There were two reactions to Nvidia's fourth-quarter results, and they had almost nothing to do with each other.
The analyst reaction: upgrade models, raise price targets, treat the numbers as fresh confirmation that the AI infrastructure boom is intact and Nvidia remains its primary beneficiary.
Wedbush lifted its price target from $230 to $300 this morning. Several other firms did similar. The quarter was, by any conventional measure, exceptional.
The market reaction: sell. Nvidia's stock fell roughly 5% on Thursday, extending a decline that has taken it well below its 52-week high of $212.19. The company reported $68.1 billion in revenue, beating the Street's $66.2 billion estimate.
It guided to $78 billion for the current quarter, a figure that was $5 billion to $10 billion above where most analysts had expected it to land. None of that guidance includes any assumed contribution from China, a market Nvidia remains locked out of by US export restrictions.
Analysts are grading the quarter. Traders are grading the economy that is supposed to make the quarter repeatable.
The numbers that should have been enough
Fourth-quarter data centre revenue came in at $62.3 billion, up 22% from the prior quarter and 75% year-over-year. Compute revenues of $51.3 billion were driven by the Blackwell and Blackwell Ultra platforms, with GB-series chips representing roughly two-thirds of data centre compute sales. The B300 has now overtaken the B200 in volume, a generational transition Nvidia expects to continue driving growth through most of fiscal 2027.
Networking was the fastester-growing line, reaching $11 billion, up 34% quarter-over-quarter and 263% year-over-year, driven by NVLink growth alongside double-digit expansion in Spectrum-X and Infiniband. Gross margins expanded to 75.2%, up from 73.6% the previous quarter. Earnings per share of $1.62 beat consensus by around 6%.
The guidance was the real surprise. Wedbush had been modelling $68.5 billion for the current quarter. Street consensus heading into Wednesday was $72.8 billion. Management guided $78 billion, and explicitly stated that number excludes any revenue from a possible resumption of China sales. Nvidia is locked out of its second-largest market and still guiding to sequential revenue growth of nearly 15%.
Wedbush calls the China exclusion a potential unpriced catalyst. If export restrictions ease even partially, any resulting shipments would be additive to a baseline that is already running well ahead of expectations.
The $95 billion procurement signal
One of the most consequential disclosures Wednesday was not a revenue figure. Nvidia's supply chain purchase commitments rose to $95.2 billion, nearly double the $50.3 billion from the prior quarter, with management stating the company has visibility into supporting shipments into calendar 2027.
Wedbush has argued for several quarters that Nvidia's early move to lock up constrained components, particularly memory, created a supply-side moat that competitors cannot close quickly. The near-doubling of commitments extends that strategy ahead of the Vera Rubin platform launch, which management confirmed is on track for production shipments in the second half of this year. Every major cloud model builder, the company said, is set to deploy Vera Rubin as it ramps.
The networking growth is also a strategic data point. Analysts at several firms noted that the $11 billion quarterly run rate, representing roughly 3.5 times year-over-year growth, signals that large customers are buying deeper into Nvidia's full stack, from chips to switching infrastructure. New greenfield data centres are increasingly choosing Nvidia across the entire build rather than mixing vendors.
Why the stock fell anyway
The gap between analyst conviction and market response comes down to what each group is actually being asked to price.
Analysts are pricing Nvidia's competitive position, which remains strong. Traders are pricing the durability of the spending cycle behind it, which is less certain. The two questions have diverged enough that a record quarter with record guidance can produce a 5% sell-off without anyone being obviously wrong.
The concern is not that Nvidia is struggling. It is that Nvidia has become the quarterly proxy for the entire AI investment thesis, and that thesis is now being scrutinised at the level of hyperscaler cash flows rather than chip demand. Large cloud providers account for more than 50% of Nvidia's data centre revenue. If those customers face investor pressure to slow capital expenditure growth, or if AI monetisation timelines extend, the impact reaches Nvidia directly and quickly.
There is also a returns question. Nvidia generated roughly $35 billion in operating cash flow during the quarter but returned $4 billion to shareholders, a ratio that reflects aggressive reinvestment into the next product cycle. In a market that is increasingly demanding near-term proof points, that posture attracts scrutiny even from investors who accept the long-term bull case.
What Wedbush's $300 target actually assumes
Wedbush's revised price target is based on 30 times its fiscal 2028 earnings estimate of $9.97 per share, plus net cash of $2.21 per share. The firm notes this multiple is below the roughly 40 times Nvidia commanded during its last sustained data centre expansion cycle, but ahead of its five-year average. At current prices, the stock trades at roughly 25 times Wedbush's fiscal 2027 estimate of $7.68 per share.
The core bear case does not require doubting the chip demand. It requires only believing that the gap between AI infrastructure spending and AI-derived revenue will remain wide enough, for long enough, that hyperscaler investors eventually push back on the pace of capital deployment. That pushback does not need to be dramatic to matter. A modest deceleration in data centre orders from Nvidia's largest customers would be visible in the numbers very quickly.
For now, the basic problem facing anyone positioned short is the same as it has been for several years: the guidance keeps arriving higher than the model. Wedbush's prior estimate for the current quarter was $68.5 billion. Management guided $78 billion, without China. That $9.5 billion revision on a single quarter's forecast is larger than the total quarterly revenue of most significant technology companies.
The analysts know this. The traders know this too. They are just asking whether it continues, and on what timeline the AI economy that funds it starts paying for itself.