OpenAI’s EMEA startups head Laura Modiano spoke at the Sifted Summit on Wednesday, 8 October.
Nurphoto | Nurphoto | Getty Images
OpenAI’s artificial intelligence chatbot ChatGPT is down for some users.
The company said it is “currently experiencing issues,” including “increased ChatGPT error rates,” according to an update on OpenAI’s status page.
“We have applied the mitigation and are monitoring the recovery,” the status page said.
OpenAI did not immediately respond to a request for comment.
Roughly 3,000 people reported issues with the chatbot on Tuesday, according to Downdetector, a website that tracks outages.
The outage comes days after OpenAI disclosed a security breach at Mixpanel one of OpenAI’s data analytics providers.
The breach compromised user information, such as names, emails and other details tied to the OpenAI API.
OpenAI did not disclose how many users were affected, saying in a blog post that an attacker “exported a dataset containing limited customer identifiable information and analytics information.”
OpenAI kickstarted the AI boom with the launch of ChatGPT three years ago. As of October, OpenAI said more than 800 million people use the chatbot each week.
The Fed’s standing repo facility (SRF) was tapped for $26bn over the course of the day on December 1, making it the largest daily uptake since $50bn in usage on October 31. Also on December 1, the tri-party general collateral repo rate (TGCR) rose to 18bp over the interest paid on reserves at the Fed (IORB). Both developments speak to tightening in funding markets heading into year end. The current situation is hardly a surprise, as we have commented on recently (see here and here).
However, on November 28, the effective federal funds rate moved up a basis point, from 4.88% to 4.89% with TGCR-IORB spreads at 18bp. This illustrates the Fed’s discomfort with such upward pressure on repo spreads. Tight liquidity conditions in the funding markets can ultimately lead to upward pressure on the Fed’s operating target, threatening rate control.
While funding pressure has materialized around specific dates (month- and quarter-ends, settlement days, tax dates), there is the risk it will happen more frequently, including on otherwise uneventful days. This is why the Fed has indicated it will eventually need to resume increasing its balance sheet. As other balance sheet liabilities increase (namely, currency in circulation), reserves will fall, exacerbating tight liquidity conditions in the funding markets. Reserves are currently reckoned to be no longer abundant, but merely ample. Reserve management operations will eventually feature in the Fed’s toolkit, although pinpointing when they might commence is difficult.
Exhibit #1 shows the daily usage of the SRF over the past half year and illustrates how its usage increases when funding is stressed. It’s worth noting that the Fed would probably prefer to see the SRF used more frequently and in larger sizes than it currently is, reducing the eventual need for open market operations. Moving to the latter presents a potential communications problem for the Fed, which would have to make it clear that such operations are not a return of quantitative easing, but are indeed reserve management policies. The SRF’s attractiveness suffers due to both internal and executive stigma, as well as a lack of central clearing.
We wrote about the Fed’s upcoming monetary policy decision last week and will write a formal preview of the FOMC next week. However, it’s worth asking here whether the Fed would announce reserve management operations at next week’s gathering. We think it’s unlikely to occur so soon. For one thing, there have so far been only vague references to such market operations in the Fed’s public communications. We would have expected more specific guidance if they were to commence soon. Furthermore, with funding market strains still mostly limited to specific dates, it could be a bit premature to set up such operations. Nevertheless, we expect them to begin early in 2026, as funding markets gradually tighten further.
Traders work on the floor of the New York Stock Exchange (NYSE) on December 02, 2025 in New York City.
Spencer Platt | Getty Images News | Getty Images
Stocks rose on Tuesday, boosted by gains in bitcoin and technology names, as traders recovered some of the ground lost in the previous session.
The Dow Jones Industrial Average gained 185.13 points, or 0.39%, to end the day at 47,474.46. The S&P 500 climbed 0.25% to settle at 6,829.37, while the Nasdaq Composite advanced 0.59% to finish at 23,413.67.
Bitcoin rose around 7% Tuesday, recouping some of its losses from the prior day. Tech players linked to the artificial intelligence trade supported the broader market as well. AI chip darling Nvidia increased almost 1%, while AI infrastructure play Credo Technology soared 12% and hit an all-time high on the back of better-than-expected earnings.
To be sure, it’s been a topsy-turvy session for stocks. The S&P 500 and Dow briefly turned negative on the day, while the Nasdaq got close to the flatline before moving back higher.
SPX intraday
The major U.S. indexes began the week in the red, ending five-day win streaks on Monday. Risk-off sentiment has pressured the bull market in recent weeks as worries of persistent inflation, elevated valuations and returns on artificial intelligence spending weigh on investors.
Although November was a mixed month for stocks, investors are watching for catalysts that could lead to a year-end rally.
Traders are currently optimistic that the Federal Reserve will announce an interest rate cut on Dec. 10 at conclusion of its next policy meeting. Markets are pricing a roughly 89% chance of a cut during the upcoming meeting, which is much higher than the odds from mid-November, according to the CME FedWatch tool.
“Markets appear to have moved away from uncertainties surrounding Fed policy and the Dec. 10 FOMC and focusing instead on better-than-expected earnings projections for the fourth quarter and calendar year 2026, in addition to looking beyond the economic soft patch we’re currently experiencing to growth accelerating later next year,” said Doug Beath, global equity strategist at Wells Fargo Investment Institute. “Seasonality also favors stocks in December, particularly after a weak November.”
According to the Stock Trader’s Almanac, the S&P 500 averages a gain of more than 1% in December, making it the third-best month of the year for the benchmark in records going back to 1950.
The Genesis Mission will be a coordinated national effort to build an integrated AI platform (the “Platform”) for Federal scientific datasets. The Platform will be used to train scientific foundation models and create AI agents for the testing of new hypotheses, automation of research workflows, and acceleration of scientific breakthroughs. The new executive order claims “The Genesis Mission will dramatically accelerate scientific discovery, strengthen national security, secure energy dominance, enhance workforce productivity, and multiply the return on taxpayer investment into research and development, thereby furthering America’s technological dominance and global strategic leadership.”
Implementation by the Secretary of Energy
The Secretary of Energy will be responsible for implementing the Genesis Mission within the Department of Energy. Within 60 days of the executive order, the Secretary will identify and submit to the Assistant to the President for Science and Technology (the “APST”) a list of at least 20 science and technology challenges that can be addressed through the Genesis Mission. These challenges may pertain to advanced manufacturing, biotechnology, critical materials, nuclear fission and fusion energy, quantum information science, or semiconductors and microelectronics.
The executive order established certain other milestones for the Secretary to complete following the issuance of the executive order:
90 days – The Secretary will identify Federal computing, storage, and networking resources necessary to support the Genesis Mission.
120 days – The Secretary will identify initial data and model assets for use in the Genesis Mission and develop a plan for the incorporation of relevant datasets from federally funded research, other agencies, academic institutions, and private-sector partners.
240 days – The Secretary will review the capabilities of the Department of Energy national laboratories and other Federal research facilities for robotic laboratories and production facilities with the ability to engage in AI-directed experimentation and manufacturing.
270 days – The Secretary will seek to demonstrate the initial operating capability of the Platform for one of the identified national science and technology challenges.
Coordination of Other Agencies
The coordination of participating executive departments and agencies will be conducted by the Assistant to the APST. As part of its coordination efforts, the APST will assist participating agencies in aligning their AI-related programs, datasets, and research and development activities with that of the Genesis Mission and identify any relevant data sources. The APST will also launch coordinated funding opportunities or prize competitions to incentivize private-sector participation in AI-driven scientific research.
Coordination with Industry
The Secretary will also, working with the APST and the White House’s Special Advisor for AI and Crypto, establish mechanisms for agency collaboration with industry participants. Specifically, the administration is seeking to work with private companies who have domain expertise concerning advanced AI, data, and computing capabilities. The executive order then directs the Secretary, in establishing a framework for collaboration, to consider issues such as trade secret protection, IP ownership and commercialization, and cybersecurity standards. These considerations will be important in order to convince industry participants to collaborate with the government.
Looking Forward
We can expect a report from the Secretary within a year regarding the state of the Platform, including its operational status and capabilities, a status of user engagement, and the scope and outcomes of public-private partnerships. The establishment of the Genesis Mission is consistent with the Trump Administration’s pro-innovation stance regarding artificial intelligence. We have yet to see any new AI regulations at the Federal level but continue to see more regulations at the state level, such as the Transparency in Frontier Artificial Intelligence Act in California, which we have discussed in a separate article. We anticipate more AI-related announcements from the Trump Administration in the near future.
The current education research-to-policy pipeline is too slow to keep pace with the urgent needs of districts and states. Researchers face steep barriers to accessing high-quality, multimodal data, while existing R&D infrastructures remain siloed and under-resourced. Without scalable, trusted, systems that enable timely and secure data use, the U.S. risks falling behind in generating actionable and evidence-based insights to guide policy and practice. In this memo, we discuss how privacy-preserving research models can be used to strengthen education R&D capacity.
Challenge and Opportunity
Learning is a lifelong and multidimensional process, yet data about learning has historically been difficult to obtain. The shift to digital learning platforms (DLPs), accelerated by COVID-19, has created a wealth of data, but accessing it remains complex and slow – especially for researchers with fewer institutional resources.
Additionally, complex privacy laws, such as the Children’s Online Privacy Protection Act (COPPA) and Family Educational Rights and Privacy Act (FERPA), alongside state-specific regulations and institutional risk aversion, create substantial barriers. These laws were not designed to accommodate privacy scenarios within the current environment of pervasive data collection and rapidly advancing AI.
As such, trusted mechanisms for safe data access that remove barriers to critical R&D, bolster global competitiveness, and leverage innovation to cultivate a skilled STEM workforce, are more important than ever. Without trusted mechanisms to ensure privacy while enabling secure data access, essential R&D stalls, educational innovation stalls, and U.S. global competitiveness suffers.
Flipping the traditional research model
The landscape of educational research and development (R&D) is rapidly evolving as digital learning platforms (DLPs) capture increasingly rich streams of data about how students learn. These multimodal data streams provide unprecedented opportunities to accelerate insights into how learning happens, for whom, and in what contexts – as well as how these processes, in turn, affect learning outcomes, engagement, and persistence. Yet, despite this potential, access to platform-generated learning data remains highly constrained – particularly for early-career researchers with minimal institutional resources and organizations outside elite academic settings.
Current challenges to accessing DLP data include privacy risks (e.g., data leaks), opaque legal environments, institutional risk aversion, and the lack of trusted third-party intermediaries to balance privacy with data utility. As a result, promising research is delayed and the research-to-policy pipeline is exacerbated – leaving decision-makers without timely evidence to address urgent needs such as learning recovery, responsible AI integration, or workforce readiness.
Privacy-preserving models offer transformative opportunities to address these barriers. Across sectors, the field is converging on trusted research environments that include secure enclaves that keep data in situ and move analysis to the data. SafeInsights, the U.S. Census’ Federal Statistical Research Data Center (FSRDC), and North Carolina Education Research Data Center (NCERDC) are examples of such systems complemented by privacy-preserving methods.
Privacy-preserving research models, such as SafeInsights, flip the traditional research model: instead of giving data to researchers, it brings researchers’ questions and analyses, encoded as software, to the data. At no point in the research process does the researcher have direct access to raw data, thereby minimizing concerns for data leaks.
Researchers instead use sample or synthetic data to craft their analyses. Once the researchers’ analysis code is submitted to the owner of the data, it is reviewed by experts for approval. This model minimizes risk, reduces delays in the research-to-policy pipeline, and unlocks data that would otherwise remain inaccessible.
Think of it as a secure research zone: a trusted third-party intermediary where researchers can run analyses using specific tools and applications, but cannot access data directly, ensuring strict security.
Rather than extracting and sharing sensitive data with researchers, privacy-preserving research models bring researchers’ analytic tools to secure data enclaves – preserving privacy while enabling rigorous, scalable, inquiry of DLP data. Through secure enclaves, transparent governance, and standardized compliance frameworks, a durable large-scale infrastructure for research can be created.
Benefits of privacy-preserving research models
Accelerate time to insight for policy and decision-makers who need rapid, evidence-based guidance. Standardized governance reduces delays arising from fragmented compliance and legal processes. For federal, state, and local level policy and decision-makers, this means actionable insights can be delivered in months rather than years, potentially informing legislative decisions and programs with greater speed.
Safely join data across platforms, enabling richer analyses of student learning. Shared infrastructure maximizes critical research infrastructure return on investments and spreads costs across funders. Secure, trusted, interoperable research environments protect privacy while enabling cumulative evidence. This aligns with federal agency priorities to modernize research infrastructure and ensure taxpayer investments translate into impact.
Democratize access and participation in complex research by lowering barriers for early-career researchers with minimal institutional resources and organizations outside elite academic settings. Lowering barriers to entry broadens the reach of federal R&D investments and supports state leaders and research organizations seeking to participate in research.
By securing cross-sector investment for embedding scalable privacy-preserving models into R&D ecosystems and infrastructures, we can expand access to high-value data while supporting long-term research scalability, security, and trust.
Such models can fill a critical gap in the R&D ecosystem by establishing a secure and sustainable research infrastructure that extends well beyond its initial NSF funding and is ideally suited to broker access between DLP developers, school districts, and researchers.
Plan of Action
Promote R&D Infrastructure Development and Sustainability
Privacy-preserving research models have the potential to offer researchers safer, faster, reliable, high-value, de-identified data analyses – while simultaneously saving DLPs and school districts time and resources on compliance reviews and privacy audits. It also creates opportunities for funders to support a sustainable research infrastructure that multiplies the impact of each dollar invested.
To move from promise to practice, interested stakeholders, including research institutions, school districts, and funders, should consider the following actions:
Recommendation 1. Lay the Foundation for Sustainable Large-Scale R&D Infrastructure
Conduct policy landscape scans, including review of state student privacy laws, to identify commonalities, constraints, and pathways for district participation.
Interview stakeholders, including district data leads, state education agencies, and platform providers, to understand pain points and demand for trusted intermediaries.
Review existing research infrastructures and operational frameworks, including research data hub governance, fee structures, data-sharing agreements, IRB support services, and services, adapting effective practices to the privacy-preserving context.
Recommendation 2. Embed Infrastructure Costs into Research Contracts and Budgets
Require researchers to include service fees for privacy-preserving infrastructure directly in grant applications, with templates to simplify proposal preparation.
Embed privacy-preserving infrastructure costs in contracting and budgeting to support scalability, drive down the marginal cost of data access across the field, and make rigorous educational research more accessible and sustainable beyond single grants.
Recommendation 3. Catalyze Scaling through Foundation and Philanthropic Support
Recommendation 4. Develop Large Scale R&D Infrastructure across Sectors
Extend privacy-preserving models across sectors, such as education, health, workforce, housing, and finance, to capture increasingly rich streams of data about how people live, learn, work, and access services.
Enable secure, interoperable, cross-sector research on questions such as how early education experiences impacts long-term workforce outcomes or how neighborhood-level educational access connects to public health disparities.
Align with federal agency efforts, such as the Federal Data Strategy, to support the linking of data ecosystems across sectors.
Conclusion
Privacy-preserving research models offer standardized, secure, and privacy-conscious ways to analyze data – helping researchers at the local, state, and federal levels understand long-term educational trends, policy impacts, and demographic disparities with unprecedented clarity.
By accelerating time-to-insight, investing in critical R&D infrastructure, and expanding participation in complex research, privacy-preserving research models offer possibilities for delivering on urgent policy priorities – building towards a modern, responsive, trustworthy education R&D ecosystem.
What kinds of research topics can be explored using privacy-preserving research models?
Privacy-preserving research models could offer the possibility to connect researchers with DLP data representing different learning contexts. DLP data is often rich and versatile, possibly enabling the exploration of multiple research topics, including:
Learning Behaviors: Analyze patterns of engagement, tool usage (e.g., text-to-speech, digital pencil), or response time.
Personalized Learning: Investigate how adaptive experiences influence outcomes.
Achievement Gaps: Study differences across subgroups (e.g., students with disabilities, English Language Learners).
Intervention Effectiveness: Test how interventions or instructional strategies impact student performance.
Learning Trajectories: Examine longitudinal progress and identify barriers to success.
read more
What kinds of data could be made available through privacy-preserving research models?
Privacy-preserving research models could facilitate connections among various types of educational data from DLP developers, each representing different aspects of K16+ teaching and learning, including administrative records, learning management systems, and curricular resource usage data.
Examples of DLP data categories include digital curricula, university data systems, and student information systems for K-12 institutions.
read more
What are some examples of privacy-preserving research models utilizing secure enclaves across different sectors?
Across sectors, the field is converging on privacy-preserving research models that utilize secure enclaves to keep data in situ and move analysis to the data. Such examples include:
Federal statistical system: the FSRDC network provides secure facilities (now including some remote access) where qualified researchers run analyses on restricted microdata under rigorous review.
Cross-agency administrative data: the Coleridge Initiative’s Administrative Data Research Facility (ADRF) is a FedRAMP-certified, cloud based platform that supports inter-state and inter-agency linkages under shared governance.
State education data enclaves: NCERDC at Duke University and the Texas Education Research Center (ERC) support secure access to longitudinal education/workforce data with well-defined agreements and masking rules.
Health: OpenSAFELY operationalizes a strict “code-to-data” model—researchers develop code on dummies, submit jobs to run against in-place EHR data, and only aggregate outputs leave the enclave. NIH’s N3C and All of Us Researcher Workbench similarly provide secure, cloud based research environments where individual-level data never leave the enclave.
These approaches are complemented by privacy-preserving release methods (e.g., differential privacy), used by the U.S. Census Bureau and supported by open-source toolkits like OpenDP/SmartNoise.
read more
How might privacy-preserving research models support research and researchers?
At the center of privacy-preserving research models is privacy-by-design that enables secure research with protected information – while alleviating technical, logistical, and collaborative challenges for researchers.
Technical
Privacy-preserving research models can offer technical components that support large-scale digital learning research such as:
Analysis options, which enable large-scale analysis of single platform data
Intervention options, which enable researchers —under appropriate agreements—to introduce different kinds of interactive activities (including surveys, assessments, and learning activities) within a partner platform’s student experience
Enclave fusion, which in some designs can enable researchers to leverage multi-platform data
Logistical
Shared data sharing agreement templates
Streamlined IRB and data-sharing processes
Consent management across different populations
Regulatory compliance with the changing data protection landscape
Community and Collaboration
Help easily surface researchers and the research that they are conducting
Bridge connections among platforms, researchers, and educational institutions to support meaningful research to inform practice
Connect researchers at different levels of their careers and different domains to support mentorship and collaboration
read more
Case Study: Turning Student Assessment into Actionable Insights
If assessment results are the scoreboard that reveals what students are learning, user data is the game film that reveals how students learn: time on task, requesting support, revising, using resources.
Using SafeInsights’ privacy-preserving tools, researchers can securely analyze real-time digital learning platform data to better understand how students engage with digital learning. Consider two students with the same score:
Student A works steadily, using hints to revise answers. This pattern suggests a need for additional content support, scaffolding, and practice.
Student B races through with rapid guessing and skipped items. This pattern suggests a need to adjust prompts, pacing, and support.
By distinguishing between these pathways, researchers, educators, and policymakers can target digital learning platform interventions more precisely—whether that means redesigning practice problems, adjusting instructional supports, or tailoring engagement strategies.
Bottom line: SafeInsights securely transforms raw data into actionable evidence, helping policymakers and practitioners invest in solutions that boost learning outcomes and improvement at scale.
read more
Education & Workforce
day one project
Privacy-Preserving Research Models Essential for Large Scale Education R&D Infrastructure
Without trusted mechanisms to ensure privacy while enabling secure data access, essential R&D stalls, educational innovation stalls, and U.S. global competitiveness suffers.
12.02.25 | 6 min read
Education & Workforce
day one project
Analytical Literacy First: A Prerequisite for AI, Data, and Digital Fluency
tudents in the 21st century need strong critical thinking skills like reasoning, questioning, and problem-solving, before they can meaningfully engage with more advanced domains like digital, data, or AI literacy.
11.07.25 | 13 min read
Improving Standardized Test Score Reporting and Administration for Students, Caregivers, and Educators
We need to overhaul the standardized testing and score reporting system to be more accessible to all of the end users of standardized tests: educators, students, and their families.
11.05.25 | 10 min read
Education & Workforce
day one project
Moving Federal Postsecondary Education Data to the States
Moving postsecondary education data collection to the states is the best way to ensure that the U.S. Department of Education can meet its legislative mandates in an era of constrained federal resources.
Your guide to what Trump’s second term means for Washington, business and the world
Shares in a US cryptocurrency miner backed by Donald Trump Jr and Eric Trump shed a third of its value on Tuesday as early investors cashed out en masse at the end of a lock-up period.
American Bitcoin was down 37 per cent by mid-afternoon in New York, wiping roughly $1bn from its market value. Trading volume in the stock was almost 40 times the daily average, according to Bloomberg data.
Eric Trump attributed the sell-off to investors in a $215mn private placement in June exercising their ability to “cash in on their profits for the first time”.
This is “why we will see volatility” in the company’s share price, the president’s son said on X. “I’m holding all my [American Bitcoin] shares — I’m 100% committed to leading the industry.”
American Bitcoin, which says its “ambition is to build the strongest and most efficient Bitcoin accumulation platform in the world”, went public in September through a reverse merger with Nasdaq-listed miner Gryphon Digital Mining.
Eric Trump is American Bitcoin’s co-founder and chief strategy officer. Donald Trump Jr was an early investor. Matt Prusak, American Bitcoin’s president, said the end of the share lock-up period “affects who can buy or sell, not the assets we operate or the work the team is doing every day”.
Donald Trump has loosened regulations on the crypto sector since returning as president this year and has vowed to make the US a digital currency “superpower”. He previously described the value of crypto as based on “thin air”.
The broader crypto market has tumbled in recent weeks as investors have pulled back from risky assets, with bitcoin down about 30 per cent since its latest peak in early October. Bitcoin steadied on Tuesday, however, up 5 per cent at about $90,000 per token.
American Bitcoin — which creates new bitcoin through the computational process known as mining and also maintains its own “strategic” bitcoin reserve — was originally called American Data Centers but rebranded in late March in a joint venture with Hut 8, another crypto miner.
As part of the deal, Hut 8 agreed to hand over all of its mining equipment in exchange for a majority interest in ADC. Hut 8’s shares were down 11 per cent on Tuesday.
American Bitcoin is among several crypto companies backed by the president’s sons, who also co-founded World Liberty Financial, which has issued billions of its own tokens and lists Donald Trump as co-founder emeritus.
The value of its WLF token has collapsed 86 per cent over the past year, according to CoinMarketCap data.
Trump Media & Technology Group, which runs the Truth Social app and is controlled by the president’s family, earlier this year said it planned to raise $1.5bn in fresh equity and another $1bn through convertible bonds to create a “bitcoin treasury”.
TMTG’s share price is down almost 70 per cent this year.
Shares in other so-called bitcoin treasury companies have tumbled in recent months amid the wider sell-off for hundreds of digital assets.
Michael Saylor’s Strategy — which pioneered the model and holds 650,000 bitcoin, equivalent to 3.1 per cent of the world’s total supply of the cryptocurrency — has fallen 40 per cent this year as investors have cooled on the company’s prolific issuance of shares, convertible debt and new preferred equity instruments to fund its bitcoin-buying spree.
Sam Altman has declared a “code red” at OpenAI to improve ChatGPT as the chatbot faces intense competition from rivals.
According to a report by tech news site the Information, the chief executive of the San Francisco-based startup told staff in an internal memo: “We are at a critical time for ChatGPT.”
OpenAI has been rattled by the success of Google’s latest AI model, Gemini 3, and is devoting more internal resources to improving ChatGPT.
Last month, Altman told employees that the launch of Gemini 3, which has outperformed rivals on various benchmarks, could create “temporary economic headwinds” for the company. He added: “I expect the vibes out there to be rough for a bit.”
OpenAI’s flagship product has 800 million weekly users but Google is also highly profitable due to its search business and has substantial data and financial resources to throw at its AI tools.
Sam Altman. Photograph: José Luis Magaña/AP
Marc Benioff, the chief executive of the $220bn (£166bn) software group Salesforce, wrote last month that he had switched allegiance to Gemini 3 and was “not going back” after trying Google’s latest AI release.
“I’ve used ChatGPT every day for 3 years. Just spent 2 hours on Gemini 3. I’m not going back. The leap is insane – reasoning, speed, images, video … everything is sharper and faster. It feels like the world just changed, again,” he wrote on X.
OpenAI is also delaying a foray into putting advertising in ChatGPT as it focuses on improving the chatbot, which celebrated its third birthday last month.
The head of ChatGPT, Nick Turley, marked the anniversary with a post on X pledging to break new ground with the product.
He wrote: “Our focus now is to keep making ChatGPT more capable, continue growing, and expand access around the world – while making it even more intuitive and personal. Thanks for an incredible three years. Lots more to do!”
Despite lacking the cash flow support enjoyed by rivals Google, Meta and Amazon, which is a big funder of competitor Anthropic, OpenAI has received substantial funding from the likes of the SoftBank investment group and Microsoft. In its latest valuation, OpenAI reached $500bn, up from $157bn last October.
OpenAI is loss-making and expects to end the year with annual revenues of more than $20bn, which Altman expects will grow to “hundreds of billion[s]” by 2030. The startup is committed to steep revenue growth after pledging to spend $1.4tn on datacentre costs to train and operate its AI systems over the next eight years.
skip past newsletter promotion
after newsletter promotion
“Based on the trends we are seeing of how people are using AI and how much of it they would like to use, we believe the risk of OpenAI of not having enough computing power is more significant and more likely than the risk of having too much,” said Altman last month.
Apple has also responded to increasingly intense competitive pressures in the sector by naming a new vice-president of AI. Amar Subramanya, a Microsoft executive, will replace John Giannandrea.
Applehas been slow to add AI features to its products in comparison with rivals such as Samsung, which have been quicker to refresh their devices with AI features.
Subramanya is joining Apple from Microsoft, where he most recently served as corporate vice-president of AI. Previously, Subramanya spent 16 years at Google, where his roles included the head of engineering for the Gemini assistant.
Earlier this year, Apple said AI improvements to its voice assistant Siri would be delayed until 2026.
Compute is the most valuable resource in the AI world we live in today. Nvidia. Google TPUs. Amazon Trainium. OpenAI and Broadcom’s partnership. Elon’s recent post about Tesla’s AI chips.
Designing the most performant chips for AI workloads sits at the heart of accelerating technological progress.
But major hurdles exist.
First, chip design is slow. It takes 12-24 months at mature nodes and 18-36 months at the leading edge for 5nm or 3nm.
Second, chip design is prohibitively expensive. It costs on average $200-250 million for 7nm, $450-500 million for 5nm, and $600-650 million for 3nm. Roughly 50-70% of that is human labor. Another 5-15% is Electronic Design Automation tooling spend in a market long dominated by Cadence and Synopsys, where each generates $5-6 billion in annual revenue and are worth approximately $90-100 billion in market cap.
AlphaChip caught my eye for these exact reasons. It gave us a peek at AI’s potential to transform the entire chip design process, showing we can cut the floorplanning step in physical design from months to hours.
What if we could extrapolate this and build AI to automate the entire flow, from architecture design to RTL to verification, all the way through physical design?
What if chip design took days, not two to three years? Every day is massively costly; some reports from August 2024 indicated that a multi-month Blackwell delay could result in more than $10 billion in lost revenue for 2025 alone. More importantly, imagine the revenue potential unlocked when new generations of chips are designed faster and shipped earlier.
What if each design didn’t cost hundreds of millions of dollars? What if chip companies didn’t need to operate large human teams on top of clunky EDA tooling?
And most exciting: what if we unlocked novel chip designs we might never have explored?
AlphaChip revealed an important human bias: in chip design, we tend to think in Manhattan grid-like structures. AlphaChip’s designs were different, more organic in shape, more like forms inspired by nature. So different, in fact, that humans wanted to reject them at first … Yet AlphaChip went on to shape four generations of the TPU.
We at Sequoia are so excited to partner with co-founders Anna Goldie and Azalia Mirhoseini, leading their very first round from the formation of Ricursive Intelligence. They pioneered AI for chip design by creating and leading the AlphaChip effort and are at the epicenter of this emerging AI for chip design ecosystem. They are visionaries with incredible clarity of thought, intensely ambitious, humble yet exceptionally accomplished, and real talent magnets who move, and inspire others to move, with urgency and velocity.
Anna and Azalia founded Ricursive Intelligence to build the frontier AI lab defining this category. In just the first weeks since company formation, they have assembled a team with the highest talent density you can imagine in the field.
Their core belief: chip design is the compute bottleneck, and progress in AI, hardware and infrastructure is capped by the speed and efficiency of silicon creation.
In their words: “If we get this right, it’s not just faster chip design cycles; it’s a fundamental expansion of what’s possible in hardware. Once chip design becomes fast and accessible, everyone will be able to customize. The automation here will unlock a flood of new hardware innovation.”
Anna and Azalia’s vision for Ricursive is to define a new movement, from “fabless” to “designless.” Fabless, meaning a company designing chips without owning expensive fabs, outsourcing production to foundries. Designless, meaning outsourcing not only manufacturing but the entire chip design process, taking an idea and converting it into a manufacturable design.
We envision a world where Ricursive helps any company design chips for its own workloads faster, more efficiently and more creatively than is possible today. In doing so, Ricursive can help revolutionize the most valuable resource in our era: compute. We could not be more excited to help build a true generational company in the making.
Dec 2 (Reuters) – Bobcat sued construction equipment rival Caterpillar (CAT.N), opens new tab in Texas federal court and at a U.S. trade tribunal on Tuesday, alleging that technology in many of Caterpillar’s dozers, excavators and other machinery infringes Bobcat’s patents.
Bobcat said in the complaints, opens new tab that Caterpillar’s construction equipment infringes patents covering technology for improved machine control and agility.
Sign up here.
Bobcat asked the court for an unspecified amount of monetary damages and the U.S. International Trade Commission for an order blocking imports of Caterpillar’s patent infringing equipment. It also filed related lawsuits against Caterpillar in German district court and at the European Union’s Unified Patent Court.
Spokespeople for Caterpillar did not immediately respond to a request for comment. Bobcat spokesperson Nadine Erckenbrack said the company seeks to “protect our patented technologies, defend fair competition, and safeguard the innovation and craftsmanship that have defined our company for more than 65 years.”
Bobcat, which specializes in compact construction equipment, was founded in North Dakota as Melroe Manufacturing Company in 1947 and acquired by South Korea-based Doosan (241560.KS), opens new tab
in 2007. The lawsuit said that Caterpillar copied Bobcat’s “skid-steer loader” technology for compact machinery.
“CAT’s use of so many of Bobcat’s patented technologies is consistent with its pattern and practice of identifying and emulating the key features in its competitors’ products,” Bobcat said in the Texas complaint.
The lawsuits are Doosan Bobcat North America Inc v. Caterpillar Inc, U.S. District Court for the Eastern District of Texas, Nos. 2:25-cv-01184 and 2:25-cv-01185; and In the Matter of Certain Skid-Steer Loaders, Compact Track Loaders, Excavators, Wheel Loaders, Dozers and Components Thereof, U.S. International Trade Commission.
For Bobcat: Sean Pak, Iman Lordgooei, Nathan Hamstra, Marc Kaplan and James Pak of Quinn Emanuel Urquhart & Sullivan
For Caterpillar: attorney information not yet available
Reporting by Blake Brittain in Washington
Our Standards: The Thomson Reuters Trust Principles., opens new tab
Purchase Licensing Rights
Blake Brittain reports on intellectual property law, including patents, trademarks, copyrights and trade secrets, for Reuters Legal. He has previously written for Bloomberg Law and Thomson Reuters Practical Law and practiced as an attorney.
