Category: 3. Business

NVIDIA (NASDAQ: NVDA) today reported record revenue for the fourth quarter ended January 25, 2026, of $68.1 billion, up 20% from the previous quarter and up 73% from a year ago. For fiscal 2026, revenue was $215.9 billion, up 65% from a year ago.

For the quarter, GAAP and non-GAAP gross margins were 75.0% and 75.2%, respectively. For fiscal 2026, GAAP and non-GAAP gross margins were 71.1% and 71.3%, respectively.

For the quarter, GAAP and non-GAAP earnings per diluted share were $1.76 and $1.62, respectively. For fiscal 2026, GAAP and non-GAAP earnings per diluted share were $4.90 and $4.77, respectively.

“Computing demand is growing exponentially — the agentic AI inflection point has arrived. Grace Blackwell with NVLink is the king of inference today — delivering an order-of-magnitude lower cost per token — and Vera Rubin will extend that leadership even further,” said Jensen Huang, founder and CEO of NVIDIA. “Enterprise adoption of agents is skyrocketing. Our customers are racing to invest in AI compute — the factories powering the AI industrial revolution and their future growth.”

During fiscal 2026, NVIDIA returned $41.1 billion to shareholders in the form of shares repurchased and cash dividends. As of the end of the fourth quarter, the company had $58.5 billion remaining under its share repurchase authorization.

NVIDIA will pay its next quarterly cash dividend of $0.01 per share on April 1, 2026, to all shareholders of record on March 11, 2026.

Q4 Fiscal 2026 Summary

GAAP
($ in millions, except earnings per share)		Q4 FY26			Q3 FY26			Q4 FY25			Q/Q		Y/Y
Revenue		$68,127			$57,006			$39,331			20 %		73 %
Gross margin		75.0 %			73.4 %			73.0 %			1.6 pts		2.0 pts
Operating expenses		$6,794			$5,839			$4,689			16 %		45 %
Operating income		$44,299			$36,010			$24,034			23 %		84 %
Net income		$42,960			$31,910			$22,091			35 %		94 %
Diluted earnings per share		$1.76			$1.30			$0.89			35 %		98 %
Non-GAAP
($ in millions, except earnings per share)		Q4 FY26			Q3 FY26			Q4 FY25			Q/Q		Y/Y
Revenue		$68,127			$57,006			$39,331			20 %		73 %
Gross margin		75.2 %			73.6 %			73.5 %			1.6 pts		1.7 pts
Operating expenses		$5,102			$4,215			$3,378			21 %		51 %
Operating income		$46,107			$37,752			$25,516			22 %		81 %
Net income		$39,552			$31,767			$22,066			25 %		79 %
Diluted earnings per share		$1.62			$1.30			$0.89			25 %		82 %

Fiscal 2026 Summary

GAAP
($ in millions, except earnings per share)						FY26				FY25				Y/Y
Revenue						$215,938				$130,497				65 %
Gross margin							71.1 %				75.0 %			(3.9) pts
Operating expenses						$23,076				$16,405				41 %
Operating income						$130,387				$81,453				60 %
Net income						$120,067				$72,880				65 %
Diluted earnings per share						$4.90				$2.94				67 %
Non-GAAP
($ in millions, except earnings per share)						FY26				FY25				Y/Y
Revenue						$215,938				$130,497				65 %
Gross margin							71.3 %				75.5 %			(4.2) pts
Operating expenses						$16,694				$11,716				42 %
Operating income						$137,300				$86,789				58  %
Net income						$116,997				$74,265				58 %
Diluted earnings per share						$4.77				$2.99				60 %

Outlook

Beginning in the first quarter of fiscal 2027, NVIDIA will include stock-based compensation expense in non-GAAP financial measures. Stock-based compensation is a foundational component of NVIDIA’s compensation program to attract and retain world-class talent.

NVIDIA’s outlook for the first quarter of fiscal 2027 is as follows:

• Revenue is expected to be $78.0 billion, plus or minus 2%. NVIDIA is not assuming any Data Center compute revenue from China in its outlook.
• GAAP and non-GAAP gross margins are expected to be 74.9% and 75.0%, respectively, plus or minus 50 basis points, inclusive of a 0.1% impact from stock-based compensation expense.
• GAAP and non-GAAP operating expenses are expected to be approximately $7.7 billion and $7.5 billion, respectively, inclusive of $1.9 billion of stock-based compensation expense.
  
   

For the full year fiscal 2027, GAAP and non-GAAP tax rates are expected to be between 17.0% and 19.0%, excluding any discrete items and material changes to NVIDIA’s tax environment.

Highlights

Data Center

• Fourth-quarter revenue was a record $62.3 billion, up 22% from the previous quarter and up 75% from a year ago, driven by the major platform shifts — accelerated computing and AI. Full-year revenue rose 68% to a record $193.7 billion.
• Unveiled the NVIDIA Rubin platform, comprising six new chips to deliver up to a 10x reduction in inference token cost, compared with the NVIDIA Blackwell platform; cloud providers Amazon Web Services (AWS), Google Cloud, Microsoft Azure and Oracle Cloud Infrastructure will be among the first to deploy Vera Rubin-based instances.
• Announced that the NVIDIA BlueField^®-4 data processor powers the NVIDIA Inference Context Memory Storage Platform, a new class of AI-native storage infrastructure for the next frontier of AI.
• Announced a multiyear, multigenerational strategic partnership with Meta spanning on-premises, cloud and AI infrastructure, including the large-scale deployment of NVIDIA CPUs, networking and millions of NVIDIA Blackwell and Rubin GPUs.
• Revealed that NVIDIA Blackwell Ultra delivers up to 50x better performance and 35x lower cost for agentic AI compared with the NVIDIA Hopper platform, according to new SemiAnalysis InferenceX benchmark results.
• Expanded AWS partnership with new technology integrations across interconnect technology, cloud infrastructure, open models and physical AI.
• Revealed that leading inference providers, including Baseten, DeepInfra, Fireworks AI and Together AI, cut AI costs by up to 10x with open source models on NVIDIA Blackwell.
• Debuted the NVIDIA Nemotron™ 3 family of open models, data and libraries designed to power transparent, efficient and specialized agentic AI development across industries; released new open models, data and tools for agentic AI, physical AI and autonomous vehicle development.
• Announced an investment and deep technology partnership with Anthropic, which is scaling its Claude model on Microsoft Azure, powered by NVIDIA systems.
• Entered into a non-exclusive licensing agreement with Groq to accelerate AI inference at global scale.
• Strengthened a collaboration with CoreWeave to accelerate the buildout of more than 5 gigawatts of AI factories by 2030.
• Announced an expanded strategic partnership with Synopsys to revolutionize engineering and design across industries.
• Announced a co-innovation AI lab with Lilly to reinvent drug discovery in the age of AI.
• Announced a major expansion of NVIDIA BioNeMo™, an open development platform that enables lab-in-the-loop workflows to develop breakthroughs in AI-driven biology and drug discovery.
• Joined the U.S. Department of Energy’s Genesis Mission as a private industry partner to support U.S. AI leadership in key areas including energy, scientific research and national security.
• Launched the NVIDIA Earth-2 family of open models — the world’s first fully open, accelerated set of models and tools for AI weather.
• Revealed that India’s global systems integrators Infosys, Persistent, Tech Mahindra and Wipro are building the next wave of enterprise agents with NVIDIA AI.
• Partnered with global industrial software leaders Cadence, Siemens and Synopsys and India’s largest manufacturers to drive India’s AI boom using applications accelerated by NVIDIA CUDA-X™ and NVIDIA Omniverse™ libraries.
  
   

Gaming and AI PC

• Fourth-quarter Gaming revenue was $3.7 billion, up 47% from a year ago, driven by strong Blackwell demand, and down 13% from the previous quarter as channel inventory naturally moderated following a season of strong holiday demand. Full-year revenue rose 41% to a record $16.0 billion.
• Announced NVIDIA DLSS 4.5, delivering major AI-powered advances in graphics quality.
• Launched NVIDIA G-SYNC^® Pulsar, extending the ultimate gaming display platform with new levels of motion clarity in esports.
• Advanced NVIDIA RTX™ AI performance and adoption, delivering up to 35% faster large language model inference in leading AI PC frameworks and up to 3x performance in AI-generated visuals.
  
   

Professional Visualization

• Fourth-quarter revenue was $1.3 billion, up 74% from the previous quarter and up 159% from a year ago, driven by exceptional demand for Blackwell. Full-year revenue rose 70% to a record $3.2 billion.
• Launched the NVIDIA RTX PRO™ 5000 72GB Blackwell GPU to power larger models and agentic workflows.
• Expanded global availability of NVIDIA DGX Spark™ for the latest open models and delivered updates for improved performance.
  
   

Automotive and Robotics

• Fourth-quarter Automotive revenue was $604 million, up 2% from the previous quarter and up 6% from a year ago, driven by continued adoption of NVIDIA’s self-driving platforms. Full-year revenue rose 39% to a record $2.3 billion.
• Unveiled the NVIDIA Alpamayo family of open AI models, simulation tools and datasets designed to accelerate the next era of safe, reasoning‑based autonomous vehicle (AV) development.
• Partnered with Mercedes-Benz on the all-new Mercedes-Benz CLA, which introduces enhanced level 2 driver assistance powered by NVIDIA DRIVE AV software, AI infrastructure and accelerated compute.
• Announced that the NVIDIA DRIVE Hyperion™ ecosystem is expanding to include tier 1 suppliers, automotive integrators and sensor partners including Aeva, AUMOVIO, Astemo, Arbe, Bosch, Hesai, Magna, Omnivision, Quanta, Sony and ZF Group.
• Announced new NVIDIA Cosmos™ and NVIDIA Isaac™ GR00T open models, frameworks and AI infrastructure for physical AI; global industry leaders including Boston Dynamics, Caterpillar, Franka Robotics, Humanoid, LG Electronics and NEURA Robotics are using the NVIDIA robotics stack.
• Expanded a strategic partnership with Siemens to build the industrial AI operating system.
• Announced a strategic partnership with Dassault Systèmes to build an industrial AI platform powering virtual twins.
  
   

CFO Commentary

Commentary on the quarter by Colette Kress, NVIDIA’s executive vice president and chief financial officer, is available at https://investor.nvidia.com.

Conference Call and Webcast Information

NVIDIA will conduct a conference call with analysts and investors to discuss its fourth quarter and fiscal 2026 financial results and current financial prospects today at 2 p.m. Pacific time (5 p.m. Eastern time). A live webcast (listen-only mode) of the conference call will be accessible at NVIDIA’s investor relations website, https://investor.nvidia.com. The webcast will be recorded and available for replay until NVIDIA’s conference call to discuss its financial results for its first quarter of fiscal 2027.

Non-GAAP Measures

To supplement NVIDIA’s condensed consolidated financial statements presented in accordance with GAAP, the company uses non-GAAP measures of certain components of financial performance. These non-GAAP measures include non-GAAP gross profit, non-GAAP gross margin, non-GAAP operating expenses, non-GAAP operating income, non-GAAP other income (expense), net, non-GAAP net income, non-GAAP net income, or earnings, per diluted share, and free cash flow. For NVIDIA’s investors to be better able to compare its current results with those of previous periods, the company has shown a reconciliation of GAAP to non-GAAP financial measures. The reconciliations for fiscal years 2025 and 2026 adjust the related GAAP financial measures to exclude stock-based compensation expense, acquisition-related and other costs, other, gains/losses from non-marketable and publicly-held equity securities, net, interest expense related to amortization of debt discount, and the associated tax impact of these items where applicable. Beginning in the first quarter of fiscal 2027, NVIDIA’s non-GAAP financial measures will no longer exclude stock-based compensation expense. Free cash flow is calculated as GAAP net cash provided by operating activities less both purchases related to property and equipment and intangible assets and principal payments on property and equipment and intangible assets. NVIDIA believes the presentation of its non-GAAP financial measures enhances the user’s overall understanding of the company’s historical financial performance. The presentation of the company’s non-GAAP financial measures is not meant to be considered in isolation or as a substitute for the company’s financial results prepared in accordance with GAAP, and the company’s non-GAAP measures may be different from non-GAAP measures used by other companies.

NVIDIA CORPORATION
CONDENSED CONSOLIDATED STATEMENTS OF INCOME
(In millions, except per share data)
(Unaudited)
			Three Months Ended								Twelve Months Ended
			January 25,				January 26,				January 25,				January 26,
				2026				2025				2026				2025
Revenue			$	68,127			$	39,331			$	215,938			$	130,497
Cost of revenue				17,034				10,608				62,475				32,639
Gross profit				51,093				28,723				153,463				97,858
Operating expenses
	Research and development			5,512				3,714				18,497				12,914
	Sales, general and administrative			1,282				975				4,579				3,491
		Total operating expenses		6,794				4,689				23,076				16,405
Operating income				44,299				24,034				130,387				81,453
	Interest income			568				511				2,300				1,786
	Interest expense			(74	)			(61	)			(259	)			(247	)
	Other income, net			5,604				733				9,022				1,034
		Total other income, net		6,098				1,183				11,063				2,573
Income before income tax				50,397				25,217				141,450				84,026
Income tax expense				7,437				3,126				21,383				11,146
Net income			$	42,960			$	22,091			$	120,067			$	72,880
Net income per share:
	Basic		$	1.77			$	0.90			$	4.93			$	2.97
	Diluted		$	1.76			$	0.89			$	4.90			$	2.94
Weighted average shares used in per share computation:
	Basic			24,304				24,489				24,359				24,555
	Diluted			24,432				24,706				24,514				24,804

 

NVIDIA CORPORATION
CONDENSED CONSOLIDATED BALANCE SHEETS
(In millions)
(Unaudited)
				January 25,			January 26,
				2026			2025
ASSETS
Current assets:
	Cash, cash equivalents and marketable securities			$	62,556		$	43,210
	Accounts receivable, net				38,466			23,065
	Inventories				21,403			10,080
	Prepaid expenses and other current assets				3,180			3,771
		Total current assets			125,605			80,126
Property and equipment, net					10,383			6,283
Operating lease assets					2,867			1,793
Goodwill					20,832			5,188
Intangible assets, net					3,306			807
Deferred income tax assets					13,258			10,979
Non-marketable equity securities					22,251			3,387
Other assets					8,301			3,038
		Total assets		$	206,803		$	111,601
LIABILITIES AND SHAREHOLDERS’ EQUITY
Current liabilities:
	Accounts payable			$	9,812		$	6,310
	Accrued and other current liabilities				21,352			11,737
	Short-term debt				999			–
		Total current liabilities			32,163			18,047
Long-term debt					7,469			8,463
Long-term operating lease liabilities					2,572			1,519
Other long-term liabilities					7,306			4,245
		Total liabilities			49,510			32,274
Shareholders’ equity					157,293			79,327
		Total liabilities and shareholders’ equity		$	206,803		$	111,601

 

NVIDIA CORPORATION
CONDENSED CONSOLIDATED STATEMENTS OF CASH FLOWS
(In millions)
(Unaudited)
			Three Months Ended								Twelve Months Ended
			January 25,				January 26,				January 25,				January 26,
				2026				2025				2026				2025
Cash flows from operating activities:
Net income			$	42,960			$	22,091			$	120,067			$	72,880
Adjustments to reconcile net income to net cash
provided by operating activities:
	Stock-based compensation expense			1,633				1,321				6,386				4,737
	Depreciation and amortization			811				543				2,843				1,864
	Gains on non-marketable equity securities and publicly-held equity securities, net			(5,491	)			(727	)			(8,918	)			(1,030	)
	Deferred income taxes			611				(598	)			(1,424	)			(4,477	)
	Other			(9	)			(138	)			(287	)			(502	)
Changes in operating assets and liabilities, net of acquisitions:
	Accounts receivable			(5,073	)			(5,370	)			(15,399	)			(13,063	)
	Inventories			(1,621	)			(2,424	)			(11,324	)			(4,781	)
	Prepaid expenses and other assets			(281	)			331				577				(395	)
	Accounts payable			1,064				867				3,096				3,357
	Accrued and other current liabilities			1,053				360				5,257				4,278
	Other long-term liabilities			533				372				1,844				1,221
Net cash provided by operating activities				36,190				16,628				102,718				64,089
Cash flows from investing activities:
	Proceeds from sales of marketable securities			14,670				177				15,157				495
	Proceeds from maturities of marketable securities			2,246				1,710				11,226				11,195
	Proceeds from sales of non-marketable equity securities			12				–				84				171
	Purchases of marketable securities			(20,540	)			(7,010	)			(40,616	)			(26,575	)
	Purchases of non-marketable equity securities			(12,800	)			(478	)			(17,502	)			(1,486	)
	Groq, Inc.			(13,000	)			–				(13,000	)			–
	Purchases related to property and equipment and intangible assets			(1,284	)			(1,077	)			(6,042	)			(3,236	)
	Acquisitions, net of cash acquired			(165	)			(542	)			(1,535	)			(1,007	)
	Other			–				22				–				22
Net cash used in investing activities				(30,861	)			(7,198	)			(52,228	)			(20,421	)
Cash flows from financing activities:
	Proceeds related to employee stock plans			–				–				644				490
	Payments related to repurchases of common stock			(3,815	)			(7,810	)			(40,086	)			(33,706	)
	Payments related to employee stock plan taxes			(2,139	)			(1,861	)			(7,948	)			(6,930	)
	Dividends paid			(243	)			(245	)			(974	)			(834	)
	Principal payments on property and equipment and intangible assets			(4	)			(32	)			(101	)			(129	)
	Repayment of debt			–				–				–				(1,250	)
	Other			(9	)			–				(9	)			–
Net cash used in financing activities				(6,210	)			(9,948	)			(48,474	)			(42,359	)
Change in cash and cash equivalents				(881	)			(518	)			2,016				1,309
Cash and cash equivalents at beginning of period				11,486				9,107				8,589				7,280
Cash and cash equivalents at end of period			$	10,605			$	8,589			$	10,605			$	8,589
Supplemental disclosures of cash flow information:
Cash paid for income taxes, net			$	6,979			$	4,129			$	20,288			$	15,118

 

NVIDIA CORPORATION
RECONCILIATION OF GAAP TO NON-GAAP FINANCIAL MEASURES
(In millions, except per share data)
(Unaudited)
			Three Months Ended												Twelve Months Ended
			January 25,				October 26,				January 26,				January 25,				January 26,
				2026				2025				2025				2026				2025
GAAP cost of revenue			$	17,034			$	15,157			$	10,608			$	62,475			$	32,639
GAAP gross profit			$	51,093			$	41,849			$	28,723			$	153,463			$	97,858
GAAP gross margin				75.0%				73.4%				73.0%				71.1%				75.0%
	Acquisition-related and other costs (A)			48				48				118				267				472
	Stock-based compensation expense (B)			69				70				53				261				178
	Other			(1	)			–				–				3				(3	)
Non-GAAP cost of revenue			$	16,918			$	15,039			$	10,437			$	61,944			$	31,992
Non-GAAP gross profit			$	51,209			$	41,967			$	28,894			$	153,994			$	98,505
Non-GAAP gross margin**				75.2%				73.6%				73.5%				71.3%				75.5%
GAAP operating expenses			$	6,794			$	5,839			$	4,689			$	23,076			$	16,405
	Stock-based compensation expense (B)			(1,564	)			(1,585	)			(1,268	)			(6,125	)			(4,559	)
	Acquisition-related and other costs (A)			(90	)			(39	)			(43	)			(204	)			(130	)
	Other			(38	)			–				–				(53	)			–
Non-GAAP operating expenses			$	5,102			$	4,215			$	3,378			$	16,694			$	11,716
GAAP operating income			$	44,299			$	36,010			$	24,034			$	130,387			$	81,453
	Total impact of non-GAAP adjustments to operating income			1,808				1,742				1,482				6,913				5,336
Non-GAAP operating income			$	46,107			$	37,752			$	25,516			$	137,300			$	86,789
GAAP total other income, net			$	6,098			$	1,926			$	1,183			$	11,063			$	2,573
	Gains from non-marketable equity securities and publicly-held equity securities, net			(5,491	)			(1,354	)			(727	)			(8,918	)			(1,030	)
	Other (C)			13				1				1				16				4
Non-GAAP total other income, net			$	620			$	573			$	457			$	2,161			$	1,547
GAAP net income			$	42,960			$	31,910			$	22,091			$	120,067			$	72,880
	Total pre-tax impact of non-GAAP adjustments			(3,670	)			389				756				(1,989	)			4,310
	Income tax impact of non-GAAP adjustments (D)			262				(532	)			(781	)			(1,129	)			(2,925	)
	Tax expense from OBBBA*			–				–				–				48				–
Non-GAAP net income**			$	39,552			$	31,767			$	22,066			$	116,997			$	74,265
Diluted net income per share
	GAAP		$	1.76			$	1.30			$	0.89			$	4.90			$	2.94
	Non-GAAP**		$	1.62			$	1.30			$	0.89			$	4.77			$	2.99
Weighted average shares used in diluted net income per share computation				24,432				24,483				24,706				24,514				24,804
GAAP net cash provided by operating activities			$	36,190			$	23,750			$	16,628			$	102,718			$	64,089
	Purchases related to property and equipment and intangible assets			(1,284	)			(1,637	)			(1,077	)			(6,042	)			(3,236	)
	Principal payments on property and equipment and intangible assets			(4	)			(24	)			(32	)			(101	)			(129	)
Free cash flow			$	34,902			$	22,089			$	15,519			$	96,575			$	60,724
*Tax expense included represents impact from OBBBA (One Big Beautiful Bill Act).
**Includes H20 charges/(releases), net, which were $4.5 billion and ($180 million) for the first and second quarter of fiscal 2026, respectively, and insignificant for both the third and fourth quarter of fiscal 2026.
(A) Acquisition-related and other costs are comprised of amortization of intangible assets, transaction costs, and certain compensation charges and are included in the following line items:
			Three Months Ended												Twelve Months Ended
			January 25,				October 26,				January 26,				January 25,				January 26,
				2026				2025				2025				2026				2025
	Cost of revenue		$	48			$	48			$	118			$	267			$	472
	Research and development		$	83			$	35			$	27			$	176			$	79
	Sales, general and administrative		$	7			$	4			$	16			$	28			$	51
(B) Stock-based compensation consists of the following:
			Three Months Ended												Twelve Months Ended
			January 25,				October 26,				January 26,				January 25,				January 26,
				2026				2025				2025				2026				2025
	Cost of revenue		$	69			$	70			$	53			$	261			$	178
	Research and development		$	1,217			$	1,206			$	955			$	4,676			$	3,423
	Sales, general and administrative		$	347			$	379			$	313			$	1,449			$	1,136
(C) Interest expense related to acquisition consideration discount to be paid in the future and amortization of debt discount.
(D) Income tax impact of non-GAAP adjustments, including the recognition of excess tax benefits or deficiencies related to stock-based compensation under GAAP accounting standard (ASU 2016-09).

 

NVIDIA CORPORATION
RECONCILIATION OF GAAP TO NON-GAAP OUTLOOK
		Q1 FY27 Outlook
		($ in millions)
GAAP gross margin			74.9%
	Impact of acquisition-related costs and other costs		0.1%
Non-GAAP gross margin*			75.0%
GAAP operating expenses		$	7,700
	Acquisition-related costs and other costs		(200	)
Non-GAAP operating expenses*		$	7,500
*Beginning in the first quarter of fiscal 2027, NVIDIA will include stock-based compensation expense in its non-GAAP financial measures. Stock-based compensation expense for the first quarter of fiscal 2027 is expected to have a 0.1% impact on non-GAAP gross margin and $1.9 billion in non-GAAP operating expenses.

A group of 14 law firms representing nearly 20,000 plaintiffs is seeking to intervene in Bayer’s proposed class action settlement of Roundup litigation, citing concerns that the deal will not be fair to cancer sufferers.

The group filed both a motion to intervene and a motion for an extension of time for court preliminary approval of the deal in St Louis city circuit court in Missouri late on February 24.

The law firms say the deal appears “unprecedented” and raises multiple “red flags”.

“It is hard to escape the impression that the proposed settlement would give Monsanto everything it desires – a near-complete release of liability for Monsanto and its parent company, Bayer AG – while giving inadequate consideration to many putative class members, who would surrender their substantive rights in exchange for settlement offers that may never result in payment,” the law firms state in their motion.

Bayer and a different group of plaintiffs’ lawyers filed the settlement proposal with the court on 17 February, with a provision to seek preliminary court approval within a 15-day period.

But the opposing firms are seeking a 60-day extension of that “fast track” time frame, saying the “sheer scale and impact of the proposed settlement, together with concerns raised by its terms and how it was negotiated, warrant broader public participation and scrutiny”.

Bayer announced the $7.25bn proposed class action settlement on February, proposing to pay amounts ranging from $10,000 to $165,000 to users of its glyphosate-based weed-killing products who have non-Hodgkin lymphoma (NHL), a type of blood cancer, or who develop it in the next several years.

Bayer, which maintains that its glyphosate herbicides do not cause cancer, has faced more than 100,000 lawsuits since buying the Roundup maker Monsanto in 2018. The company has so far paid billions of dollars in settlements and jury verdicts to tens of thousands of people suffering from NHL that they blame on exposure to Roundup and other Monsanto glyphosate-based herbicide brands.

The law firms seeking to intervene in the proposed settlement alleged that it “heavily favors” occupational Roundup users such as farmers or commercial landscapers over residential users.

Under the proposed payment schedule, an occupational claimant diagnosed before age 60 with aggressive NHL could receive, on average, $165,000, while a residential user with the same traits would average $40,000, the motion to intervene points out.

Additionally, they object to Bayer’s request that the court stay the thousands of lawsuits pending against the company in Missouri.

The law firms say the 600-page settlement agreement was “negotiated behind closed doors” and does not adequately represent the interests of the plaintiffs they represent.

In response to the opposition filings, Bayer said in a statement that it is “not surprised” and fully expects a “robust debate” about the settlement proposal.

“We remain confident that the long-term and well-financed proposed class settlement plan, which is supported by plaintiff law firms representing thousands of potential class members, is fair to all claimants, and warrants approval by the court,” the company said.

The plaintiffs’ legal team that negotiated the settlement proposal with Bayer said in a statement: “It is obvious that the proposed intervenors’ have reviewed the agreement over the last week, and are hopefully working as hard to communicate its terms to their clients as they are trying to delay compensation for the tens of thousands of Roundup victims who have waited a decade for justice.”

The team said in their statement that without a settlement, plaintiffs face “the risks of Monsanto filing for bankruptcy”, among other events that could hinder their ability to collect from the company in the future.

Bayer is hoping a looming US supreme court review will incentivize plaintiffs to agree to the settlement because if Bayer prevails in the case, future lawsuits against the company could be severely hampered.

Bayer maintains that federal law pre-empts failure-to-warn claims against the company. Because the Environmental Protection Agency (EPA) has not established a definitive cancer link and has approved labels with no cancer warning, lawsuits claiming the company should have provided a cancer warning must be barred, the company says.

The supreme court agreed to weigh in on that issue and has set a hearing for 27 April.

The company filed its opening brief in the case on Monday. In its filing with the high court, Bayer cites support from Donald Trump and US regulators while renewing a threat to stop sales of glyphosate-based herbicides to farmers if it does not prevail with the justices.

• This story is co-published with the New Lede, a journalism project of the Environmental Working Group

Introduction

Participant recruitment is a critical component of medical research studies, and the rate of participation varies significantly depending on the type of study, the setting, and the population involved. As an example, recent data suggest that enrollment rate in cancer trials is around 6.3% to 7.1% [,]. Enrollment rates may also differ among different populations; specifically, they can be lower among minority, pediatric, and geriatric populations [,]. Poor recruitment often results in underpowered studies, which lack the necessary sample size to detect meaningful differences between distinct groups. This can lead to statistically nonsignificant results even when there are clinically relevant effects []. Despite these efforts and the growing adoption of digital recruitment tools, recent analyses confirm that these challenges persist with participation patterns continuing to vary by age, sex, and socioeconomic status (SES) []. This underscores a critical point that digital methods may not fully mitigate traditional barriers and could introduce new inequities that influence who participates in research.

Digital tools, including social media platforms, mobile apps, electronic health records, patient portals, and electronic consent, have expanded the reach of research recruitment efforts by enabling targeted outreach to specific demographics and geographic areas [-]. These approaches can engage populations previously hard to reach through traditional methods, but evaluations report inconsistent effects on diversity and efficiency across settings []. A recent systematic review cataloged the spectrum of digital technologies deployed for recruitment and highlighted the still-limited evidence that any one approach reliably improves inclusion of underrepresented groups [].

A central challenge for digital recruitment is the digital divide, defined as persistent inequities in broadband access, device availability, and digital literacy that map closely onto socioeconomic, geographic, and age-related lines []. These inequities can depress response rates and create systematic attrition in specific subgroups even when outreach is delivered digitally. Recent studies and reviews caution that digitalization can reproduce or widen participation gaps if considerations are not built into recruitment strategies []. Within health care systems, portal-based recruitment shows promise but exhibits differential response patterns across patient characteristics, underscoring the need to track impacts on disparity as programs scale. Moreover, scoping reviews in specific domains suggest that mixed, multichannel recruitment strategies (digital and offline) may improve inclusivity compared with relying on a single modality []. These disparities are often more pronounced among disadvantaged groups, including racial and ethnic minorities and older adults, who may have limited access to the internet and lower digital literacy levels [-]. For instance, studies have shown that older adults and African American patients are less likely to use digital health portals compared to their younger and White counterparts, highlighting a significant gap in technology use []. Here, the term “White” reflects the racial classification used in the original study, which reported race and ethnicity separately and did not provide ethnicity‑specific breakdowns. Similarly, individuals from lower SES neighborhoods often have reduced access to the internet and lower health literacy, which can hinder their ability to engage with digital health technologies effectively []. While digital tools have the potential to improve the representativeness of trial participants, there is limited evidence supporting their effectiveness in recruiting underrepresented groups [,]. This underscores the need for targeted interventions and strategies to bridge the digital divide and ensure equitable access to digital health resources [,].

Digital recruitment has become especially popular with the increased interest in artificial intelligence (AI) in health care, which requires large and representative datasets to train. For example, the Bridge2AI-Voice program has the explicit goal to create “an ethically sourced flagship dataset to enable future research in artificial intelligence” []. Such datasets are typically assembled through digital recruitment workflows, underscoring the need to understand how these methods shape participant diversity. Our group has similarly endeavored to obtain a large bank of speech recordings focused on neurological disorders, with the goal of subsequently using these data to train AI models, through a primarily digital recruitment approach. Given concerns about the digital divide and its potential impact on representativeness, the broader speech capture study offered the ideal setting to formally investigate these issues in depth. This study aims to characterize recruitment and attrition patterns in a remote neurology cohort, quantify associations with sociodemographic factors (age, sex, neighborhood deprivation, housing-based SES, and urbanicity), and identify drop-off points to inform strategies for digital recruitment for all participants.

Methods

Study Design and Setting

This analysis was nonexperimental and observational, using a longitudinal cohort design to evaluate sociodemographic factors influencing recruitment and attrition in a remote speech capture study. This study was conducted in accordance with the American Psychological Association (APA) Journal Article Reporting Standards (JARS; American Psychological Association, 2018; refer to for the completed JARS-Quant checklist) []. The overarching speech capture study aims to remotely collect speech samples from patients with neurologic diseases to develop an easy-to-use and cost-effective screening tool for predicting disease progression. The parent speech capture study extends beyond the March-July 2024 analysis window; for this report, we used data collected within that period. The research was conducted at Mayo Clinic.

Inclusion and Exclusion Criteria

Individuals were eligible if they met the following criteria: (1) adults aged 18 years or older, (2) residing in the United States, and (3) able to communicate in English via spoken language.

Data Collection and Participant Characteristics

Patient identification was conducted using the Mayo Clinic Electronic Health Record (Epic, developed by Epic Systems Corporation), and invitations to complete an eligibility survey were sent via the patient portal using Qualtrics (developed by Qualtrics, LLC) []. The eligibility survey also assessed participants’ understanding of the study and asked whether they had a legally authorized representative responsible for financial or health care decisions. Once interest and eligibility were confirmed, participants received a PDF of the consent form to sign electronically via AdobeSign (Adobe Inc) through Mayo Clinic–developed Participant Tracking System (PTrax). PTrax is an institutional research software program designed to streamline informed consent processes, manage participant status, track enrollments and accruals, and provide reporting and analytics. Recruitment followed a convenience sampling approach, targeting patients with upcoming neurology appointments accessible via the institutional patient portal. After consent was obtained, participants were sent a secure message in the patient portal with a link to the speech recording platform and instructions.

We exported longitudinal record data for all eligible patients invited to the speech capture study between March and July 2024. This resulted in a total sample of 5846 participants, reflecting all patients meeting the inclusion criteria during the recruitment window. Race was extracted from the electronic health record, where “White” reflects a race category and is distinct from ethnicity. Ethnicity (eg, Hispanic or Latinx origin) was not included in the dataset for this secondary analysis; therefore, individuals categorized as “White” may include participants of diverse ethnic backgrounds. The invited cohort had a median (IQR) age of 63 (48-72) years, with 56.2% (3283/5846) female and 93.7% (5478/5846) identifying as White. Urbanicity distribution was 56.5% (3303/5846) urban, 23.3% (1363/5846) rural, and 20.2% (1180/5846) urban cluster. This secondary analysis included all eligible patients invited during the recruitment window of the overarching speech capture study. No a priori power calculation was performed because the analysis was observational and descriptive. Precision was conveyed using 95% CIs for medians and IQRs. Data were drawn from Epic (demographics), Qualtrics (survey responses), PTrax (consent tracking), and the speech recording platform (task completion).

Measures and Covariates

SES was assessed using the Housing-based Socioeconomic Status (HOUSES) index and the Area Deprivation Index (ADI) national rank. The HOUSES index is a practical and adaptable tool for assessing SES using housing data. It effectively correlates with traditional SES measures and predicts various health outcomes []. Higher HOUSES index values indicate higher SES, while lower values indicate lower SES. The ADI national rank measures neighborhood socioeconomic disadvantage based on factors such as income, education, employment, and housing characteristics [,]. Higher scores indicate greater disadvantage. ADI is widely used in public health research to assess the impact of socioeconomic context on health outcomes.

Primary outcomes included study completion and time to enrollment. Exposures and predictors included age, sex, urbanicity (urban, rural, or urban cluster), ADI national rank, and the HOUSES index. Potential confounders considered were device type and urbanicity. Urbanicity was treated as an exposure in primary analyses and as a potential confounder in models examining associations between socioeconomic indices and participation outcomes. The collected data included participants’ age, sex assigned at birth (hereafter referred to as sex), date of invitation, recruitment process checkpoints and dates, and devices used for speech recording. Residence information was used to measure the HOUSES and ADI index and to classify participants by urbanicity (urban, rural, or urban cluster).

Quality of Measurements

This analysis relied exclusively on data exported from established operational systems (Epic for demographics, Qualtrics for eligibility responses, PTrax for participant status tracking, and the recording platform for task completion). No new training, instrumentation, or measurement procedures were implemented specifically for this analysis. All timestamps and statuses were generated by the source systems as part of routine workflows. Data were deidentified before analysis.

Analytic Strategy

The longitudinal time series data for each patient undergoing the participation process were standardized to align with specific checkpoints, as demonstrated in . At each step, it was possible for patients not to respond to the research coordination team, which was defined as the “No Response” stage. After providing consent, participants were also free to withdraw it at any time. While most individuals followed one of the typical pathways depicted in , there were 83 cases (1.4% of the total 5846 cases) in which participants deviated from these paths, necessitating intervention by research coordinators. These atypical cases were primarily attributed to personal circumstances or the involvement of a legally authorized representative.

‎

Kruskal-Wallis and Wilcoxon rank-sum tests were used to compare the median age, socioeconomic indices, and time taken to reach different steps of the study. These nonparametric tests were selected due to non-normal distribution of key variables, as confirmed by the Anderson-Darling test (P<.001). They are appropriate for comparing medians across groups and handling skewed data. Analysis was conducted at 2 levels: at each checkpoint and through an end-to-end investigation of participants who completed the study. For each path, a participant could take at each checkpoint, age, ADI national rank, and HOUSES index were compared to identify statistically significant differences, using a 2‑sided alpha level of .05. At the end-to-end level, additional comparisons were made across sex, urbanicity (urban, rural, and urban cluster), ADI national rank, and HOUSES index to evaluate differences in time taken to complete participation, from initial invitation to accrual, and whether participants completed the study, regardless of the path taken. Analyses were conducted using BlueSky Statistics version 10.3.4 (developed by BlueSky Statistics, LLC) and Python SciPy package version 1.16. Exact P values are reported, with P<.001 used where appropriate. No effect sizes were calculated for analyses.

Data Diagnostics

Since both Kruskal-Wallis and Wilcoxon rank-sum tests require at least 5 data points in each comparison category, pathways with fewer than 5 participants without missing values were excluded from the analysis. This threshold was applied to ensure statistical validity and avoid unreliable comparisons in small subgroups, as no randomized assignments or masking strategies were implemented. To further assess the nature of missing data across key sociodemographic variables, including age, ADI national rank, and HOUSES index, we conducted Little’s missing completely at random (MCAR) test []. This test evaluates whether data are MCAR, which informs the appropriate handling strategy. Based on the results, we adopted a pairwise deletion approach for statistical analyses, allowing each test to include all available cases for the specific variable of interest. This method was chosen to preserve sample size and maintain statistical power, particularly in subgroups with limited data. This method was chosen to preserve sample size and maintain statistical power, particularly in subgroups with limited data. No outlier removal or variable transformations were applied.

Ethical Considerations

The overarching speech capture study was reviewed and approved by the Mayo Clinic Institutional Review Board (#22-002430). Informed consent was obtained electronically for the primary speech capture study via Adobe Sign. For this secondary analysis of recruitment and attrition patterns, which used deidentified data from the primary study, the institutional review board determined that additional approval was not required. All data were deidentified prior to analysis to ensure participant confidentiality. No compensation was provided for participation. No identifiable images of participants are included in the manuscript or any supplementary materials.

Results

Overview of the Invited Cohort

A total of 5846 patients were invited to participate in the study between March and July 2024. Of these participants, 3283 (56.2%) were female, 2560 (43.8%) were male, and 3 (0.1%) were unknown. Most participants (5478/5846, 93.7%) identified as White. The age distribution ranged from 18 to 96 years, with a median (IQR) age of 63 (48-72) years (N=5846; 95% CI 62‐63). The narrow CI suggests high precision in estimating the median age of the invited cohort. Regarding urbanicity, 56.5% (3303/5846) of invited participants resided in urban areas, 23.3% (1363/5846) in rural areas, and 20.2% (1180/5846) in urban clusters (). Following accrual completion, participants used various devices to access the recording platform. Apple-based mobile devices (iPhone and iPad) were most frequently used (141/415, 34.0%), followed by Windows-based computers (134/415, 32.3%), Apple-based computers (82/415, 19.8%), Android-based mobile devices (52/415, 12.5%), and other devices (6/415, 1.4%).

The ADI national rank of participants spanned the entire range from 1 to 100, with a median (IQR) of 44 (28-61; n=5403; 95% CI 43‐45), indicating representation across diverse socioeconomic backgrounds. Similarly, the HOUSES index percentile ranged from 1 to 100, with a median (IQR) of 70 (43-88; n=5439; 95% CI 69‐71), demonstrating considerable variability in housing conditions among participants. The tight CI reflects reliable estimation despite the wide IQR, which indicates heterogeneity in housing-based SES. Neither age, ADI national rank, nor HOUSES index exhibited normal distribution according to the Anderson-Darling test (P<.001 for all variables), confirming the appropriateness of nonparametric statistical methods used in subsequent analyses. illustrates the distribution of the HOUSES index percentile and ADI national rank, with panel A showing ADI national ranks and panel B showing HOUSES index percentiles.

‎

To assess the nature of missing data, Little MCAR test was conducted on age, ADI national rank, and HOUSES index and yielded χ²₃=3.45; P=.24. This nonsignificant result indicates that missingness was not systematically related to other variables, supporting the assumption that data were MCAR. Accordingly, pairwise deletion was applied in subsequent analyses, a strategy chosen to preserve sample size and maintain statistical power while ensuring valid comparisons between study completers and noncompleters.

Comprehensive Participation Analysis

Significant differences were observed between participants who completed the study and those who did not across several demographic and socioeconomic factors. Age was higher among completers (median 66.4, IQR 56.0-72.5; 95% CI 65.1‐67.6 years) compared to noncompleters (median 62.8, IQR 47.5-72.7; 95% CI 62.2‐63.2 years; P<.001), suggesting that older individuals were more likely to participate (). The narrow CIs for both groups indicate high precision in estimating age differences, reinforcing the robustness of this finding. Participants who completed the study also resided in slightly less socioeconomically disadvantaged areas, as indicated by lower ADI national ranks (median 41.0, IQR 27.0-56.0; 95% CI 39.0‐43.0 vs median 44.5, IQR 28.0-62.0; 95% CI 43.0‐45.0; P=.04). Although the difference is statistically significant, the overlapping CIs suggest that the magnitude of this effect is modest and should be interpreted cautiously. No significant differences were found in HOUSES percentiles, indicating that individual HOUSES may exert less influence on participation compared to neighborhood-level disadvantage.

Sex differences in enrollment time were significant, with female participants taking longer to enroll than males (median 38.5, IQR 14.8-66.3; 95% CI 35.0‐41.0 vs median 32.0, IQR 8.0-57.5; 95% CI 29.0‐38.0 days; P=.01; ). The relatively narrow CIs for both groups indicate precise estimates of enrollment time differences, reinforcing the statistical significance of this finding. Additionally, urbanicity influenced the time to complete enrollment. Participants from urban areas enrolled more quickly than those from rural or urban cluster regions (median 32.0, IQR 9.0-58.0; 95% CI 31.0‐37.0 vs median 41.0, IQR 22.0-65.0; 95% CI 37.0‐49.0 days and median 40.0, IQR 13.0-71.0; 95% CI 33.0‐49.0, respectively; P=.01). The wider CIs for rural and urban cluster groups suggest greater variability in enrollment time compared to urban participants, possibly reflecting differences in access or engagement. However, no significant differences were found in completion time across sex or urbanicity. No significant associations were detected between HOUSES indices and study completion. Additionally, no significant relationship was found between the device type used for task completion and the completion or participation time. These null findings suggest that SES and device type did not influence completion dynamics in this cohort.

Step-by-Step Participation Analysis

The majority of invited participants either did not read or did not respond to the initial invitation via Epic (n=2736) or expressed no interest (n=1752). Among the 1358 participants who initially expressed interest, 415 (30.6%) ultimately completed the study in its entirety. Throughout various stages of the recruitment process, a total of 3346 participants failed to respond to follow-up communications from the research coordination team.

Analysis of participant age across different pathways revealed that individuals who did not respond to the invitation or eligibility check were significantly younger than those who proceeded toward study completion. This age disparity contributed to an increase in the median age of participants completing the study (66.4, IQR 56.0-72.5; 95% CI 65.1‐67.6) years compared to the overall invited cohort (62.8, IQR 47.5-72.7; 95% CI 62.2‐63.2) years. The narrow CIs for both estimates indicate high precision, reinforcing confidence in the observed age-related attrition pattern. A similar pattern was observed among the 95 participants who withdrew consent after initially providing it but before being accrued for the recording session, with these participants having a median age of 55.8 years compared to 66.3 years for those who continued with the recording. This suggests that younger participants were disproportionately represented among early dropouts.

While no significant differences in the HOUSES index were observed across different participation pathways, participants who did not respond to the initial invitation had significantly higher ADI national ranks (median 45.0, IQR 29.0-63.0; 95% CI 44.0‐46.0) compared to those who expressed interest (median 42.0, IQR 27.0-59.0; 95% CI 39.0‐43.0), indicating residence in more socioeconomically disadvantaged neighborhoods. The narrow CIs for these ADI estimates suggest precise measurement of this disparity, underscoring the influence of neighborhood-level socioeconomic disadvantage on initial engagement. No other significant differences in ADI national ranks were observed across subsequent recruitment steps. details the participant flow and checkpoint-specific distributions.

‎

Discussion

Summary of Principal Findings

This study provides a comprehensive analysis of participant recruitment pathways in a digital speech research study, revealing important associations between sociodemographic factors and participation outcomes. Understanding these pathways is critical for improving equity and efficiency in digital clinical research []. First, recruitment remains a fundamental challenge in clinical research, with participation rates often below 10 percent in specialized studies, limiting generalizability and slowing innovation [,]. Second, as digital recruitment methods become increasingly prevalent, it is essential to assess whether these approaches truly reduce barriers or inadvertently perpetuate existing disparities, an area where evidence remains limited []. By analyzing longitudinal recruitment data from 5846 invited patients, our findings demonstrate that systematic pathway analysis can uncover patterns of participation and attrition that may not be apparent when examining only final enrollment outcomes. These results support our primary hypothesis that sociodemographic factors, including age, urbanicity, and neighborhood disadvantage, are associated with recruitment and attrition in digital research workflows. This finding challenges the assumption that digital methods inherently improve inclusivity and highlights the need for targeted strategies, such as age-specific engagement and rural digital support, to promote retention in remote research.

Age-Related Participation Patterns

The significant age differences observed at various dropout points suggest that digital recruitment methods may be less effective for younger populations. Our analysis showed that the median age increased from 63 (IQR 48-72; 95% CI 62‐63) years in the invited cohort to 66.4 (IQR 56.0-72.5; 95% CI 65.1‐67.6) years among completers, indicating a robust and precise trend toward older participant retention. This finding challenges the conventional wisdom that digital methods inherently appeal to younger participants and suggests that age-specific engagement strategies may be necessary throughout the recruitment process. While younger individuals may be more comfortable with technology, our results and others suggest that younger participants are more likely to disengage or drop out of digital studies over time [,]. Possible explanations include competing priorities, lower perceived relevance of neurological research, and reduced tolerance for multistep enrollment processes []. This emphasizes the importance of age-sensitive retention strategies that extend beyond initial recruitment, including tailored messaging and incentives to support long-term engagement.

Disparities by Urbanicity in Enrollment Timing

The observation that participants from urban areas completed enrollment significantly faster than those from rural areas or urban clusters highlights potential disparities by urbanicity in research accessibility. Urban participants enrolled in a median of 32.0 (IQR 9.0-58.0; 95% CI 31.0‐37.0) days compared to 41.0 (IQR 22.0-65.0; 95% CI 37.0‐49.0) days for rural and 40.0 (IQR 13.0-71.0; 95% CI 33.0‐49.0) days for urban clusters, underscoring a consistent and precise difference. This finding aligns with broader concerns about the urban-rural digital divide and suggests that digital recruitment, while theoretically boundaryless, may still be influenced by infrastructure, digital literacy, or health care engagement [,]. Rural participants may face slower internet speeds, limited device availability, and less familiarity with patient portals, which could delay enrollment []. This contrasts with some findings in telehealth adoption, which suggest that technology can overcome location-based care barriers []. To mitigate these disparities, future strategies should include technical support for rural participants, alternative enrollment options (eg, phone-based consent), and targeted outreach through local health care networks.

Socioeconomic Disadvantage and Participation

Perhaps most notably, our analysis revealed that participants from neighborhoods with higher socioeconomic disadvantage (higher ADI national ranks) were significantly less likely to respond to initial invitations. This may reflect limited broadband access, lower digital literacy, and competing priorities in disadvantaged neighborhoods []. This finding is consistent with prior research demonstrating that individuals from more disadvantaged areas experience greater barriers to engaging with digital health, including higher no-show rates and lower uptake of telehealth services [,]. Our study adds nuance by comparing neighborhood-level (ADI) and housing-based (HOUSES) SES measures, an approach rarely examined in digital recruitment literature. This finding suggests that digital recruitment methods may perpetuate existing socioeconomic disparities in research participation if not specifically designed to address these barriers. To mitigate these barriers, recruitment strategies should include targeted outreach in areas with high ADI, provide technical support, and incorporate offline options. The absence of significant differences in the HOUSES index across participation pathways, despite differences in the ADI national rank, suggests that neighborhood context likely influences digital engagement more than individual housing characteristics because infrastructure and community resources shape access and literacy.

Sex Differences in Enrollment Dynamics

Females took longer to complete the enrollment process than males, a difference that warrants further investigation. This may reflect differences in time availability, competing responsibilities, or engagement with digital health platforms that could impact recruitment strategies. Prior studies suggest that social and structural factors, such as unequal distribution of caregiving and household responsibilities, affect time availability for research participation, particularly in remote studies []. Recruitment workflows could incorporate flexible scheduling and simplified enrollment steps to reduce time burden. While underexplored in the current literature, these insights underscore the need for sex- and gender-aware design in digital recruitment, which may improve inclusivity and reduce attrition.

Recruitment Funnel Attrition

Of 5846 individuals invited, a large proportion did not respond or declined participation, and only 415 of all the invited individuals (7.1%) completed the study. This significant attrition, consistent with patterns observed in other digital recruitment efforts [,], may reflect perceived complexity of enrollment, lack of immediate incentives, and competing priorities among participants. These findings highlight the need for iterative, multitouch recruitment strategies that re-engage potential participants and address barriers to active enrollment. Such strategies could include reminder messages, simplified consent processes, and personalized follow-ups to maintain engagement. Our findings reinforce prior evidence that digital recruitment alone is insufficient and highlight the importance of hybrid approaches combining digital and traditional outreach strategies [].

Limitations

Several limitations should be considered when interpreting these findings. First, this was an observational analysis conducted within a single academic health system, which may limit generalizability to other settings with different patient demographics or digital infrastructure. Second, the study relied on electronic health record–based digital recruitment and patient portal messaging, which presumes access to broadband internet and digital literacy; these factors were not directly measured and may have influenced participation patterns. Findings may generalize to similar academic health systems using portal-based recruitment but may differ in settings with lower portal adoption or different sociodemographic profiles. Third, the cohort was predominantly White (93.7%), limiting the ability to examine racial and ethnic disparities in digital recruitment. Future studies should prioritize inclusion of more diverse populations to assess whether similar sociodemographic patterns persist across racial and ethnic groups. Fourth, while we examined neighborhood-level (ADI) and housing-based (HOUSES) socioeconomic measures, other dimensions of socioeconomic status, such as income, education, and employment, were not available and could provide additional insights. Fifth, attrition analysis was based on recruitment checkpoints rather than qualitative data on participant motivations or barriers, which constrains the interpretation of underlying causes for dropout. Finally, the study period was relatively short (March to July 2024), and findings may not reflect seasonal or long-term trends in digital recruitment dynamics.

To address these challenges, we recommend several strategies to enhance inclusivity and effectiveness in digital recruitment. First, adopting a multichannel approach is essential. Digital methods should be complemented with traditional outreach, particularly for populations with limited online access, such as those in rural areas or socioeconomically disadvantaged communities. Second, age-specific engagement should be prioritized by developing tailored messaging and user experiences for different age groups, with special attention to reducing dropout rates among younger participants. Similarly, geographic barriers must be addressed by providing technical support and offering alternative participation options for both rural and urban clusters. To ensure continuous improvement, it is critical to track recruitment analytics. Monitoring data will help identify dropout points and demographic trends, enabling real-time adjustments to recruitment strategies. Additionally, efforts should be made to minimize participation burden by streamlining enrollment processes to reduce time demands, which is particularly beneficial for individuals with limited availability, such as those with caregiving responsibilities. Socioeconomic factors also require consideration. Recruitment materials should be inclusive, and incentives or support should be offered to offset participation costs, thereby improving socioeconomic accessibility. Furthermore, while device type did not affect completion time in our findings, platforms should still be optimized for device diversity, ensuring mobile-friendly and cross-device accessibility. Ultimately, intentional design and continuous evaluation are key to ensuring that digital methods promote inclusivity rather than hinder it. Future work should explore the mechanisms behind demographic and socioeconomic disparities, test interventions to address these patterns, and evaluate whether similar dynamics occur across other clinical domains.

Conclusion

This study demonstrates that digital recruitment methods in neurological research are subject to demographic, urbanicity, and socioeconomic influences that affect the representativeness of study populations. By mapping these factors, this study provides actionable insights for designing recruitment strategies that improve participation in remote neurological research and similar digital health initiatives. Our findings suggest that although digital recruitment expands reach, it does not eliminate traditional barriers and introduces new challenges. The digital divide appears to manifest in nuanced ways throughout the recruitment process, potentially influencing who participates in neurological research and, consequently, who benefits from its findings. The relatively low overall completion rate (7.1% of invited participants) underscores the persistent challenge of recruitment in specialized medical research, even with digital methods. Collectively, these findings reinforce the importance of refining digital recruitment strategies to bridge the persistent digital divide and promote research participation.

News Highlights

• AMD and Nutanix sign multi-year agreement to accelerate adoption of Nutanix-powered agentic AI platform on AMD accelerated compute infrastructure for enterprise AI and service providers
• AMD to invest and fund up to $250 million in Nutanix shares, and R&D and go-to-market for integrated solutions
• Joint roadmap to integrate AMD ROCm™ and AMD Enterprise AI software into the Nutanix Cloud Platform and the Nutanix Kubernetes Platform using AMD EPYC™ CPUs and AMD Instinct™ GPUs with support from a broad set of OEM server providers 

SANTA CLARA, Calif. and SAN JOSE, Calif., Feb. 25, 2026 (GLOBE NEWSWIRE) — AMD (NASDAQ: AMD) and Nutanix (NASDAQ: NTNX) today announced a multi-year strategic partnership to jointly develop an open, full-stack AI infrastructure platform designed to power agentic AI applications, everywhere. This agreement aligns to both companies’ commitment to an open ecosystem for AI, providing customers with choice and easy-to-deploy, production-ready, high-performance, and efficient solutions that are optimized for agentic AI, at the edge, inside enterprises, and across the cloud.

The partnership aligns silicon innovation, open runtime software and enterprise cloud orchestration technologies for AI to deliver scalable, production-ready agentic AI platforms across data center, hybrid and edge environments. By optimizing the Nutanix Cloud and Nutanix Kubernetes Platforms on AMD EPYC™ CPUs and AMD Instinct™ GPUs, and integrating the AMD ROCm™ software ecosystem and the AMD Enterprise AI platform into Nutanix AI full-stack solutions, the companies are developing an open solution for agentic AI platforms using high-performance infrastructure and supported by a broad set of OEM partners.

As part of the agreement, AMD will make a strategic investment of $150 million in Nutanix common stock at a purchase price of $36.26 per share, and fund up to $100 million for Nutanix to support joint engineering initiatives and go-to-market collaboration to accelerate the adoption of AMD and the Nutanix-powered agentic AI platform, everywhere. The equity investment is expected to close in the second quarter of 2026, subject to regulatory approvals and customary closing conditions.

“Enterprise customers need the freedom to run the models and workloads that matter most to their business, without compromise,” said Dan McNamara, senior vice president and general manager of Compute and Enterprise AI at AMD. “Through our partnership with Nutanix we’re building a scalable, full-stack AI platform rooted in openness, designed to give enterprises and service providers the flexibility to innovate, extend and grow AI deployments across Enterprises.”

“Our partnership with AMD reflects a shared vision for scalable, production-ready AI infrastructure,” said Tarkan Maner, President and Chief Commercial Officer, Nutanix. “Together, we are delivering full-stack, integrated platforms optimized for inference and agentic applications across hybrid environments for enterprises and service providers.”

Advancing the Open Ecosystem for Enterprise AI

Enterprise AI infrastructure is entering a phase where inference workloads dominate and openness is essential for long-term innovation. AMD is committed to advancing an AI ecosystem built on open standards, interoperable software frameworks and architectural choice, which are essential requirements for Enterprises.

The first jointly-developed agentic AI platform from this partnership is expected to come to market beginning in late 2026, underscoring the companies’ commitment to rapid execution and delivery.

As AI inference becomes foundational to enterprise computing, infrastructure must deliver performance, efficiency and operational simplicity at scale. The co-engineered platform will be designed to provide high-performance inference acceleration powered by AMD Instinct GPUs and EPYC™ CPUs, high-core-density compute and orchestration through AMD EPYC™ processors, and unified lifecycle management via Nutanix Enterprise AI — enabling enterprises to deploy open-source and commercial AI models without dependency on vertically integrated AI stacks.

Together, AMD and Nutanix are defining a new class of open AI infrastructure designed to support enterprise AI agents, multimodel inference services and industry-specific intelligent applications.

About AMD

AMD (NASDAQ: AMD) drives innovation in high-performance and AI computing to solve the world’s most important challenges. Today, AMD technology powers billions of experiences across cloud and AI infrastructure, embedded systems, AI PCs and gaming. With a broad portfolio of AI-optimized CPUs, GPUs, networking and software, AMD delivers full-stack AI solutions that provide the performance and scalability needed for a new era of intelligent computing. Learn more at www.amd.com.

AMD Forward-Looking Statements

This press release contains forward-looking statements concerning Advanced Micro Devices, Inc. (AMD) such as the strategic partnership between Nutanix and AMD; the joint development of an open, full‑stack AI infrastructure platform; the expected benefits, impact, performance, features, and functionality of the jointly developed platform; the parties’ commitment to an open ecosystem for AI; the expected timeline for the availability of the platform; expectations regarding market demand and adoption; expectations regarding the development of enterprise AI infrastructure within the industry; anticipated R&D and go‑to‑market funding commitments; and the closing of AMD’s strategic investment in Nutanix, which are made pursuant to the Safe Harbor provisions of the Private Securities Litigation Reform Act of 1995. Forward-looking statements are commonly identified by words such as “would,” “may,” “expects,” “believes,” “plans,” “intends,” “projects” and other terms with similar meaning. Investors are cautioned that the forward-looking statements in this press release are based on current beliefs, assumptions and expectations, speak only as of the date of this press release and involve risks and uncertainties that could cause actual results to differ materially from current expectations. Such statements are subject to certain known and unknown risks and uncertainties, many of which are difficult to predict and are generally beyond AMD’s control, that could cause actual results and other future events to differ materially from those expressed in, or implied or projected by, the forward-looking information and statements. Material factors that could cause actual results to differ materially from current expectations include, without limitation, the following: the ability of AMD and Nutanix to execute on their respective obligations under the strategic partnership; the ability to successfully integrate technologies and develop the jointly engineered platform on the anticipated timeline or at all; risks that anticipated R&D and go‑to‑market funding commitments, including the timing, amount, and conditions thereof, may not be fully realized; risks that the strategic partnership may not generate anticipated revenue synergies, if any; risks related to market acceptance, customer adoption, and participation by third parties in an open AI ecosystem; intense competition in the AI infrastructure market; the ability to obtain required regulatory clearances, including the expiration or termination of any applicable waiting period under the Hart‑Scott‑Rodino Antitrust Improvements Act of 1976; impact of government actions and regulations such as export regulations, import tariffs, trade protection measures, and licensing requirements; competitive markets in which AMD’s products are sold; the cyclical nature of the semiconductor industry; market conditions of the industries in which AMD products are sold; AMD’s ability to introduce products on a timely basis with expected features and performance levels; loss of a significant customer; economic and market uncertainty; quarterly and seasonal sales patterns; AMD’s ability to adequately protect its technology or other intellectual property; unfavorable currency exchange rate fluctuations; ability of third party manufacturers to manufacture AMD’s products on a timely basis in sufficient quantities and using competitive technologies; availability of essential equipment, materials, substrates or manufacturing processes; ability to achieve expected manufacturing yields for AMD’s products; AMD’s ability to generate revenue from its semi-custom SoC products; potential security vulnerabilities; potential security incidents including IT outages, data loss, data breaches and cyberattacks; uncertainties involving the ordering and shipment of AMD’s products; AMD’s reliance on third-party intellectual property to design and introduce new products; AMD’s reliance on third-party companies for design, manufacture and supply of motherboards, software, memory and other computer platform components; AMD’s reliance on Microsoft and other software vendors’ support to design and develop software to run on AMD’s products; AMD’s reliance on third-party distributors and add-in-board partners; impact of modification or interruption of AMD’s internal business processes and information systems; compatibility of AMD’s products with some or all industry-standard software and hardware; costs related to defective products; failure to maintain an efficient supply chain as customer demand changes; AMD’s ability to rely on third party supply-chain logistics functions; AMD’s ability to effectively control sales of its products on the gray market; impact of climate change on AMD’s business; AMD’s ability to realize its deferred tax assets; potential tax liabilities; current and future claims and litigation; impact of environmental laws, conflict minerals related provisions and other laws or regulations; evolving expectations from governments, investors, customers and other stakeholders regarding corporate responsibility matters; issues related to the responsible use of AI; restrictions imposed by agreements governing AMD’s notes, the guarantees of Xilinx’s notes and the revolving credit agreement; AMD’s ability to satisfy financial obligations under guarantees and other commercial commitments; impact of acquisitions, joint ventures and/or investments on AMD’s business and AMD’s ability to integrate acquired businesses; impact of any impairment of the combined company’s assets; political, legal and economic risks and natural disasters; future impairments of technology license purchases; AMD’s ability to attract and retain key employees; and AMD’s stock price volatility. Investors are urged to review in detail the risks and uncertainties in AMD’s Securities and Exchange Commission filings, including but not limited to AMD’s most recent reports on Forms 10-K and 10-Q. These forward‑looking statements speak only as of the date of this press release, and AMD undertakes no obligation to update or revise any forward‑looking statements, whether as a result of new information, future events, or otherwise, except as required by law.

About Nutanix

Nutanix (NASDAQ: NTNX) is a hybrid multicloud computing leader, offering organizations a unified software platform for running applications and AI and managing data anywhere. With Nutanix, organizations can simplify operations for traditional and modern applications, freeing them to focus on business goals. Trusted by more than 30,000 customers worldwide, Nutanix helps empower organizations to transform digitally and power hybrid multicloud environments consistently, simply, and cost-effectively. 

© 2026 Nutanix, Inc. All rights reserved. Nutanix, the Nutanix logo, and all Nutanix product and service names mentioned herein are registered trademarks or unregistered trademarks of Nutanix, Inc. (“Nutanix”) in the United States and other countries. Other brand names or marks mentioned herein are for identification purposes only and may be the trademarks of their respective holder(s). This press release is for informational purposes only and nothing herein constitutes a warranty or other binding commitment by Nutanix.

Nutanix Forward-Looking Statements

This press release contains express and implied forward‑looking statements, including, but not limited to, statements regarding the strategic partnership between Nutanix and AMD; the joint development of an open, full‑stack AI infrastructure platform; the expected benefits, impact, performance, features, and functionality of the jointly developed platform; the parties’ commitment to an open ecosystem for AI; the expected timeline for the availability of the platform; expectations regarding market demand and adoption; expectations regarding the development of enterprise AI infrastructure within the industry; anticipated R&D and go‑to‑market funding commitments; and the closing of AMD’s strategic investment in Nutanix. These forward‑looking statements are not historical facts and are based on Nutanix’s current expectations, estimates, assumptions, opinions, and beliefs. Actual results may differ materially from those expressed or implied by these forward‑looking statements as a result of various risks and uncertainties, including, but not limited to: the ability of Nutanix and AMD to execute on their respective obligations under the strategic partnership; the ability to successfully integrate technologies and develop the jointly engineered platform on the anticipated timeline or at all; risks that anticipated R&D and go‑to‑market funding commitments, including the timing, amount, and conditions thereof, may not be fully realized; risks that the strategic partnership may not generate anticipated revenue synergies, if any; risks related to market acceptance, customer adoption, and participation by third parties in an open AI ecosystem; intense competition in the AI infrastructure market; the ability to obtain required regulatory clearances, including the expiration or termination of any applicable waiting period under the Hart‑Scott‑Rodino Antitrust Improvements Act of 1976; and other risks and uncertainties described in Nutanix’s filings with the Securities and Exchange Commission, including its Annual Report on Form 10‑K for the fiscal year ended July 31, 2025 and subsequent Quarterly Reports on Form 10‑Q and other filings. These forward‑looking statements speak only as of the date of this press release, and Nutanix undertakes no obligation to update or revise any forward‑looking statements, whether as a result of new information, future events, or otherwise, except as required by law. Many of the products and features described herein, including the jointly developed AI infrastructure platform and its related features and functionalities, remain in various stages and will be offered on a when‑and‑if‑available basis. The development, release, and timing of any such products, features or functionalities are subject to change. Nutanix will not have any liability arising from reliance on this press release for any failure to deliver, or delay in the delivery of, any such products, features or functionalities. Any future product or product feature information is intended to outline general product directions, and is not a commitment, promise or legal obligation for Nutanix to deliver any functionality. This information should not be used when making a purchasing decision.

Nutanix Media Contact:
Jennifer Massaro
pr@nutanix.com

AMD Media Contact:
Aaron Grabein
aaron.grabein@amd.com

Source: Nutanix, Inc.

Released February 25, 2026