorol. J., 1, 166–172, 1884. 

Glade, T., Crozier, M., and Smith, P.: Applying probability determination to refine landslide-triggering rainfall thresholds using an empirical “Antecedent Daily Rainfall Model,” Pure Appl. Geophys., 157, 1059–1079, https://doi.org/10.1007/s000240050017, 2000. 

Godt, J. W., Baum, R. L., and Chleborad, A. F.: Rainfall characteristics for shallow landsliding in Seattle, Washington, USA, Earth Surf. Proc. Land., 31, 97–110, https://doi.org/10.1002/esp.1237, 2006. 

Goetz, J. N., Brenning, A., Petschko, H., and Leopold, P.: Evaluating machine learning and statistical prediction techniques for landslide susceptibility modeling, Comput. Geosci., 81, 1–11, https://doi.org/10.1016/j.cageo.2015.04.007, 2015. 

Guzzetti, F., Carrara, A., Cardinali, M., and Reichenbach, P.: Landslide hazard evaluation: a review of current techniques and their application in a multi-scale study, Central Italy, Geomorphology, 31, 181–216, https://doi.org/10.1016/S0169-555X(99)00078-1, 1999. 

Guzzetti, F., Peruccacci, S., Rossi, M., and Stark, C. P.: Rainfall thresholds for the initiation of landslides in central and southern Europe, Meteorol. Atmos. Phys., 98, 239–267, https://doi.org/10.1007/s00703-007-0262-7, 2007. 

Hanley, J. A. and McNeil, B. J.: The meaning and use of the area under a receiver operating characteristic (ROC) curve, Radiology, 143, 29–36, https://doi.org/10.1148/radiology.143.1.7063747, 1982. 

Heidke, P.: Berechnung Des Erfolges Und Der Güte Der Windstärkevorhersagen Im Sturmwarnungsdienst, Geogr. Ann., 8, 301–349, https://doi.org/10.1080/20014422.1926.11881138, 1926. 

Hersbach, H., Bell, B., Berrisford, P., Hirahara, S., Horányi, A., Muñoz-Sabater, J., Nicolas, J., Peubey, C., Radu, R., Schepers, D., Simmons, A., Soci, C., Abdalla, S., Abellan, X., Balsamo, G., Bechtold, P., Biavati, G., Bidlot, J., Bonavita, M., De Chiara, G., Dahlgren, P., Dee, D., Diamantakis, M., Dragani, R., Flemming, J., Forbes, R., Fuentes, M., Geer, A., Haimberger, L., Healy, S., Hogan, R. J., Hólm, E., Janisková, M., Keeley, S., Laloyaux, P., Lopez, P., Lupu, C., Radnoti, G., de Rosnay, P., Rozum, I., Vamborg, F., Villaume, S., and Thépaut, J. N.: The ERA5 global reanalysis, Q. J. Roy. Meteor. Soc., 146, 1999–2049, https://doi.org/10.1002/qj.3803, 2020. 

Hosmer, D. W. and Lemeshow, S.: Applied Logistic Regression, John Wiley & Sons, Inc., Hoboken, NJ, USA, https://doi.org/10.1002/0471722146, 2000. 

Huang, F., Cao, Z., Guo, J., Jiang, S. H., Li, S., and Guo, Z.: Comparisons of heuristic, general statistical and machine learning models for landslide susceptibility prediction and mapping, Catena, 191, 104580, https://doi.org/10.1016/j.catena.2020.104580, 2020. 

Hungr, O.: A review of landslide hazard and risk assessment methodology, Landslides Eng. Slopes. Exp. Theory Pract., 1, 3–27, https://doi.org/10.1201/b21520-3, 2016. 

Hürlimann, M., Guo, Z., Puig-Polo, C., and Medina, V.: Impacts of future climate and land cover changes on landslide susceptibility: regional scale modelling in the Val d’Aran region (Pyrenees, Spain), Landslides, 19, 99–118, https://doi.org/10.1007/s10346-021-01775-6, 2022. 

Hwang, C.-L. and Yoon, K.: Multiple Attribute Decision Making: methods and applications a state-of-the-art survey, Springer Sci. Bus. Media, 186, https://doi.org/10.1007/978-3-642-48318-9, 1981. 

INTERREG IIC: Descrizione dei principali eventi alluvionali che hanno interessato la regione Piemonte, Liguria e nella Spagna Nord Orientale, 90–94 pp., https://www.arpa.piemonte.it /pubblicazione/descrizione-dei-principali-eventi-alluvionali-che-hanno-interessato-regione-piemonte (last access: 25 March 2024), 1998. 

ISPRA: Dissesto idrogeologico in Italia: pericolosità e indicatori di rischio, 183 pp., https://www.isprambiente.gov.it /it/pubblicazioni/rapporti/dissesto-idrogeologico-in-italia-pericolosita-e-indicatori-di-rischio-edizione-2021 (last access: 20 January 2024), 2021. 

Iverson, R. M.: Landslide triggering by rain infiltration, Water Resour. Res., 36, 1897–1910, https://doi.org/10.1029/2000WR900090, 2000. 

Jacobs, L., Kervyn, M., Reichenbach, P., Rossi, M., Marchesini, I., Alvioli, M., and Dewitte, O.: Regional susceptibility assessments with heterogeneous landslide information: Slope unit- vs. pixel-based approach, Geomorphology, 356, 107084, https://doi.org/10.1016/j.geomorph.2020.107084, 2020. 

Jenks, G. F.: The data model concept in statistical mapping, Int. Yearb. Cartogr., 7, 186–190, 1967. 

Jones, J. N., Boulton, S. J., Bennett, G. L., Stokes, M., and Whitworth, M. R. Z.: Temporal Variations in Landslide Distributions Following Extreme Events: Implications for Landslide Susceptibility Modeling, J. Geophys. Res.-Earth, 126, 1–26, https://doi.org/10.1029/2021JF006067, 2021. 

Kirschbaum, D. and Stanley, T.: Satellite-Based Assessment of Rainfall-Triggered Landslide Hazard for Situational Awareness, Earth’s Future, 6, 505–523, https://doi.org/10.1002/2017EF000715, 2018. 

Knevels, R., Petschko, H., Proske, H., Leopold, P., Maraun, D., and Brenning, A.: Event-based landslide modeling in the styrian basin, Austria: Accounting for time-varying rainfall and land cover, Geosciences, 10, 1–27, https://doi.org/10.3390/geosciences10060217, 2020. 

Licata, M., Buleo Tebar, V., Seitone, F., and Fubelli, G.: The Open Landslide Project (OLP), a New Inventory of Shallow Landslides for Susceptibility Models: The Autumn 2019 Extreme Rainfall Event in the Langhe-Monferrato Region (Northwestern Italy), Geosciences, 13, 289, https://doi.org/10.3390/geosciences13100289, 2023. 

Lombardo, L. and Mai, P. M.: Presenting logistic regression-based landslide susceptibility results, Eng. Geol., 244, 14–24, https://doi.org/10.1016/j.enggeo.2018.07.019, 2018. 

Lombardo, L., Opitz, T., Ardizzone, F., Guzzetti, F., and Huser, R.: Space-time landslide predictive modelling, Earth-Sci. Rev., 209, 103318, https://doi.org/10.1016/j.earscirev.2020.103318, 2020. 

Lu, N. and Godt, J.: Infinite slope stability under steady unsaturated seepage conditions, Water Resour. Res., 44, 1–13, https://doi.org/10.1029/2008WR006976, 2008. 

Luino, F.: The Flood and Landslide Event of November 4-6 1994 in Piedmont Region (Northwestern Italy): Causes and Related Effects in Tanaro Valley, Phys. Chem. Earth, Part A Solid Earth Geod, 24, 123–129, https://doi.org/10.1016/S1464-1895(99)00007-1, 1999. 

Luino, F.: Sequence of instability processes triggered by heavy rainfall in the Northern Italy, Geomorphology, 66, 13–39, https://doi.org/10.1016/j.geomorph.2004.09.010, 2005. 

Malamud, B. D., Turcotte, D. L., Guzzetti, F., and Reichenbach, P.: Landslide inventories and their statistical properties, Earth Surf. Proc. Land., 29, 687–711, https://doi.org/10.1002/esp.1064, 2004. 

Mandarino, A., Luino, F., and Faccini, F.: Flood-induced ground effects and flood-water dynamics for hydro-geomorphic hazard assessment: the 21–22 October 2019 extreme flood along the lower Orba River (Alessandria, NW Italy), J. Maps, 17, 136–151, https://doi.org/10.1080/17445647.2020.1866702, 2021. 

Maraun, D., Knevels, R., Mishra, A. N., Truhetz, H., Bevacqua, E., Proske, H., Zappa, G., Brenning, A., Petschko, H., Schaffer, A., Leopold, P., and Puxley, B. L.: A severe landslide event in the Alpine foreland under possible future climate and land-use changes, Commun. Earth Environ., 3, 1–11, https://doi.org/10.1038/s43247-022-00408-7, 2022. 

Marc, O., Gosset, M., Saito, H., Uchida, T., and Malet, J. P.: Spatial Patterns of Storm-Induced Landslides and Their Relation to Rainfall Anomaly Maps, Geophys. Res. Lett., 46, 11167–11177, https://doi.org/10.1029/2019GL083173, 2019. 

Marino, P., Peres, D. J., Cancelliere, A., Greco, R., and Bogaard, T. A.: Soil moisture information can improve shallow landslide forecasting using the hydrometeorological threshold approach, Landslides, 17, 2041–2054, https://doi.org/10.1007/s10346-020-01420-8, 2020. 

Mason, P. J. and Rosenbaum, M. S.: Geohazard mapping for predicting landslides: An example from the Langhe Hills in Piemonte, NW Italy, Q. J. Eng. Geol. Hydroge., 35, 317–326, https://doi.org/10.1144/1470-9236/00047, 2002. 

Micheletti, N., Lambiel, C., and Lane, S. N.: Investigating decadal-scale geomorphic dynamics in an alpine mountain setting, J. Geophys. Res. F Earth Surf., 120, 2155–2175, https://doi.org/10.1002/2015JF003656, 2015. 

Montgomery, D. R. and Dietrich, W. E.: A physically based model for the topographic control on shallow landsliding, Water Resour. Res., 30, 1153–1171, https://doi.org/10.1029/93WR02979, 1994. 

Montrasio, L., Schilirò, L., and Terrone, A.: Physical and numerical modelling of shallow landslides, Landslides, 13, 873–883, https://doi.org/10.1007/s10346-015-0642-x, 2016. 

Moreno, M., Lombardo, L., Crespi, A., Zellner, P. J., Mair, V., Pittore, M., van Westen, C., and Steger, S.: Space-time data-driven modeling of precipitation-induced shallow landslides in South Tyrol, Italy, Sci. Total Environ., 912, 169166, https://doi.org/10.1016/j.scitotenv.2023.169166, 2024. 

Muñoz Sabater, J.: ERA5-Land hourly data from 1950 to present, Copernicus Climate Change Service (C3S) Climate Data Store (CDS), https://doi.org/10.24381/cds.e2161bac, 2019. 

Muñoz-Sabater, J., Dutra, E., Agustí-Panareda, A., Albergel, C., Arduini, G., Balsamo, G., Boussetta, S., Choulga, M., Harrigan, S., Hersbach, H., Martens, B., Miralles, D. G., Piles, M., Rodríguez-Fernández, N. J., Zsoter, E., Buontempo, C., and Thépaut, J.-N.: ERA5-Land: a state-of-the-art global reanalysis dataset for land applications, Earth Syst. Sci. Data, 13, 4349–4383, https://doi.org/10.5194/essd-13-4349-2021, 2021. 

Nam, K., Kim, J., and Chae, B.: Exploring class imbalance with under-sampling, over-sampling, and hybrid sampling based on Mahalanobis distance for landslide susceptibility assessment: a case study of the 2018 Iburi earthquake induced landslides in Hokkaido, Japan, Geosci. J., 28, 71–94, https://doi.org/10.1007/s12303-023-0033-6, 2024. 

Ozturk, U., Pittore, M., Behling, R., Roessner, S., Andreani, L., and Korup, O.: How robust are landslide susceptibility estimates?, Landslides, 18, 681–695, https://doi.org/10.1007/s10346-020-01485-5, 2021. 

Peirce, C. S.: The numerical measure of the success of predictions, Science, 4, 453–454, https://doi.org/10.1126/science.ns-4.93.453-a, 1884. 

Persichillo, M. G., Bordoni, M., and Meisina, C.: The role of land use changes in the distribution of shallow landslides, Sci. Total Environ., 574, 924–937, https://doi.org/10.1016/j.scitotenv.2016.09.125, 2017. 

Petley, D.: Global patterns of loss of life from landslides, Geology, 40, 927–930, https://doi.org/10.1130/G33217.1, 2012. 

Piana, F., Fioraso, G., Irace, A., Mosca, P., D’Atri, A., Barale, L., Falletti, P., Monegato, G., Morelli, M., Tallone, S., and Vigna, G. B.: Geology of Piemonte region (NW Italy, Alps–Apennines interference zone), J. Maps, 13, 395–405, https://doi.org/10.1080/17445647.2017.1316218, 2017. 

Poggio, L., De Sousa, L. M., Batjes, N. H., Heuvelink, G. B. M., Kempen, B., Ribeiro, E., and Rossiter, D.: SoilGrids 2.0: Producing soil information for the globe with quantified spatial uncertainty, Soil, 7, 217–240, https://doi.org/10.5194/soil-7-217-2021, 2021. 

Provost, F. and Fawcett, T.: Robust classification for imprecise environments, Mach. Learn., 42, 203–231, https://doi.org/10.1023/A:1007601015854, 2001. 

Provost, F., Fawcett, T., and Kohavi, R.: The case against accuracy estimation for comparing induction algorithms, Int. Conf. Mach. Learn., 445, https://dl.acm.org/doi/10.5555/645527.657469 (last access: 24 November 2023), 1998. 

Raghavan, V., Bollmann, P., and Jung, G. S.: A Critical Investigation of Recall and Precision as Measures of Retrieval System Performance, ACM T. Inform. Syst., 7, 205–229, https://doi.org/10.1145/65943.65945, 1989. 

Regione Piemonte: Geoportale Piemonte, https://www.geoportale.piemonte.it, last access: 4 March 2024. 

Reichenbach, P., Rossi, M., Malamud, B. D., Mihir, M., and Guzzetti, F.: A review of statistically-based landslide susceptibility models, Earth-Sci. Rev., 180, 60–91, https://doi.org/10.1016/j.earscirev.2018.03.001, 2018. 

Saito, T. and Rehmsmeier, M.: The precision-recall plot is more informative than the ROC plot when evaluating binary classifiers on imbalanced datasets, PLoS One, 10, 1–21, https://doi.org/10.1371/journal.pone.0118432, 2015. 

Sala, G., Lanfranconi, C., Frattini, P., Rusconi, G., and Crosta, G. B.: Cost-sensitive rainfall thresholds for shallow landslides, Landslides, 18, 2979–2992, https://doi.org/10.1007/s10346-021-01707-4, 2021. 

Samia, J., Temme, A., Bregt, A. K., Wallinga, J., Stuiver, J., Guzzetti, F., Ardizzone, F., and Rossi, M.: Implementing landslide path dependency in landslide susceptibility modelling, Landslides, 15, 2129–2144, https://doi.org/10.1007/s10346-018-1024-y, 2018. 

Segoni, S., Tofani, V., Rosi, A., Catani, F., and Casagli, N.: Combination of rainfall thresholds and susceptibility maps for dynamic landslide hazard assessment at regional scale, Front. Earth Sci., 6, https://doi.org/10.3389/feart.2018.00085, 2018. 

Šilhán, K.: Dendrogeomorphological analysis of landslides on the undercut river terrace bank (a case study in Czech Republic), Landslides, 19, 621–635, https://doi.org/10.1007/s10346-021-01833-z, 2022. 

Smith, H. G., Spiekermann, R., Betts, H., and Neverman, A. J.: Comparing methods of landslide data acquisition and susceptibility modelling: Examples from New Zealand, Geomorphology, 381, 107660, https://doi.org/10.1016/j.geomorph.2021.107660, 2021. 

Smith, H. G., Neverman, A. J., Betts, H., and Spiekermann, R.: The influence of spatial patterns in rainfall on shallow landslides, Geomorphology, 437, 108795, https://doi.org/10.1016/j.geomorph.2023.108795, 2023. 

Steger, S., Brenning, A., Bell, R., and Glade, T.: The propagation of inventory-based positional errors into statistical landslide susceptibility models, Nat. Hazards Earth Syst. Sci., 16, 2729–2745, https://doi.org/10.5194/nhess-16-2729-2016, 2016. 

Thomas, M. A., Mirus, B. B., and Collins, B. D.: Identifying Physics-Based Thresholds for Rainfall-Induced Landsliding, Geophys. Res. Lett., 45, 9651–9661, https://doi.org/10.1029/2018GL079662, 2018. 

Tiranti, D., Nicolò, G., and Gaeta, A. R.: Shallow landslides predisposing and triggering factors in developing a regional early warning system, Landslides, 16, 235–251, https://doi.org/10.1007/s10346-018-1096-8, 2019. 

Trigila, A. and Iadanza, C.: The national landslide inventory, landslide events, impacts and mitigation measures in Italy, in: Landslides and Engineered Slopes. Protecting Society through Improved Un- derstanding vol. 1, edited by: Eberhardt, E., Froese, C., Turner, A.K., and Leroueil, S., CRC Press, Taylor & Francis, Banff, Alberta, Canada, pp. 273–278, 2012. 

Van Den Eeckhaut, M., Poesen, J., Govers, G., Verstraeten, G., and Demoulin, A.: Characteristics of the size distribution of recent and historical landslides in a populated hilly region, Earth Planet. Sc. Lett., 256, 588–603, https://doi.org/10.1016/j.epsl.2007.01.040, 2007. 

van Westen, C. J., Castellanos, E., and Kuriakose, S. L.: Spatial data for landslide susceptibility, hazard, and vulnerability assessment: An overview, Eng. Geol., 102, 112–131, https://doi.org/10.1016/j.enggeo.2008.03.010, 2008. 

Varnes, D. J.: IAEG Commission on Landslides and other Mass-Movements: Landslide Hazard Zonation: A Review of Principles and Practice, The UNESCO Press, Paris, 63 pp., 1984. 

Wu, W. and Sidle, R. C.: A distributed slope stability model for steep forested basins, Water Resour., 31, 2097–2110, 1995. 

Yordanov, V. and Brovelli, M. A.: Comparing model performance metrics for landslide susceptibility mapping, Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLIII-B3-2020, 1277–1284, https://doi.org/10.5194/isprs-archives-XLIII-B3-2020-1277-2020, 2020. 

Yule, G. U.: On the association of attributes in statistics, Philos. T. Roy. Soc. Lond. A, 194, 257–319, https://doi.org/10.1098/rsta.1900.0019, 1900. 

Zhao, Z., He, Y., Yao, S., Yang, W., Wang, W., Zhang, L., and Sun, Q.: A comparative study of different neural network models for landslide susceptibility mapping, Adv. Space Res., 70, 383–401, https://doi.org/10.1016/j.asr.2022.04.055, 2022. 

From life sciences to financial services and IT operations, Akkodis demonstrates how applied AI is driving measurable business outcomes, enabling strategic expansion while helping companies achieve cost savings and sustainable workforce transformation worldwide.

ZURICH, Nov. 10, 2025 /PRNewswire/ — Akkodis, a global leader in digital engineering consulting, today announced a series of successful technology implementations demonstrating its deep expertise and strong demand for AI-enabled transformation across industries. Leveraging its global footprint and technical depth, Akkodis helps businesses become not just AI-capable but future-ready—equipping them with the agility to respond faster and stay ahead in a world of constant change. Through its comprehensive suite of AI and data analytics solutions spanning digital engineering, R&D and IT, Akkodis enables clients to realize tangible, scalable transformation.

“We’re focused on using AI as a practical lever to solve complex problems, elevate quality and empower people to work in new ways,” said Jo Debecker, President and CEO, Akkodis. “These projects show how we bring together human ingenuity and advanced technology to deliver transformation that lasts.”

These examples demonstrate how AI-led innovation powered by Akkodis Intelligence drives real-world impact:

1. AI reduces life sciences production scheduling time from five days to seconds

Akkodis partnered with a global healthcare manufacturer to integrate AI into supply and demand planning, aligning complex forecasts with production of critical equipment. Using advanced combinatorial optimization and a human-in-the-loop approach, the solution delivers rapid, bias-free scheduling recommendations, reducing scheduling time from five days to seconds and enabling enterprise-wide scalability. The next phase will introduce LLM-based agents, allowing managers to express priorities in natural language and further enhance agility, efficiency and decision-making across the organization.

2. U pskilling engineers & data scientists in AI: Supporting responsible AI in banking

In partnership with Microsoft Worldwide Learning, Akkodis Academy created a bespoke AI enablement program for the Commonwealth Bank of Australia featuring customized technical bootcamps, webinars and targeted hands-on training. The program helped teams to rapidly adopt AI tools such as GitHub Copilot, with approximately 30% of AI-generated code accepted, cutting development time and boosting accuracy.

3. Scaling AI and automation through IT: 2,000 employees AI-proficient, 15,000 hours saved

Akkodis Japan launched a program using generative AI and low-code tools to foster a hands-on, field-led approach to digital transformation. The initiative advanced operational excellence through automation and change management—saving over 15,000 hours annually by automating claims submissions and sales operations processes. Within just 10 months, more than 2,000 employees of Akkodis Japan (81% of those focused on internal operations) became proficient in AI tools. This large-scale success now serves as a blueprint for clients pursuing responsible, scalable AI transformation worldwide.

These outcomes underscore Akkodis’ continued dedication to combining advanced technology, domain expertise and human insight to enable transformation across industries. Grounded in Akkodis Intelligence – its commitment to bringing technology and human potential together to drive meaningful, measurable impact – Akkodis will continue to introduce new products and solutions that advance responsible, AI-driven innovation in the months ahead.

“At Akkodis, we deliver AI solutions that are not only powerful but responsible,” said Joshua Morley, Akkodis Group AI Officer. “By uniting deep domain expertise with robust governance and cutting-edge technology, we help clients build the confidence and capability to embed AI responsibly across their organizations, translating ambition into measurable real-world outcomes.”

Media contacts

Anne Friedrich
SVP, Global Head of Communications, Akkodis
E. [email protected]

Lisa Bushka
VP, External Communications, Akkodis
E. [email protected]

About Akkodis 

Akkodis is a global digital engineering consulting company that enables organizations to innovate and accelerate by applying technology to redefine how processes and products are developed, powered and optimized. With deep expertise across AI, data, cloud, edge and software engineering, we combine technology and talent to deliver end-to-end solutions, from strategy and consulting to talent development and implementation. Our commitment to Akkodis Intelligence helps businesses connect the exponential power of technology with the irreplaceable strengths of human thinking and collaboration. Part of the Adecco Group and headquartered in Switzerland, Akkodis brings together 50,000 engineers and tech consultants in over 30 countries with services that span Consulting, Talent, Solutions, and Academy. With a cross-sector view and strong delivery capabilities, Akkodis empowers businesses to solve complex challenges and achieve sustainable impact. akkodis.com | LinkedIn | Instagram | Facebook| X

About the Adecco Group

The Adecco Group is the world’s leading talent company. Our purpose is making the future work for everyone. Through our three global business units – Adecco, Akkodis and LHH – across 60 countries, we enable sustainable and lifelong employability for individuals, deliver digital and engineering solutions to power the Smart Industry transformation and empower organisations to optimise their workforces. The Adecco Group leads by example and is committed to an inclusive culture, fostering sustainable employability, and supporting resilient economies and communities. The Adecco Group AG is headquartered in Zurich, Switzerland (ISIN: CH0012138605) and listed on the SIX Swiss Exchange (ADEN). www.adeccogroup.com

Photo- https://mma.prnewswire.com/media/2817844/Akkodis_Intelligence.jpg
Logo – https://mma.prnewswire.com/media/2598662/5611051/Akkodis_Logo.jpg

SOURCE Akkodis

Contents

Key Takeaways 1

Introduction. 3

Modeling the Economic Impact of Semiconductor Export Controls 4

Impact of Export Controls on U.S. Semiconductor Firms 6

Medium-Term Foregone Growth and Resulting Market 10

Long-Term Foregone Growth and Resulting Market 14

Impact on ICT Industries 17

Conclusion. 19

Appendix: Methodology 20

Endnotes 25

Semiconductors constitute the foundation of the global digital economy, generating $631 billion in global sales in 2024, and is projected to exceed $1 trillion by 2030, while enabling more than $7 trillion in annual downstream economic activity in fields such as artificial intelligence (AI), cloud computing, and advanced manufacturing.^[1] This represents roughly 7 percent of the world’s economic activity.[2]

This report examines a hypothetical scenario in which the United States implements broader export controls and policies that effectively decouple U.S. semiconductor firms from the Chinese market. In reality, the United States has only implemented a limited set of export controls aimed at slowing China’s access to advanced semiconductors and related technologies for military and high-performance computing applications. For example, in October 2022, the Biden administration issued export controls restricting the sale of AI chips to China, along with technology to manufacture those chips.^[3] Following that, in January 2025, the Department of Commerce announced a proposed AI diffusion rule that implemented a three-tiered system for access to advanced AI hardware, capping the number of advanced chips that foreign nations could receive.^[4] However, those proposed rules were rescinded by the Trump administration.[5] On the other hand, the Chinese government could also impose similar restrictions on U.S. firms, resulting in a lost of access to the Chinese market. Indeed, because the Chinese government banned Micron products in critical infrastructure in 2023, the U.S. chip company intends to stop supplying server chips to China altogether.[6]

Even within this limited framework, restrictions carry potential risks. The Information Technology and Innovation Foundation’s (ITIF’s) economic model shows that reducing access to Chinese commercial markets in a decoupling scenario would reduce revenue for U.S. chipmakers, which in turn could lower investment in research and development (R&D), the primary driver of next-generation chips and long-term competitiveness in the industry. Slower revenue growth and innovation would not only weaken the global position of U.S. firms, but could also ripple across the broader economy by affecting high-skill jobs in semiconductor design and manufacturing and downstream employment in industries reliant on chips. In short, reduced revenues and fewer jobs would only end up harming American chip firms and American workers.

This report develops an economic model to estimate the impact of U.S. semiconductor export controls if they were to lead to a full decoupling from the Chinese market. It provides the logic of the economic model followed by a summary of the semiconductor industry in 2024, which serve as the baseline for the model. Subsequently, the report estimates the initial impact of full decoupling on U.S. semiconductor firms’ revenues and R&D investments, before examining the impact on industry jobs and downstream jobs in the U.S. economy. In addition, the report estimates the market share and foregone revenue, as well as R&D investments and job losses, in both the short term and long term, assuming a full decoupling scenario. It then provides summaries of the impact from other scenarios of decoupling, such as 25 percent decoupling. The report concludes by exploring the impact of decoupling on specific information and communications technology (ICT) industry sectors.

ITIF developed an economic model to estimate the impact of U.S. semiconductor export controls on China. This model estimates U.S. semiconductor firms’ revenue, R&D investments, and jobs impact from four export control/decoupling scenarios: 1) full decoupling or export ban (U.S. companies are completely prohibited from selling to China); 2) 50 percent decoupling (U.S. semiconductors restricted selling to China, leading to a 50 percent revenue loss from the Chinese market); 3) 25 percent decoupling (some U.S. semiconductors cannot be sold to China, leading to a 25 percent revenue loss from the Chinese market); and 4) Entity List-based decoupling (U.S. companies cannot sell to a list of specific Chinese companies, leading to a 10 percent revenue loss from the Chinese market). (See the appendix for the full methodology.)

The rationale behind the four decoupling scenarios is as follows. The United States will try to place export controls on the most important semiconductor products first before expanding the list. As such, the Entity List will first cut off some Chinese companies from a few key semiconductor products, leading U.S. firms to lose 10 percent of revenue from the Chinese market. However, gradually, the United States will expand export controls to encompass all semiconductor products, leading to a full decoupling between the two nations. This would mean that U.S. semiconductor firms would lose all their revenue from the Chinese market.

The economic model measures the extent of negative impact U.S. export controls could have on U.S. semiconductor firms’ revenue and share of the global industry. Assuming that the United States does not utilize the Foreign Direct Product Rule, U.S. firms’ losses could be gained by other nations, including China, the EU, Japan, Taiwan, South Korea, and others. Using the Boston Consulting Group (BCG) and the Semiconductor Industry Association’s (SIA’s) “Emerging Resilience in the Semiconductor Supply Chain” report, ITIF estimated the global market shares of each nation for each semiconductor product in 2022.[7] (See figure 1.)

Figure 1: Global market share of each semiconductor product type[8]

Using this as a proxy for future global shares by nation, ITIF estimated the extent to which other nations could benefit from the lost U.S. revenue by removing the United States and scaling the remaining nations’ shares.

As a result of U.S. firms’ revenue losses, their investment in R&D would also be reduced. As SIA found, U.S. semiconductor companies invested 17.7 percent of their revenue into R&D in 2024.[9]

Furthermore, the revenue losses would also mean that U.S. semiconductor companies may be unable to offer as many jobs, leading to a decline in the industry’s employment numbers. This decline would subsequently result in fewer downstream jobs being created in the U.S. economy. Indeed, as SIA found, every job in the semiconductor industry supports an additional 5.7 downstream jobs in the U.S. economy.[10]

In sum, revenue losses would result in fewer jobs, reduced R&D investment, and a lower global market share for U.S. chipmakers. Figure 2 illustrates the analytical framework ITIF’s model uses to estimate the economic impact of U.S. semiconductor export controls on U.S. semiconductor firms and the U.S. economy.

Figure 2: ITIF’s analytical framework for modeling the effects of U.S. semiconductor export controls on China

The following sections provide background information on the global semiconductor market, the U.S. semiconductor market, China’s semiconductor market, and U.S. semiconductor firms’ share of the Chinese market in 2024. Afterward, they examine the initial impact of a full decoupling with China on U.S. firms’ revenue and other nations’ firms’ revenue. In addition, the section provides a summary of the impact of the three other decoupling scenarios: 50 percent decoupling, 25 percent decoupling, and Entity List-based export controls. The report later examines the extent to which each decoupling scenario would impact U.S. semiconductor firms’ R&D investment levels, before turning to the impact on jobs.

In 2024, the global semiconductor industry generated $630.5 billion in revenue, of which U.S. semiconductor firms accounted for 50.4 percent.[11] Meanwhile, Chinese semiconductor firms accounted for 4.5 percent, or $28.4 billion.[12] China demands more chips than it produces. In 2024, China demanded 24 percent, or $151.3 billion, of the global semiconductor market.[13] Given U.S. semiconductor firms’ large market share, they supplied 50.7 percent of China’s demand in 2024.[14] (See table 1.)

Table 1: Semiconductor market measures, 2024[15]

Measure	Value (Billions)	Share
Global semiconductor market	$631	100.0%
Chinese firms’ global market share	$28	4.5%
China’s demand of global market	$151	24.0%
U.S. firms’ global market share	$318	50.4%
U.S. firms’ share of China market	$77	50.7%

China’s demand for semiconductors varies by type. For example, China’s computing and information technology (IT) industries may require logic chips such as central processing units, while a phone maker would need more memory chips and an electric vehicle (EV) manufacturer more sensors and microcontrollers. As such, using the SIA 2025 Factbook’s share of global demand for each semiconductor product type as a proxy, ITIF estimated China’s demand for each product type in 2024.[16] The top three semiconductor product types most demanded include logic chips (34 percent), memory chips (26 percent), and analog chips (13 percent). (See figure 3.)

Figure 3: China’s chip demand by semiconductor product types[17]

Since U.S. firms fulfill a significant share of China’s demand, export controls or full decoupling could have a severe impact on U.S. semiconductor firms’ revenues. They would eliminate, or at least partially reduce, U.S. firms’ ability to sell to China, thereby reducing U.S. firms’ overall revenue. In this section, ITIF uses 2024 as a baseline to estimate the revenue U.S. firms could lose and other nations’ firms could gain in the initial year after decoupling with China.

Table 3 shows the potential revenue losses of U.S. firms by semiconductor product type after a full decoupling from China and the nations that could gain from U.S. firms’ losses. In the event of a complete decoupling, U.S. firms would no longer sell to China, meaning they would lose 100 percent of their revenue from China. In this scenario, U.S. firms could lose about $77 billion from complete decoupling with China. Meanwhile, mainland Chinese firms could gain about $9 billion, EU firms $15 billion, Japanese firms $12 billion, Taiwanese firms $14 billion, South Korean firms $21 billion, and other nations’ firms $5 billion of those losses.

Table 2: Foreign firms’ potential gains from U.S. semiconductor firms’ losses after full decoupling (billions)[18]

Product Type	United States	Mainland China	Japan	Taiwan	EU	South Korea	Others
Logic	-$26.3	$3.8	$3.0	$8.3	$6.8	$2.3	$2.3
Memory	-$20.1	$0.8	$1.9	$1.1	$0.1	$16.1	$0.1
Analog	-$9.7	$1.5	$3.0	$0.8	$2.8	$0.7	$1.0
MPU	-$6.6	$0.9	$0.8	$2.1	$1.7	$0.6	$0.6
Opto	-$5.0	$0.8	$1.5	$0.4	$1.4	$0.3	$0.5
Discretes	-$3.8	$0.6	$1.2	$0.3	$1.1	$0.3	$0.4
MCU	-$2.7	$0.4	$0.3	$0.8	$0.7	$0.2	$0.2
Sensor	-$2.3	$0.4	$0.7	$0.2	$0.7	$0.2	$0.2
DSP	-$0.3	$0.0	$0.0	$0.1	$0.1	$0.0	$0.0
Total	-$76.7	$9.1	$12.3	$14.1	$15.3	$20.6	$5.3

Any form of decoupling could exert negative impacts on U.S. firms’ revenues while benefitting other nations. U.S. firms could lose as much as $38.4 billion in revenue in the face of 50 percent decoupling with China. Of this, mainland Chinese firms could gain about $5 billion, South Korean firms $10 billion, and Taiwanese firms $7 billion. Under 25 percent decoupling, U.S. firms could lose about $19 billion while mainland Chinese firms could gain about $2 billion, Japanese ones $3 billion, and Taiwanese firms $4 billion. Lastly, under an Entity Listing against Chinese companies, U.S. firms could lose about $8 billion while mainland Chinese firms gain $1 billion, the EU’s firms $2 billion, and South Korea’s firms $2 billion. (See table 3.)

Table 3: Foreign semiconductor firms’ potential gains from U.S. firms’ losses (billions)[19]

Decoupling Scenarios	United States	Mainland China	Japan	Taiwan	EU	South Korea	Others
Full decoupling	-$76.7	$9.1	$12.3	$14.1	$15.3	$20.6	$5.3
50% decoupling	-$38.4	$4.5	$6.2	$7.0	$7.7	$10.3	$2.7
25% decoupling	-$19.2	$2.3	$3.1	$3.5	$3.8	$5.1	$1.3
Entity List	-$7.7	$0.9	$1.2	$1.4	$1.5	$2.1	$0.5

Impact on U.S. Semiconductor Firms’ R&D Spending

The U.S. semiconductor industry invests a significant share of its revenue in R&D to develop the next generation of globally competitive chips. In 2024 alone, the industry invested 17.7 percent of its revenue in R&D, and, on average, invested 18.3 percent from 2015 to 2024. As such, given that U.S. semiconductor firms’ revenues would decline from decoupling with China, R&D investments would also decline. Generally, the share of revenue invested in R&D would also decline when revenue declines, but ITIF kept the share constant for a more conservative estimate.

The U.S. semiconductor industry invested $56.2 billion in R&D in 2024. However, if the United States fully decouples from China, U.S. firms’ R&D investments could decrease to about $43 billion, roughly $14 billion lower than the status quo. This is calculated by multiplying the industry’s R&D intensity by the lost revenues from precluded sales to China. Meanwhile, U.S. firms would likely reduce R&D investments to about $49 billion, or about $7 billion lower than the status quo, under 50 percent decoupling. Further, U.S. firms could reduce R&D investments to about $53 billion (about $3 billion lower than status quo) and $55 billion (about $1 billion lower) from 25 percent decoupling or export control entity listing, respectively. (See figure 4.)

Figure 4: U.S. semiconductor firms’ future investments in R&D, by decoupling scenario[20]

The robust U.S. semiconductor industry provides hundreds of thousands of jobs to the U.S. economy while simultaneously supporting downstream job creation in other sectors. For instance, a semiconductor process engineer position is one job in the economy, but it also indirectly supports logistics providers who ship the chips the engineer helps produce. Moreover, when the engineer spends their money in the economy, this further supports other jobs. For example, engineers support restaurant employees when they spend money at a restaurant. Indeed, as SIA’s 2025 Factbook highlights, a semiconductor industry job supports 5.7 additional downstream jobs in the U.S. economy. In 2024, the U.S. semiconductor industry supported approximately 345,000 direct jobs and indirectly supported nearly 2 million additional ones.

However, when U.S. semiconductor firms’ revenues decline, they can no longer support as many industry jobs, leading to both direct and indirect job losses. Indeed, the U.S. semiconductor industry could support over 80,000 fewer industry jobs and almost 500,000 fewer downstream jobs if the United States fully decoupled from China. Meanwhile, the U.S. industry could support over 40,000 fewer industry jobs and more than 200,000 fewer downstream jobs in a scenario of 50 percent decoupling. A 25 percent decoupling could lead to about 20,000 fewer industry jobs and over 100,000 fewer downstream jobs. Finally, an export control entity listing could lead to over 8,000 fewer industry jobs and almost 50,000 fewer downstream jobs. (See figure 5.) In sum, significant export controls or decoupling with China could negatively impact U.S. households from decreased employment opportunities in well-paying high-tech jobs.

Figure 5: U.S. semiconductor and additional job losses by decoupling scenario[21]

The following section first examines the U.S. and Chinese global semiconductor market shares and revenues five years after a one-time full decoupling scenario. Afterward, it quantifies how a one-time full decoupling would negatively impact R&D investments in subsequent years, before examining the foregone jobs.

If normal growth trajectories had prevailed (e.g., expected growth of China’s semiconductor market and stable U.S. market share) then after five years under a normal scenario, U.S. semiconductor companies could have been expected to earn about $84 billion from China’s market. With those revenues now lost in this full decoupling scenario, the sales would likely be accrued by other nations’ semiconductor firms. Chinese firms could gain about $10 billion, EU firms $17 billion, South Korean firms $22 billion, Japanese firms $13 billion, Taiwanese firms $15 billion, and other nations’ firms $6 billion. (See table 4.)

Table 4: Foreign firms’ potential gains from U.S. firms’ losses five years after full decoupling (billions)[22]

Product Type	United States	Mainland China	Japan	Taiwan	EU	South Korea	Others
Logic	-$28.6	$4.1	$3.3	$9.0	$7.4	$2.5	$2.5
Memory	-$21.9	$0.9	$2.0	$1.2	$0.1	$17.5	$0.1
Analog	-$10.6	$1.6	$3.2	$0.9	$3.0	$0.7	$1.1
MPU	-$7.2	$1.0	$0.8	$2.3	$1.8	$0.6	$0.6
Opto	-$5.4	$0.8	$1.7	$0.5	$1.6	$0.4	$0.6
Discretes	-$4.1	$0.6	$1.3	$0.3	$1.2	$0.3	$0.4
MCU	-$2.9	$0.4	$0.3	$0.9	$0.7	$0.2	$0.2
Sensor	-$2.5	$0.4	$0.8	$0.2	$0.7	$0.2	$0.3
DSP	-$0.4	$0.1	$0.0	$0.1	$0.1	$0.0	$0.0
Total	-$83.6	$9.9	$13.4	$15.4	$16.7	$22.4	$5.8

As a result of the revenue reduction, U.S. firms’ global market share would be lower than without decoupling after five years. Without decoupling, the U.S. semiconductor industry’s global revenue could be about $396 billion, or a market share of 48 percent. In comparison, with full decoupling, the U.S. industry’s market share would likely fall to about 38 percent, equating to about $313 billion in revenue. (See figure 6.)

Figure 6: U.S. semiconductor firms’ global market share five years after full decoupling versus the status quo[23]

In contrast, Chinese semiconductor firms could gain some of U.S. firms’ lost revenue and grow their overall global revenue and market share. Without decoupling, Chinese firms’ global revenue could be about $52 billion, or about 6.3 percent. However, with full decoupling, Chinese firms’ global revenue could grow to about $62 billion (about 7 percent) after five years, approximately $10 billion higher than without decoupling. (See figure 7.)

Figure 7: Chinese semiconductor firms’ global market share five years after full decoupling[24]

Due to the lower-than-expected revenue, U.S. firms’ R&D investments after five years would be smaller than with no decoupling. Without decoupling, U.S. firms would likely invest about $72 billion in R&D investments. Yet, U.S. firms could reduce their investments to about $57 billion under full decoupling, leading to $15 billion in foregone R&D investments. (See figure 8.)

Figure 8: U.S. semiconductor firms’ investment in R&D five years after full decoupling versus the status quo[25]

U.S. semiconductor firms’ lower-than-expected revenue also means that they would be able to support fewer industry and downstream jobs in the economy. Under the status quo scenario, the U.S. semiconductor industry could support about 400,000 industry jobs and 2.3 million downstream jobs. However, under full decoupling, U.S. firms could only support about 300,000 industry jobs and 1.8 million downstream jobs, leading to almost 100,000 foregone industry jobs and over 550,000 foregone downstream jobs. (See figure 9.)

Figure 9: U.S. semiconductor and related jobs five years after full decoupling versus the status quo[26]

In the medium term, U.S. firms would continue to forego revenue, R&D investments, and jobs in the event of any of the contemplated decoupling scenarios. Under 50 percent decoupling, U.S. firms could forego about $42 billion in revenue—resulting in a global market share of about 43 percent compared with 48 percent under the status quo—about $8 billion in R&D investments, about 50,000 industry jobs, and almost 300,000 downstream jobs after five years. Chinese firms could gain some of the U.S. firms’ losses, resulting in about a 7 percent global market share rather than 6 percent under the status quo. Under a 25 percent decoupling, U.S. firms would forego about $21 billion in revenue, $4 billion in R&D investments, almost 25,000 industry jobs and about 140,000 downstream jobs. U.S. firms’ global market share could fall to about 45 percent, while Chinese firms’ share could rise to about 7 percent. Under an Entity List scenario, U.S. firms could still forego about $8 billion in revenue, $2 billion in R&D, almost 10,000 industry jobs, and over 56,000 downstream jobs—U.S. firms’ global market share could fall to 47 percent while Chinese firms’ share could rise to 6 percent. (See table 5.)

Table 5: Summary of decoupling impact after five years[27]

Impact	Entity List	25% Decoupling	50% Decoupling	Full Decoupling
U.S. firms’ share of the Chinese market	28.3%	23.6%	15.7%	0.0%
U.S. firms’ global market share	46.9%	45.4%	42.9%	37.8%
Chinese firms’ global market share	6.4%	6.6%	6.9%	7.5%
U.S. firms’ revenue foregone	$8.4B	$20.9B	$41.8B	$83.6B
U.S. firms’ R&D spending foregone	$1.5B	$3.8B	$7.6B	$15.3B
U.S. industry jobs foregone	9,888	24,720	49,440	98,881
U.S. downstream jobs foregone	56,362	140,905	281,810	563,620

Similar to the previous section, this one examines the U.S. and Chinese global semiconductor market shares and revenue 10 years after a one-time full decoupling with China. Then, it examines the impact on U.S. semiconductor firms’ share of and revenue from the Chinese market, U.S. firms’ R&D investments, and U.S. industry and downstream jobs.

Over the long run, the missing supply of semiconductors otherwise provided by U.S. firms to the Chinese market would be gained by foreign competitors, especially by China’s domestic semiconductor industry. As a result, these countries’ firms would have larger revenues to reinvest in R&D, accelerating their innovation capacity. Moreover, Chinese and other nations’ firms could steadily close the technological gap with the United States. In contrast, U.S. firms, constrained by smaller markets and diminished R&D capacity, could risk falling behind. This erosion of leadership could undermine the U.S. semiconductor industry’s central role in sustaining national competitiveness and technological primacy.

ITIF estimates that, if current growth trends had prevailed and there had been no decoupling, U.S. semiconductor companies could have likely earned about $91 billion in revenues from the Chinese market 10 years from now. Again, those revenues are lost in this full decoupling scenario and instead would be accrued by firms from other nations. Here, ITIF estimates that South Korean firms could gain about $24 billion, EU firms $18 billion, and mainland Chinese firms $11 billion. Japan, Taiwan, and other nations could gain the remainder of the foregone revenue. (See table 6.)

Table 6: Other nations’ gains and U.S. semiconductor firms’ foregone revenue by semiconductor product type 10 years after full decoupling (billions)[28]

Under the status quo, U.S. firms could maintain their global market share of 48 percent, which would result in about $520 billion in global revenues 10 years on. In contrast, under full decoupling, U.S. firms’ market share would likely drop to about 40 percent, resulting in about $429 billion of global revenue. (See figure 10.)

Figure 10: U.S. semiconductor firms’ global market share 10 years after full decoupling versus the status quo[29]

In comparison, Chinese semiconductor firms’ global market share and revenue would be higher than no decoupling, as they would gain some of the U.S. firms’ lost revenue. Without decoupling, Chinese firms’ share could be about 6 percent, equating to a revenue of about $68 billion. Yet, with full decoupling, Chinese firms’ revenue could rise to $79 billion, leading to about a 7 percent global market share and about $11 billion more than if the United States does not fully decouple from China. (See figure 11.)

Figure 11: Chinese semiconductor firms’ global market share after full decoupling versus the status quo[30]

As a result of the foregone revenue, U.S. firms would invest less in R&D compared with no decoupling. Under the status quo scenario, U.S. firms could invest about $95 billion in R&D in the 10th year. In comparison, under full decoupling, U.S. firms could invest only about $78 billion, resulting in about $17 billion of foregone R&D investments. (See figure 12.)

Figure 12: U.S. semiconductor firms’ investment in R&D 10 years after full decoupling versus the status quo[31]

Moreover, U.S. semiconductor firms would support fewer jobs. Under a status quo growth scenario, the U.S. semiconductor industry could likely support about 500,000 industry jobs and 2.8 million downstream jobs in the U.S. economy in 10 years’ time. But in a full decoupling scenario, after 10 years, U.S. firms would only be able to support about 370,000 industry jobs and 2.1 million downstream jobs, foregoing almost 120,000 industry jobs and over 650,000 downstream jobs.(See figure 13.)

Figure 13: U.S. semiconductor jobs and additional jobs 10 years after full decoupling versus the status quo[32]

In the long term, regardless of the extent of decoupling, U.S. firms could lose more than they gain. Accordingly, under 50 percent decoupling, U.S. firms could forego about $46 billion in revenue, $8 billion in R&D investments, almost 60,000 industry jobs, and over 300,000 downstream jobs 10 years after the initial decoupling. As a result, U.S. firms’ market share could be about 44 percent compared to about 48 percent without decoupling, while Chinese firms’ share could be about 7 percent compared with 6 percent without decoupling. Under 25 percent decoupling, U.S. firms could forego about $23 billion in revenue, $4 billion in R&D investments, almost 30,000 industry jobs, and almost 170,000 downstream jobs. Under an Entity List scenario, U.S. firms could still forego about $9 billion in revenue, $2 billion in R&D investments, over 10,000 industry jobs, and almost 70,000 downstream jobs. (See table 7.)

Table 7: Summary of decoupling impact after 10 years[33]

Decoupling Scenarios	Entity List	25% Decoupling	50% Decoupling	Full Decoupling
U.S. firms’ share of the Chinese market	23.5%	19.6%	13.1%	0.0%
U.S. firms’ global market share	47.1%	45.8%	43.7%	39.5%
Chinese firms’ global market share	6.4%	6.5%	6.8%	7.3%
U.S. firms’ revenue foregone	$9.1B	$22.8B	$45.5B	$91.1B
U.S. firms’ R&D spending foregone	$1.7B	$4.2B	$8.3B	$16.6B
U.S. industry jobs foregone	11,739	29,346	58,693	117,386
U.S. downstream jobs foregone	66,910	167,275	334,549	669,099

The U.S. ICT sector is deeply reliant on the continuous advancement of semiconductor technology. Advanced chips represent the backbone of modern ICT infrastructure, powering everything from data centers and cloud computing platforms to AI systems and high-performance computing applications. Indeed, as an ITIF report estimates, an average data center relies on approximately 340,000 semiconductors to function.[34] These chips enable efficiency in processing speed, energy efficiency, and data handling capabilities, which are essential for the evolving demands of the digital economy. However, in the consistently evolving ICT sector, which requires greater efficiency every day, the U.S. semiconductor industry needs to consistently innovate and develop the next generation of chips in order to keep up with customer demand.

As this report explains, the U.S. semiconductor industry’s ability to innovate could be hindered due to export controls or decoupling with China by depriving the industry of the revenues necessary to finance the R&D needed to produce the next generation of innovative chips. Corroborating this, a BCG study estimates that a technology decoupling between the U.S. semiconductor industry and China would lead to a decline of $12 billion, or 30 percent, in R&D investment.[35]

While other countries’ semiconductor firms could eventually supply advanced chips, it would take significant time for them to catch up to the already sophisticated designs and capabilities of U.S. firms.[36] During that period, U.S. industries would face higher costs, limited access to cutting-edge technology, and reduced competitiveness.

Below are five examples of ICT industries that would be negatively affected by decreasing R&D in the semiconductor industry:

1. Cloud computing and data centers. Cloud providers such as Amazon Web Services, Microsoft Azure, and Google Cloud rely on high-performance chips to run large-scale data centers efficiently. Less-efficient semiconductors increase energy costs and reduce processing speed, limiting the ability to offer faster, cheaper, and more scalable cloud services.

2. AI and machine learning. AI systems, including natural language processing and computer vision, require advanced GPUs and AI accelerators. Slower or less-efficient chips reduce the speed of model training and inference, thereby constraining innovation and the adoption of AI tools across various industries.

3. Telecommunications and 5G networks. Next-generation 5G infrastructure depends on powerful chips to handle massive amounts of data with low latency. Reduced chip performance can slow network deployment, limit bandwidth, and degrade service quality for consumers and enterprises.

4. Enterprise software and high-performance computing (HPC). Applications such as financial modeling, scientific simulations, and big data analytics rely on fast, energy-efficient processors. Less-capable semiconductors increase computation time and costs, lowering productivity and innovation potential for these industries.

5. Consumer electronics and smart devices. Smartphones, laptops, smart home devices, and wearables depend on efficient semiconductors for battery life, speed, and advanced features. Reduced R&D in next-generation chips could delay product innovation, increase costs, and make U.S. products less competitive globally.

The U.S. ICT sector’s reliance on advanced semiconductor technology necessitates continuous investment in R&D. Export controls and other measures that restrict market access could impede the development of next-generation chips, potentially hindering the growth and competitiveness of this critical industry. Policymakers should consider the long-term implications of such policies on semiconductor innovation and the broader digital economy. Although national security concerns are very important, strategic approaches that balance these concerns with the need for technological advancement are essential to sustain the U.S. ICT sector’s leadership in the global market.

U.S. semiconductor export controls or broader decoupling from China would weaken the U.S. semiconductor industry and the wider economy. Limiting market access would reduce revenue for U.S. chipmakers, which would directly constrain their ability to fund R&D. Reduced R&D activity would hinder the creation of next-generation semiconductors, technologies that power critical ICT sectors such as cloud computing, AI, telecommunications, and high-performance computing. Reduced innovation in semiconductors cascades through these downstream industries, slowing technological advancement and weakening the global competitiveness of U.S. ICT firms.

Lower revenue in the semiconductor industry would also force U.S. semiconductor companies to cut jobs, as they would be unable to maintain the same workforce without sufficient financial resources. Since each semiconductor job supports additional employment across the economy, the contraction in industry jobs would also reduce broader U.S. employment and economic activity. The U.S. semiconductor industry needs to continue to maintain its share of the Chinese semiconductor market in order to sustain ongoing R&D investment, which helps it maintain leadership in the industry and enables the United States to maintain its leadership in the digital economy. As such, U.S. policymakers should keep semiconductor export controls to a minimum in order to enable U.S. semiconductor firms to obtain high levels of revenue to invest in R&D.

To estimate the impact of export controls or a decoupling with China on U.S. semiconductor companies, ITIF first obtained the revenue of the global semiconductor market in 2024; China’s aggregate global demand and its market share; U.S. semiconductor firms’ share of and revenue from the Chinese market; U.S. semiconductor firms’ global market share and revenue; and Chinese semiconductor firms’ global share and revenue for our baseline assumptions. (See table 8.) These values come from SIA’s publications.

Table 8: Economic model’s key variables and descriptions

Variable	Description
Global semiconductor market	Size of the global semiconductor market
China’s demand of global market	China’s portion of global demand
U.S. firms’ share of China market	Estimated share of China’s market held by U.S. firms
U.S. firms’ global market share	U.S. semiconductor firms’ global market share
Chinese firms’ global market share	Chinese semiconductor firms’ global market share

Using SIA’s 2025 Factbook, ITIF then obtained global semiconductor sales by type of product (e.g., logic, memory, etc.).[37] We then estimated the total global semiconductor market for each type of semiconductor product. Assuming that the Chinese market demands the same share of each type of semiconductor product, China’s demand for each type of product in the global semiconductor market was estimated.

Using U.S. semiconductor firms’ share of China’s market, ITIF also estimated U.S. firms’ revenue from the Chinese market for each type of product. These estimates were done for four different decoupling or export control scenarios: full decoupling (100 percent reduction in Chinese market access to U.S. semi), 50 percent decoupling (50 percent reduction), 25 percent decoupling (25 percent reduction), and Entity List (assuming Entity List only reduces U.S. semiconductor revenue from the Chinese market by 10 percent). (See table 9.) To simplify the analysis, ITIF assumed that U.S. firms have the same proportion of the Chinese market for all product types (though, obviously, this varies).

Table 9: Economic model’s decoupling scenarios and U.S. firms’ share loss of Chinese market

Next, ITIF used data from BCG and SIA’s “Emerging Resilience in the Semiconductor Supply Chain” report to obtain the global semiconductor market share that China, the EU, Japan, Taiwan, South Korea, the United States, and others hold for each semiconductor product type.[38] Logic semiconductor product types include general purpose logic chips, microprocessor unit chips (MPUs), microcontroller unit chips (MCUs), and digital signal processor chips (DSPs). Memory semiconductor product types only include memory chips. Discrete, analog, and optoelectronics (DAO) semiconductor product types include analog chips, optoelectronic (opto), discretes, and sensors. (See table 10.) With these observations, ITIF reduced the United States’ share and scaled the remaining nation’s share proportionally to 100 percent. In other words, the remaining nations are assumed to have taken over the United States’ shares based on their starting proportions (not including the United States).

Table 10: Semiconductor product categories and product types for each category

Logic	Memory	DAO
▪ General purpose logic chips ▪ Microprocessor units ▪ Microcontroller units ▪ Digital signal processor chips	▪ Memory chips	▪ Analog chips ▪ Optoelectronic chips ▪ Discretes ▪ Sensors

Using the scaled shares of the remaining nations (not including the United States) and the U.S. firms’ revenue from China’s market for each type of product, ITIF obtained the redistribution of the U.S. firms’ revenue from China’s market to these remaining nations for each product type. This process was completed for each decoupling scenario.

To estimate the impact of U.S. semiconductor firms’ revenue losses on R&D investments for the four decoupling scenarios, ITIF first obtained the share of sales invested in R&D for U.S. firms and their corresponding value in 2024.

Then, for each decoupling scenario, ITIF subtracted the U.S. semiconductor firms’ revenue loss from the 2024 U.S. semiconductor industry’s revenue before taking the share of sales invested in R&D for U.S. firms. We then subtracted this value from U.S. semiconductor firms’ original R&D spending in 2024 to obtain the potential foregone R&D investment.

To estimate the impact of U.S. semiconductor firms’ revenue losses on the semiconductor industry jobs, ITIF first obtained the total number of semiconductor industry employees in 2024. Then U.S. firms’ revenue per employee was used to estimate job losses. Since it was not possible to determine the elasticity of demand for semiconductor firm’s employees, this model assumes a linear relationship between chipmakers’ employment demand and revenues.

ITIF first obtained the U.S. semiconductor firms’ revenue per employee by dividing the U.S. semiconductor firms’ global revenue by the number of employees. Next, the U.S. semiconductor firms’ revenue loss for each decoupling scenario was divided by the U.S. semiconductor firms’ revenue per employee to obtain the number of employees foregone from the decline in revenue.

To assess the impact of U.S. semiconductor firms’ revenue loss on non-semiconductor industry jobs, ITIF first determined the semiconductor job multiplier in 2024, which was estimated to be 5.7.[39] This means every job in the semiconductor industry generated 5.7 additional jobs in the U.S. economy. Next, the number of foregone semiconductor jobs was multiplied by 5.7 to determine the number of foregone downstream jobs.

U.S. Semiconductor Firms’ Revenue Foregone

To estimate U.S. semiconductor firms’ revenue foregone from decoupling with China after 5 and 10 years, ITIF first estimated the average annual growth rate of U.S. semiconductor firms’ revenue from the Chinese market from 2020 to 2024.

Then, for each decoupling scenario, ITIF took U.S. semiconductor firms’ losses for each semiconductor product and multiplied it by (1 + the average annual growth rate) raised to the power of 5 or 10 (years) to determine the U.S. semiconductor firms’ foregone revenues 5 and 10 years after decoupling.

Moreover, the U.S. semiconductor firms’ losses that were shifted to other nations were also multiplied by the average annual growth rate to estimate how much each nation gained after 5 and 10 years from U.S. semiconductor firms’ losses from decoupling in the base year. This was done for each decoupling scenario and each semiconductor type.

R&D Spending Foregone

To determine U.S. semiconductor firms’ foregone R&D investments, ITIF first estimated the average annual share of sales invested in R&D (i.e., R&D intensity) by U.S. semiconductor firms from 2015 to 2024. The rate was then multiplied by U.S. semiconductor firms’ foregone revenue after 5 and 10 years for each decoupling scenario.

Semiconductor and Additional Jobs Foregone

To estimate the foregone semiconductor jobs from decoupling in the base year, ITIF first estimated the average annual growth rate from 2015 to 2024 for U.S. semiconductor jobs. Then the number of jobs lost after decoupling in the base year was multiplied by (1 + the average annual growth rate). This was done for each decoupling scenario.

To determine the additional jobs in the economy foregone, ITIF multiplied these values by the job multiplier (5.7). This was done for each decoupling scenario.

New Overall Market Shares After Decoupling

To estimate the impact of decoupling after 5 and 10 years, ITIF first obtained the average annual growth rate for the global semiconductor market from 2005 to 2024 and U.S. firms’ revenue from China’s market from 2020 to 2024. ITIF also obtained the average China market demand of the global market, U.S. firms’ global market share, and Chinese firms’ global market share from 2015 to 2024.

Using the growth rates, ITIF first estimated the new global semiconductor market after 5 and 10 years. It was then assumed that China’s demand of the global market would continue to be the average of its last 10 years. Next, the new demand was multiplied by the new global semiconductor market value to determine the revenue from China’s demand of the global market.

Then, ITIF estimated U.S. semiconductor firms’ share of and revenue from China’s market for each decoupling scenario after 5 and 10 years. To do so, U.S. semiconductor firms’ losses for each decoupling scenario were first subtracted from the U.S. semiconductor firms’ revenue from the China market in Year 0 before decoupling to determine the remaining Chinese market that U.S. semiconductor firms retained after decoupling in Year 0. That value was then multiplied by (1 + the annual average growth rate of U.S. firms’ revenue from the China market) for 5 or 10 years after decoupling to determine the new revenue from the Chinese market for U.S. firms. These values were divided by China’s demand of the global market to determine U.S. semiconductor firms’ share of China’s market.

Next, the U.S. semiconductor firms’ global market share was estimated. To do so, the average U.S. firms’ market share in the last 10 years was first multiplied by the global semiconductor market to determine U.S. firms’ new global revenue. With that value, ITIF subtracted U.S. semiconductor firms’ losses for each decoupling scenario in Year 0 multiplied by (1 + the average annual growth rate of U.S. firms’ revenue from the China market) raised to 5 or 10 (years). These values were divided by the total global semiconductor market to determine U.S. semiconductor firms’ market share.

ITIF also estimated China’s market share 5 and 10 years after decoupling. To calculate this, the average Chinese firm’s market share over the last 10 years was first multiplied by the global semiconductor revenue to determine the new global revenue of Chinese firms. With that value, ITIF added the U.S. semiconductor firms’ losses that were gained by Chinese firms for each decoupling scenario in Year 0 multiplied by (1 + the average annual growth rate of U.S. firms’ revenue from the China market) raised to 5 or 10 (years). Then, these values were divided by the total global semiconductor market to determine Chinese firms’ market share.

R&D Spending After Decoupling

To estimate the impact of decoupling with China on U.S. semiconductor firms’ R&D spending, ITIF first estimated the average annual share of sales spent on R&D by U.S. semiconductor firms from 2015 to 2024. That average was then multiplied by the new U.S. global market share to determine U.S. firms’ new R&D amount 5 and 10 years after an initial decoupling. This was done for each decoupling scenario.

Semiconductor Jobs

To estimate the impact of decoupling with China on U.S. semiconductor industry jobs, ITIF first estimated the average annual growth rate from 2015 to 2024 for U.S. semiconductor firms’ total employees.

Next, the total number of U.S. semiconductor firms’ employees in year 0 after decoupling was estimated by subtracting U.S. semiconductor firms’ job losses from each decoupling scenario from the U.S. semiconductor firms’ total employees before decoupling in year 0. These values were then multiplied by (1 + the average annual growth rate) raised to the power of 5 or 10 (years) to determine the new number of semiconductor industry employees 5 and 10 years after an initial decoupling.

Additional Jobs

To estimate the additional jobs created 5 and 10 years after an initial decoupling, ITIF first assumed the job multiplier did not change. Then, the new number of semiconductor industry employees 5 and 10 years after decoupling was multiplied by the job multiplier for each decoupling scenario.

Apple has removed two of the most popular gay dating apps in China from the App Store after receiving an order from China’s main internet regulator and censorship authority, WIRED has learned. The move comes as reports of Blued and Finka disappearing from the iOS App Store and several Android app stores circulated on Chinese social media over the weekend. The apps appear to still be functional for users in the country who already have them downloaded.

“We follow the laws in the countries where we operate. Based on an order from the Cyberspace Administration of China, we have removed these two apps from the China storefront only,” an Apple spokesperson said in an email. Apple clarified that the apps have not been available in other countries for some time. “Earlier this year, the developer of Finka elected to remove the app from storefronts outside of China, and Blued was available only in China.”

Most international LGBTQ+ dating apps are already blocked in China. Grindr was removed from Apple’s Chinese App Store in 2022.

China decriminalized homosexuality in the 1990s, but the government does not recognize same-sex marriage. In recent years, China’s LGBTQ+ community has increasingly come under pressure as the Chinese Communist Party tightens its control over civil society and free expression. Several prominent gay rights organizations in China have shut down, and social media companies now frequently censor LGBTQ+ content and accounts.

The Chinese embassy in Washington, DC did not immediately respond to a request for comment.

In July, Blued abruptly stopped new user registration without giving an explanation, according to Chinese social media posts. For a month, Chinese users who wanted to get on the platform were paying as much as $20 for secondhand Blued accounts on ecommerce websites. But registration resumed in mid-August.

In 2020, BlueCity, the parent company of Blued, went public. It announced that the app had over 49 million registered users and over 6 million monthly active users. The same year, BlueCity said it was acquiring Finka, its main competitor in China, for about $33 million. The company delisted in 2022 and was acquired by Newborn Town, a Hong Kong-listed social media firm. Most of the longtime employees of Blued, including its founder Ma Baoli, left the company after the acquisition, says a former Blued employee who asked not to be named for privacy reasons.

By Megumi Fujikawa

  Bank of Japan policy board member Junko Nakagawa reaffirmed the bank's stance of seeking further interest-rate hikes but also stressed the necessity of moving carefully given lingering uncertainties. 

  "If its outlook for economic activity and prices is realized, the bank will continue to raise the policy interest rate and adjust the degree of monetary accommodation," Nakagawa said in a speech Monday. 

  She said medium- to long-term inflation expectations are moderately rising toward the bank's target of 2%. "The rise in such expectations over the past few years has been more pronounced than in the past," she said. 

  The former Nomura executive also acknowledged a high degree of uncertainty over the economic outlook, including around the impact of higher U.S. tariffs. 

  She then warned of the possibility that a negative shift in expectations for artificial intelligence, which has been fueling a rise in global stock markets, could significantly affect asset prices and slow the U.S. economy. 

  "The bank will make monetary policy decisions as appropriate by continuing to carefully assess data and information that becomes available," she said. 

  At its latest meeting in October, the Japanese central bank held its policy rate steady at 0.5%, keeping it at the same level since its last hike in January. A summary of opinions from that meeting released Monday suggested that the BOJ policy board is getting ready for the next interest-rate hike. 

  "It is likely that conditions for taking a further step toward the normalization of the policy interest rate have almost been met," one policy board member was quoted as saying. 

  Write to Megumi Fujikawa at megumi.fujikawa@wsj.com 

  (END) Dow Jones Newswires

  November 09, 2025 23:54 ET (04:54 GMT)
Copyright (c) 2025 Dow Jones & Company, Inc.

Stay informed with free updates

Simply sign up to the Climate change myFT Digest — delivered directly to your inbox.

“Our industry . . . made a huge mistake,” Douglas Flint, the outgoing chair of UK asset manager Aberdeen, said earlier this year. “It became a marketing thing: let’s tell everyone we’re saving the world.”

Flint was referring to a wave of financial-sector enthusiasm for climate action that reached a conspicuous climax at COP26 in November 2021. At the UN summit in Glasgow, former Bank of England governor Mark Carney announced that $130tn in financial sector assets had been committed to climate action under the Glasgow Financial Alliance for Net Zero (Gfanz).

In the lead-up to this year’s COP30 summit in Brazil the mood has been very different, with Gfanz and its sectoral alliances falling into crisis. Last month the Net-Zero Banking Alliance voted to cease operations, following a string of exits by big US and UK banks. A similar rush of departures prompted the alliance for the asset management sector to suspend operations in January. The insurance sector group Net-Zero Insurance Alliance disbanded in 2024.

A major factor behind the climate alliances’ woes has been political pressure from US Republican officials, who have said financial companies belonging to these groups may be in breach of fiduciary duty to their clients, as well as antitrust rules. Some US state governments withdrew business from financial institutions that were part of the coalitions — including BlackRock, the world’s largest asset management company, whose departure from the Net Zero Asset Managers’ initiative earlier this year helped precipitate its suspension.

Another issue has been consistent strength in world fossil fuel production, defying expectations of government action that would hit output during the 2020s. This has been reflected in the financing patterns of big banks, which have been reluctant to sacrifice their business with an oil and gas sector still chasing growth. The 65 largest global banks increased their financing for fossil fuels by $162.5bn last year to $869bn, reversing the falls in the two prior years.

Yet one segment of the financial sector continues to strengthen its engagement with climate change: pension funds, and other long-term investors known as “asset owners” (institutions that directly own investment assets on behalf of members). In contrast to the other coalitions, the Net Zero Asset Owner Alliance has only two fewer members than a year ago.

“Climate change remains as much in focus for us as ever,” says Laura Hillis, director of climate and environment at the Church of England Pensions Board. “If anything, the increasing physical impacts and updated scientific assessments we’re seeing in 2025 should raise more alarm bells across our sector.”

This year, pension funds, particularly in Europe, have adopted a more assertive approach on climate risk with asset managers. In February, 26 asset owners controlling a total of $1.5tn warned their asset managers they would risk being dropped if they did not engage more strongly on climate risk with companies. Two weeks later, the UK’s People’s Pension fund pulled a £28bn investment mandate from US asset manager State Street, citing concerns over sustainability. In September, Dutch fund PFZW moved about €14bn from BlackRock for similar reasons. The pension funds’ moves reflect a widening divide between European and US asset managers, with the former showing far greater support for environmental shareholder proposals than the latter.

They also come as long-term investors grapple with the financial implications of increasingly severe climate effects. Climate change “will impact the companies we invest in and the value of the fund”, Carine Smith Ihenacho, chief governance and compliance officer at Norway’s $1.8tn wealth fund, said last month. “Our analysis suggests risk of meaningful losses at the portfolio level.”

Some asset managers are seeking to capitalise on pension funds’ concern about climate risks. UK-based Resolution Investors launched in September promising to invest in companies with robust business models as well as strong climate credentials.

This approach contrasts with a wave of green investment around 2021, when fund managers succumbed to “euphoria” around the energy transition without sufficient emphasis on business quality, argues David Lowish, a co-founder at the firm.

That euphoria had already dissipated before Donald Trump returned to the US presidency, as rising interest rates hit the capital-intensive renewable energy sector, as well as many green start-ups. Trump has taken a far more hostile tone towards renewable energy than during his first term, deriding the sector as a “joke” and cutting back its tax credits.

Yet many green investment strategies have been flourishing this year. US clean energy stocks have been surging: the Nasdaq Clean Edge Green Energy benchmark index is up over 30 per cent since January 1, driven partly by tech companies’ eagerness to power data centres for artificial intelligence. They have shown appetite for renewable energy, partly because solar and wind plants are now cost-competitive due to technical advances and economies of scale.

Green technology adoption “is being driven less and less by politics and policy, and more and more by markets and economics”, says Daniel Weiss, managing partner of Angeleno Group, a US venture capital firm focused on low-carbon businesses. “It is definitely a confusing and turbulent time in the capital markets around climate and sustainability. But there are very interesting pockets of opportunity.”

We often assume that FDA relapses when they are exposed to drug-related stimuli. For what it could be, the FDA may also experience drug cravings even when there is no stimulus. Based on previous reports, we hypothesized that the brain areas implicated in drug craving are significantly activated at rest in FDA [16,17,18,19,20]. We used the resting-state functional MRI (rs-fMRI) technique as it allowed us to measure the brain activity of the FDA in a safe and non-invasive way. The novelty of this work is that focuses on individuals who have been abstinent from drug use, whereas most studies have mainly focused on active drug users. Additionally, while studies have been performed by exposing the participants to drug-related stimuli, the present work examined the brain activities of the FDA in the absence of such stimuli. This method would allow us to better understand the neural mechanisms underlying drug craving more straightforwardly. With this understanding, an evidence-based treatment plan or rehabilitation programme could potentially be developed to effectively inhibit drug craving at the neural level, thus minimizing the relapse rate.

When drug-abstinent individuals are exposed to drug-related stimuli, the brain areas responsible for craving, reward, motivation, memory, and cognitive control show significant activations [11]. The significant activations in these brain areas (commonly referred to as the reward pathway) have been found to induce drug cravings in drug-abstinent individuals, thus increasing their tendency to relapse [34]. However, less is known about whether similar activation would be observed in drug-abstinent individuals even when there are no drug-related stimuli. Therefore, the primary purpose of this study was to investigate whether the brain areas implicated in drug cravings show significant activation in drug-abstinent individuals without exposing them to drug-related stimuli. This study found seven significantly activated brain clusters when drug-abstinent individuals were at rest. These clusters comprised seven brain areas: the right PrG, left middle MOG, right SPL, right MCgG, bilateral AnG, left SFG, and left SMG. We interpret these findings in relation to those reported previously in the neuroscience literature.

The brain area with the highest peak of activation is the right PrG. The PrG is part of the primary motor cortex and is known for controlling the body’s voluntary motor movement [35]. Activation of the right PrG has been reported in many rs-fMRI studies [36,37,38,39]. In these studies, activation of the PrG during rs-fMRI was attributed to its role in the salience network—a brain network crucial for detecting and responding to internal and external stimuli. The salience network is constantly activated even when one is at rest. Similarly, we attribute activation of the right PrG to the processing and coordination of neural resources. Moreover, studies have not reported the role of the right PrG in drug addiction. The second-highest peak of activation was found for the left MOG, which is primarily responsible for visual processing. The specific role of the left MOG in drug addiction has been under-reported in the literature. Of those that did, most reported a decrease in activation of the left MOG rather than an increase. For instance, it was found that individuals with major depressive disorder and active cocaine addicts had decreased activations of the left MOG [40, 41]. As most studies mainly focused on active drug addicts, there is no available report on the activation of the left MOG in the FDA. As such, we attributed this increase in activation to visual processing, where our participants had to keep their eyes open during the rs-fMRI scans to avoid falling asleep [42,43,44].

Next, we found significant activation of the right SPL, which is responsible for many neuronal activities including visuomotor, cognitive, sensory, working memory, higher order, and attentional [45]. The significant activation of the right SPL in this study aligns with an earlier study demonstrating that drug-related stimuli evoked activation of SPL in heroin-abstinent individuals [11]. The authors mentioned that SPL played a role in motor preparation and output, suggesting that its activation may induce drug craving. Consequently, when viewing drug-related images, heroin-abstinent individuals tend to imagine the action of taking drugs, thus triggering the SPL. Although our study did not require participants to view drug-related images, the significant activation of SPL may suggest a residual neural activity related to motor imagery or planning, possibly linked to their history of drug use.

Apart from that, we also found a significant activation of the right MCgG. This region is part of the limbic system which is involved in reward processing, motivation, emotion, cognitive, motor function, and memory [46]. It has been proposed that the cingulate gyrus serves as a central “hub” in addiction-related neural networks of cognitive functions [47]. In addition to this, increased activation of the right MCgG has been suggested to modulate the reward circuit in individuals with cocaine use disorder [48]. In support of this, a study found that increased activation of this region was associated with increased cocaine craving in cocaine-dependent individuals [49]. Therefore, the significant activation of the right MCgG suggests that drug-abstinent individuals may still crave drugs even at rest. However, this assumption is arguable considering that the participants reported no feeling of craving. The role of the right MCgG in drug addiction remains elusive and warrants further investigations.

Subsequently, this study showed significant activation in bilateral AnG, left SFG, and left SMG. Previous rs-fMRI studies highlighted the role of these three regions as the primary center of DMN, which is mainly involved in mind wandering and internal thinking and is more highly activated during resting or passive conditions rather than during task-specific conditions [41, 50]. They are expected to be involved in internal mental states and become prominent when participants do not engage in external interaction. Significant activation of these regions may be attributed to their role in monitoring internal and external environments [51]. This explanation is plausible given that the participants did not engage in any task-specific condition during the rs-fMRI scans. Moreover, previous studies have not implicated these regions in drug addiction.

In summary, our findings did not support our hypothesis as we initially expected to see significant activation in the reward-related pathway including the insula cortex, basal ganglia, insula, amygdala, parahippocampal gyrus, and DLPFC [16,17,18,19,20]. We firmly believe that the reason behind this may likely be due to the demographics of our participants, which are very different from those of the previous studies. More precisely, our study involved participants who had been drug abstinent for approximately six years, while the participants in the earlier studies on which we based our hypothesis were either active users [16, 17] or had been drug abstinent for less than three months [20]. Therefore, the activation of the reward pathway documented in these studies may be due to their participants’ brains being still sensitive to drug craving. This is plausible provided evidence showing significant brain activations in FDA who had been drug abstinent for up to two years [34]. Our findings suggest that prolonged abstinence may reduce drug-specific brain responses. This is supported by previous reports linking prolonged abstinence with reduced brain reactivity and diminished activation of the reward pathway [51, 52]. The diminished activity in the reward pathway may be due to neuroplasticity—the ability of the brain to recover and restore its normal brain function in affected brain regions over time [53]. This neural adaptation was found in the brains of former marijuana and methamphetamine users following a long period of abstinence for at least 14 mon [54, 55]. Additionally, longitudinal neuroimaging studies provide strong evidence that prolonged abstinence from drug use leads to functional and neurochemical recovery in brain regions associated with reward and craving. For instance, a study found that heroin-dependent individuals with long-term abstinence showed significantly reduced activation in the medial prefrontal cortex, anterior cingulate cortex, and caudate in response to drug cues, compared to those with short-term abstinence [56]. This attenuation of cue-reactivity observed in the long-term abstinence participants was also associated with more stable subjective craving. Similarly, a systematic review examined 45 longitudinal neuroimaging studies involving treatment-seeking individuals with substance use disorders who underwent abstinence for at least two measurement time points [57]. The results consistently showed at least partial neurobiological recovery following abstinence. Structurally, significant regeneration occurred primarily within the frontal cortex, insula, hippocampus, and cerebellum, suggesting substantial restoration of gray matter integrity. Functionally and neurochemically, improvements were observed in prefrontal regions as well as subcortical areas such as the midbrain, striatum, and thalamus. Notably, structural recovery tended to emerge first, followed soon after by neurochemical normalization, and functional improvements, which required longer periods of abstinence. Overall, these findings underline the brain’s remarkable capacity for neuroplastic recovery during sustained abstinence and highlight the importance of timing in treatment interventions and relapse prevention. Based on these reports, the non-significant activation of the reward pathway reported in this study may indicate that the brains of the FDA rewire themselves and begin to normalize after a long period of abstinence. Together, these findings support the interpretation that neuroplastic changes during extended abstinence contribute to the normalization of brain function and reduced vulnerability to relapse. The findings also underscore the importance of long-term abstinence in inhibiting drug cravings and lowering relapse rates.

It is worth noting that this study has several limitations that should be considered when interpreting its results. Firstly, we only reported the brain areas that were significantly activated at rest and did not extend our investigation to connectivity analyses. The reason was that there are currently limited studies investigating the brain networks of the FDA, particularly in those using methamphetamine. Most studies have largely focused on heroin and cocaine users. In this study, we focused on methamphetamine as it is the most abused substance in the authors’ country [58]. As such, we focused on identifying and understanding the activated brain regions before extending our analyses to examine the functional connectivity of these regions. Additionally, as we only focused on methamphetamine, we cannot generalize our findings to other types of substance use (e.g. heroin or cocaine).

Secondly, most studies on this topic focus on active substance users. Of those that examined abstinent individuals, very few focused on methamphetamine. Therefore, the interpretation and discussion of our results are very limited and include findings that are not specific to methamphetamine alone (e.g., including results from cocaine and heroin studies). This limitation indicates that more research is warranted in exploring the brain functions of methamphetamine users, particularly in those who are no longer active users.

The third limitation is the demographics of our participants where we only included male participants, which limits the generalizability of the results to female FDA. Therefore, future studies should consider including both male and female FDA as their participants and compare the findings of both genders. A gender balanced study is recommended to reduce the potential confounding effects, ensure consistency, and enhance the validity of the results. Thus, a more comprehensive understanding of how the brain induces drug craving in male and female FDA could be achieved, allowing for the planning of a treatment that considers possible gender differences.

The fourth limitation is that the non-significant finding of the reward pathway activation may reflect the low sensitivity due to the small sample size. At the beginning of the study, we calculated that the optimal sample size for this study was 24 participants. However, after excluding those who did not meet the eligibility criteria, we were left with only 20 participants. Based on an earlier study suggesting that a sample size of 16 to 32 participants is adequate to achieve good sensitivity in an fMRI study [22], we believed that our number of participants was optimal, even after excluding the four participants. However, given that the results were not statistically significant at the stringent threshold, we should reconsider our understanding of this matter. Moreover, a much more recent fMRI study recommended a sample size ranging between 46 and 72 participants to yield results with high statistical power [59]. Therefore, future fMRI studies on addiction in particular should consider using a large sample size of 46 to 72 participants. Moreover, we did not include a healthy group for comparison as our primary objective was to investigate neural activation patterns within the FDA group. However, the absence of a control group limits the ability to determine whether the observed activations are specific to the FDA or reflect typical resting-state activity. Future studies should include matched healthy controls to clarify this distinction.

While the absence of reward pathway activation in the FDA group may be interpreted as a sign of neuroplastic recovery following long-term abstinence from methamphetamine use, it is important to consider alternative explanations for this finding. One possibility is that the observed null effect reflects limited statistical power, particularly given the relatively small sample size and the variability often observed in neural responses among individuals with a history of substance use. Additionally, task-related factors or residual effects of prior drug use could have influenced the observed activation patterns. Without a healthy control group, it is difficult to determine whether the absence of activation in reward-related regions is specific to the FDA group or reflects a normative resting-state pattern. Although a formal Bayesian analysis could have provided a quantitative estimate of the evidence for the null hypothesis (i.e., absence of activation), we chose not to conduct such an analysis due to the exploratory nature of the study, the single-group design, and the limited sample size. Under these conditions, Bayesian results could be sensitive to prior assumptions and potentially difficult to interpret without appropriate control comparisons. Nevertheless, we acknowledge the value of Bayesian approaches for evaluating null effects and recommend that future studies adopt this framework, particularly in combination with larger, well-powered samples and the inclusion of appropriate control groups. Such designs would allow for a more precise interpretation of the absence of activation in the reward pathway and contribute to a clearer understanding of the neural mechanisms underlying recovery in addiction.

Lastly, we acknowledge that examining functional connectivity represents a key strength of resting-state fMRI and could provide additional insights into the neural mechanisms underlying addiction and recovery. However, due to the scope and objectives of the current research that primarily focused on activation patterns within specific brain regions, functional connectivity analyses were not conducted. Future studies incorporating both activation and connectivity approaches would be valuable to more comprehensively characterize the alterations in brain networks associated with long-term abstinence from methamphetamine use. Despite these limitations, the findings of this study have advanced our understanding of the neural mechanisms underlying addiction in long-term abstinent individuals and contribute to the limited literature on methamphetamine addiction.

Study design

This retrospective cohort recruited 776 patients hospitalized in three tertiary teaching hospitals in Shiraz, Iran, from March to September 2020. We included hospitalized adult patients with COVID-19 confirmed by positive results of real-time RT-PCR from nasopharyngeal oropharyngeal swabs and underwent at least one chest CT imaging without any contrast agents. The patients below 18 years of age and those who lacked positive PCR results or did not undergo chest CT at the initial stage of their hospital admission were excluded. Applying the eligibility criteria, we excluded 87 patients due to a lack of positive PCR report and 123 due to a lack of chest CT or assessable CT imaging. Hence, a total number of 566 patients were included in this study. Afterward, the study population was divided into two groups based on in-hospital mortality status: survivors and nonsurvivors.

All the patients underwent initial laboratory testing such as CBC, coagulation markers, and kidney and liver function tests. Patients at each center received treatment and managed based on a similar national guideline released by Iran’s Ministry of Health and Medical Education (DTFC: Diagnostic Therapeutic Flowchart for COVID-19) [17]. Indications for ICU admission, mechanical ventilation use, and discharging criteria have been described in guideline [17].

Variables

This study collected data on three domains of patients’ demographic features, initial clinical manifestations, disease outcomes, and radiologic findings. Clinically relevant variables included sex, fever, dry cough, dyspnea, admission to intensive care units, and mortality status. Regarding laboratory variables, leukopenia was defined as WBC count < 4.0 × 10⁹ per liter, thrombocytopenia as PLT count < 150 × 10⁹ per liter, and lymphocytopenia as lymphocyte count < 1.5 × 10⁹ per liter. Reviewing CT images, we recorded radiologic findings, including crazy paving pattern, halo sign, reversed halo sign, air bronchogram, pleural effusion, GGO/consolidation, and mediastinal lymphadenopathy (Fig. 1).

Validation of visual severity scores

We searched the electronic database using the keywords of “prognostic modeling,” AND “COVID-19,” AND “CT findings”. We reviewed 13 related papers with the development of a prognostic model identifying COVID-19 patients at high risk of mortality. Then, three articles were chosen based on sample size, methodology, and their proposed prognostic models. Upon quality assessment of the radiologic-based scoring component of these selected models, three CT severity scoring systems [12,13,14] were finally selected to be validated in this study, and their respective variables were investigated. Table 1 summarizes the details about the function and interpretation of the selected radiologic models. All the patients were evaluated by three prognosis-predictive models of COVID-19. We extend our analyses to propose the most fitting model according to results obtained from our study population (Table 1).

The data required to validate selected prognostic models include patients’ demographic features, clinical history, laboratory results, and radiologic findings. All data were registered in patients’ medical records at the hospital admission course. Also, these patients received the necessary medical treatment and supportive care during their hospitalization course. The review of the selected articles clarified three systems of radiologic severity scoring as follows:

• 15-Score model Wang et al. [13] introduced a prognostic model with a maximum score of 15

• 20-Score model Tabatabaei et al. [14] defined a CT-based predictive model with a sum score ranging from 0 to 20

• 24-Score model Zheng et al. [12] introduced a prognostic nomogram based on clinical and CT features with a sum score ranging from 0 to 24

CT-scan acquisition

All the patients underwent chest CT-scan without contrast injection with a single inspiratory phase while holding their breath. The acquisition of images in all centers was conducted in the supine position of the patients using a multidetector 16-section scanner (GE Medical Systems, Milwaukee, WI, USA). The tube voltage was set at 120 kV for all scans. The CT images included the whole extent of the chest and the upper portion of the abdomen. The CT apparatus photographed the images with a 1.25–2-mm thickness and intervals of 1.25 mm.

The patients’ CT images were reviewed and analyzed by a certified expert radiologist (with more than 15 years of experience) blinded to the disease outcomes, clinical presentations, demographic features, and laboratory profile. The radiologist stratified the severity of pulmonary involvement according to each scoring system, as explained in Table 1. Each scan was evaluated for GGO, consolidation, crazy paving, halo sign, reversed halo sign, air bronchogram, pleural effusion, and lymphadenopathy. The definition of these CT manifestations had been previously explained in Fleischner’s Glossary of Terms for Thoracic CT [18]. Further, the CT images were evaluated for changes in liver density as described previously [12,13,14]. We categorized opacification patterns into GGO, consolidation, and mixed GGO and consolidation. Also, mediastinal lymphadenopathy was considered when the axial diameter of lymph nodes exceeded 1 cm.

Statistical analysis

The data obtained were analyzed by SPSS v.24 software. All the qualitative data were presented in frequency and percentages. These categoric variables were compared between groups using the Chi-square test. On the other hand, quantitative data were presented in median (IQ2) due to non-normal distribution. These continuous variables were compared using a nonparametric Mann–Whitney U test. We compared the CT findings across the survivor and nonsurvivor groups using linear-by-linear association.

The external validation was performed using a logistic multivariate regression model. At this step, each radiologic severity scoring system was independently assessed with three variables of age, gender, and lymphocyte count by multivariate logistic regression model. Our statistical goal was to predict the mortality outcome (Y or dependent variable) based on predictor factors (X or independent variables). Four variables were chosen to build a multivariable prediction curve to avoid overfitting bias. It means that the independent variables should minimally show inter-dependence on each other. We conducted a correlation analysis between the predictors to avoid multicollinearity in the regression model. The correlation for each pair of these four variables was separately calculated before including them in the logistic regression model. Finally, to identify the COVID-19 patients with a high risk of mortality in the most accurate fashion, we conducted ROC analysis to determine the most optimal system of radiologic severity scoring. A significance was considered when P < 0.05 for all the analytic tests.