NVIDIA posts record $68.1 billion quarter, but Wall Street wants more

NVIDIA reported record quarterly revenue of $68.1 billion for its fiscal fourth quarter ended January 25, 2026, beating Wall Street estimates by roughly $2 billion.^[1] The results were up 73% year-over-year and 20% quarter-over-quarter, driven overwhelmingly by its data center business.^[1] The company guided first-quarter fiscal 2027 revenue of $78 billion, plus or minus 2% — approximately $5 billion above consensus estimates.^[1][2]

Investors initially rewarded the beat, pushing shares up roughly 3.5% in after-hours trading within minutes of the results posting.^[5] By Thursday's close, the reaction had reversed sharply: NVIDIA fell 5.5% to $184.89, its worst single-day decline since April.^[4]

The numbers

NVIDIA's data center segment — now accounting for over 91% of total revenue — posted $62.3 billion for the quarter, up 75% year-over-year and 22% sequentially.^[1][2] Full-year data center revenue reached $193.7 billion, up 68% from fiscal 2025's $115.2 billion.^[1]

GAAP gross margin came in at 75.0%, beating guidance of 74.8% and recovering from 73.4% in the prior quarter.^[1][3] Non-GAAP gross margin was 75.2%.^[1] GAAP net income for the quarter was $43.0 billion, up 94% year-over-year, while non-GAAP net income was $39.6 billion.^[1]

Adjusted earnings per share were $1.62, above the LSEG consensus estimate of $1.53.^[2] Full-year revenue hit $215.9 billion, up 65%, with net income of $120.1 billion.^[1] Free cash flow for the quarter was $34.9 billion and $96.6 billion for the full fiscal year.^[1]

The company returned $41.1 billion to shareholders during fiscal 2026 through share repurchases and dividends, with $58.5 billion remaining under its buyback authorization.^[1]

Outside the core data center business, professional visualization revenue surged 159% year-over-year to $1.32 billion, driven by demand for Blackwell-based workstation GPUs.^[1][2] Gaming revenue was $3.7 billion, up 47% year-over-year but down 13% sequentially as holiday-driven channel inventory normalized.^[1] Automotive came in at $604 million, up 6% year-over-year but slightly below analyst expectations of $654.8 million.^[2]

What NVIDIA said

CEO Jensen Huang framed the results around what he called three structural platform shifts: from CPUs to GPU-driven computing, from traditional machine learning to generative AI, and from generative AI to agentic AI.^[3]

"Computing demand is growing exponentially — the agentic AI inflection point has arrived," Huang said in the earnings release. "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."^[1]

NVIDIA announced it shipped its first Vera Rubin GPU samples to customers this week.^[1] The next-generation platform promises up to a 10x reduction in inference token cost compared to Blackwell, and cloud providers AWS, Google Cloud, Microsoft Azure, and Oracle Cloud Infrastructure will be among the first to deploy Vera Rubin-based instances.^[1]

The company also disclosed a multiyear, multigenerational strategic partnership with Meta spanning on-premises, cloud, and AI infrastructure, including the deployment of millions of NVIDIA Blackwell and Rubin GPUs.^[1]

On the earnings call, Huang pushed back directly on concerns about hyperscaler spending sustainability. "I am confident in their cash flow growing," he said. "We have now seen the inflection of agentic AI and the usefulness of agents across the world and enterprises everywhere, and you're seeing incredible compute demand because of it."^[3]

The China question

NVIDIA's guidance notably excludes any data center compute revenue from China. CFO Colette Kress addressed this directly on the call: "While small amounts of H200 products for China-based customers were approved by the U.S. government, we have yet to generate any revenue, and we do not know whether any imports will be allowed into China."^[6]

The company's cash flow statement showed a $4.5 billion H200 charge in Q1 of fiscal 2026, with a partial $180 million release in Q2, related to Chinese export restrictions.^[1] No material charges were recorded in Q3 or Q4.^[1]

Why the stock fell

The paradox of NVIDIA's report — a comprehensive beat followed by a 5.5% selloff — reflects a market that has moved beyond evaluating quarterly performance and is now interrogating the durability of the AI infrastructure cycle itself.^[4]

Bloomberg reported that the decline was NVIDIA's worst in ten months, noting that the company's forecast "failed to dispel fears of an AI bubble" despite exceeding estimates.^[4] Morgan Stanley analyst Joseph Moore called it the "largest, cleanest beat and raise in the history of the semis industry" and said he was surprised by the muted reaction.^[5]

Several analysts pointed to the broader context. Hyperscaler capital expenditures are now projected to approach $700 billion combined in 2026, according to analyst estimates compiled by CNBC.^[2] A recent Moody's report flagged $662 billion in future data center lease commitments from major cloud providers that remain off their balance sheets.^[3] Total supply-related commitments on NVIDIA's own balance sheet nearly doubled in one quarter, from $50.3 billion at the end of Q3 to $95.2 billion at the end of Q4.^[3]

"The debate has shifted away from near-term results and toward the sustainability of AI capex spending, amid concerns around its quantum, monetisation, and potential cashflow degradation," Richard Clode, portfolio manager at Janus Henderson Investors, told CNBC.^[2]

The selloff extended across the semiconductor sector. Broadcom, Lam Research, Western Digital, and Applied Materials all declined on Thursday.^[2]

The wider landscape

NVIDIA's earnings landed during a week already dominated by the intersection of AI and infrastructure politics.

On Tuesday, President Trump used his State of the Union address to announce a "Rate Payer Protection Pledge," under which major technology companies would commit to building, buying, or generating their own electricity for new AI data centers rather than drawing on the existing grid.^[7] The White House confirmed that Amazon, Google, Meta, Microsoft, xAI, Oracle, and OpenAI will gather at the White House on March 4 to formally sign the pledge.^[7]

"We have an old grid. It could never handle the kind of numbers, the amount of electricity that's needed," Trump said during the address. "So I'm telling them, they can build their own plant."^[7]

Anthropic separately pledged to cover 100% of electricity price increases that consumers face from its data centers.^[8]

Energy experts were skeptical about the pledge's enforceability. Ari Peskoe, director of the Electricity Law Initiative at Harvard Law School, told Reuters that the pledge was "meaningless until we see utilities file contracts with state and federal regulators that allocate all costs of serving data centers to data centers."^[8] He noted that most current cost pressure comes from transmission, distribution, and system readiness — costs that persist even when data centers self-supply generation.^[8]

A structural shift in how NVIDIA reports

Starting in Q1 of fiscal 2027, NVIDIA will include stock-based compensation expense in its non-GAAP financial measures, a change from its previous practice of excluding it.^[1] The company said stock-based compensation is "a foundational component of NVIDIA's compensation program to attract and retain world-class talent."^[1] For Q1, the impact is estimated at 0.1% on non-GAAP gross margin and $1.9 billion in non-GAAP operating expenses.^[1]

This is a meaningful transparency shift. By including SBC in non-GAAP figures, NVIDIA's future non-GAAP margins will appear lower on a like-for-like basis compared to historical periods, even if underlying economics remain unchanged.

Why this matters

NVIDIA's quarterly earnings have become more than a corporate report. They function as the single most closely watched indicator of whether the AI infrastructure buildout is sustainable or overextended.

The numbers themselves leave little room for debate: demand remains intense, margins are recovering, the product roadmap is accelerating, and guidance crushed expectations.^[1][2] CFO Kress reiterated that NVIDIA sees visibility to $500 billion in revenue from Blackwell and Rubin offerings through the end of calendar year 2026, with total AI infrastructure investment potentially reaching $3 to $4 trillion annually by 2029 or 2030.^[3]

But the market's reaction suggests a growing gap between NVIDIA's operational execution and investors' confidence in what happens next. The question is no longer whether NVIDIA can sell its chips. It is whether its customers — spending at unprecedented levels — will generate enough return on those investments to sustain the cycle.

As Kiplinger's James Demmert put it, investors were "extremely hungry for confirmation that AI is still a positive storyline for stocks."^[5] NVIDIA gave them that confirmation. It was not enough.