Nvidia just posted one of the most dominant quarters in semiconductor history and got punished for it.
The chipmaker reported fiscal fourth-quarter revenue of $68.13 billion on Wednesday, beating Wall Street’s $66.21 billion consensus by nearly $2 billion. Earnings per share came in at $1.62 versus expectations of $1.53. Revenue surged 73% year-over-year. The data center business alone generated $62.3 billion, up 75% from a year ago. And the company guided first-quarter fiscal 2027 revenue to a staggering $78 billion, blowing past the Street’s $71.5 billion estimate.
Shares fell as much as 5.6% on Thursday, marking Nvidia’s worst single-day drop since November.
Welcome to the expectations treadmill, where beating every number on the page still isn’t enough if the forward story doesn’t feel like acceleration on top of acceleration on top of acceleration.
The Numbers Were Pristine. The Market Shrugged.
Let’s be clear about what happened here. Nvidia didn’t stumble. It printed a record quarter on virtually every metric that matters. Full-year fiscal 2026 revenue hit $215.9 billion, up 65%. Data center revenue for the full year reached $193.7 billion. GAAP gross margin expanded to 75%, with non-GAAP at 75.2%. The company returned $41.1 billion to shareholders through buybacks and dividends during the fiscal year, with $58.5 billion still authorized for future repurchases.
CEO Jensen Huang used the earnings call to frame the next chapter around inference economics and agentic AI, calling the shift an “inflection point” and declaring that enterprise adoption of AI agents is “skyrocketing.” He unveiled details on the Vera Rubin platform, a six-chip system that Nvidia claims will deliver up to a 10x reduction in inference token cost compared to Blackwell. First samples have already shipped to customers, with AWS, Google Cloud, Microsoft Azure, and Oracle Cloud Infrastructure lined up as early deployers.
That $78 billion Q1 guidance? It assumes zero data center compute revenue from China, meaning Nvidia is projecting nearly 15% sequential growth while effectively writing off one of the world’s largest markets.
And still the stock dropped.
The Real Debate Has Moved Past Quarterly Beats
Richard Clode, a portfolio manager at Janus Henderson Investors, put it plainly: the conversation 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.”
That’s the whole game now. Nobody serious doubts that Nvidia will keep printing monster quarters in 2026. The question is what happens when the hyperscalers, collectively planning roughly $650 billion in AI capital expenditures this year, eventually need to show their own shareholders a return on all that spending. The AI infrastructure boom has entered what Bridgewater’s Greg Jensen called a “more dangerous phase,” one marked by exponentially rising investments in physical infrastructure and growing reliance on outside capital.
Nvidia’s stock isn’t being valued on what it earned last quarter. It’s being valued on whether the entire AI investment cycle has legs into 2028, 2029, and beyond. At a forward P/E of roughly 27, Nvidia looks cheap relative to its growth. But that valuation assumes the growth doesn’t slow materially for years. Any deceleration, even from 73% year-over-year to 40% year-over-year, reprices the entire thesis.
Cracks In The Periphery
While the data center business remained untouchable, some of Nvidia’s smaller segments showed stress. Gaming revenue of $3.7 billion missed analyst expectations of $4 billion and declined 13% sequentially. The company warned that memory constraints would remain a headwind to its gaming business “in the first quarter of fiscal 2027 and beyond,” essentially confirming speculation that Nvidia is prioritizing AI chip production over consumer GPUs when it comes to allocating scarce HBM supply.
Automotive revenue came in at $604 million, below the $654.8 million estimate. Not catastrophic numbers, but a reminder that Nvidia’s growth story is almost entirely a data center story right now, with over 91% of total revenue coming from that single segment.
Professional visualization was the bright spot outside data center, surging 159% year-over-year to $1.32 billion. But even that number is a rounding error relative to the company’s core AI business.
The China Wildcard
Nvidia’s decision to strip China data center compute revenue entirely from its Q1 outlook speaks volumes. Chinese customs authorities have reportedly instructed agents to block Nvidia’s H200 chips from entering the country, and Beijing has moved to pause certain orders. The company framed its guidance conservatively on this front, but the geopolitical variable remains live and unpredictable.
If those restrictions ease, Nvidia’s Q1 could blow past even the $78 billion midpoint. If they tighten further, it validates the bears who argue that Nvidia’s addressable market is shrinking at the margins even as absolute demand grows.
Competition Is No Longer Theoretical
AMD is preparing to ship Helios, its first rack-scale AI system, later this year. Meta just committed to deploying up to 6 gigawatts of AMD GPUs. Amazon and Google continue expanding their custom silicon programs. Intel, somehow, is up 27% this year on its own restructuring narrative.
None of these competitors threaten Nvidia’s dominance in the near term. Nvidia’s software ecosystem, CUDA moat, and annual platform cadence give it structural advantages that won’t evaporate overnight. But the market isn’t pricing Nvidia for the near term. It’s pricing for a decade of AI infrastructure buildout, and even small shifts in competitive dynamics can change the math on a stock valued at nearly $5 trillion.
What Actually Matters Now
Nvidia’s quarter wasn’t the problem. Nvidia’s quarter was extraordinary by any historical standard. The problem is that Nvidia has become a market object, a referendum on whether the entire AI trade holds together. A good quarter gets a shrug. A great quarter gets a shrug. The market is daring Nvidia to deliver a story about the next three years, not the last three months.
Gene Munster of Deepwater Asset Management framed the core question investors are now asking: “The real debate is what growth looks like in 2027 and 2028.” If you believe we’re in the second inning of the AI buildout, Nvidia at these levels is a steal. If you think 2026 marks peak AI capex, then the stock has already priced in most of the upside.
Jensen Huang is betting everything on the former. He committed to delivering a new full AI infrastructure platform every single year, positioning Nvidia not as a chipmaker but as the operating system for the AI industrial revolution. Vera Rubin is the next act. The Rubin platform promises a step-change in cost-per-token economics for inference workloads, exactly the kind of efficiency gain that could justify another wave of infrastructure spending even after the initial training buildout plateaus.
The numbers say buy. The stock price says the market has already heard the pitch. The 5% drop says the crowd wanted fireworks, got them, and still asked if the show was big enough.
For Nvidia, being good is no longer good enough. The bar is transcendence, every single quarter, forever. That’s the price of being the most important company in the most important trade on Wall Street.
