That pattern is the story. When the biggest chipmaker on earth clears every bar the Street sets and still cannot hold a bid, the bar is no longer about results. It is about what comes next.
The Numbers That Mattered
Adjusted earnings hit $1.87 per share versus the $1.77 consensus. Revenue of $81.6 billion topped the $79.2 billion forecast. Data center revenue reached $75.2 billion, nearly doubling from a year ago and making up 92% of total sales.
The Q2 guide of $91 billion, give or take 2%, cleared the $87 billion Street estimate and even topped the $89 to $90 billion buy-side whisper. That is a $4 billion beat on forward guidance alone.
The board authorized $80 billion in new share buybacks with no expiration date. They also raised the quarterly dividend from one cent to 25 cents. Revenue grew 85% year over year.
Jensen Huang Says the Demand Has Gone Parabolic
CEO Jensen Huang closed the earnings call with a line that will get quoted for the rest of the year: "Agentic AI has arrived." He described demand as "parabolic," driven by companies deploying AI agents that do real work across industries.
The next hardware platform, called Vera Rubin, ships in the third quarter of fiscal 2027 and is purpose-built for these agentic workloads. Nvidia is no longer just selling training chips. It is selling the infrastructure for AI systems that operate on their own.
That shift matters for how you value the company. Training demand has a ceiling. Inference and agentic deployment, where AI does ongoing work rather than one-time learning, could be a much larger and more recurring revenue stream.
China Remains a Question Mark
CFO Colette Kress confirmed that Nvidia has generated zero revenue from chip sales to China so far. The H200 was technically cleared for export after Trump's Beijing summit, but Beijing blocked purchases on the other end.
China represents a $50 billion annual AI chip market. The fact that Nvidia just posted $91 billion in guidance without any contribution from that market tells you two things. The rest of the world is spending fast enough to compensate. And any resolution in China becomes pure upside.
Why the Stock Fell on a Perfect Quarter
BofA analyst Vivek Arya noted that Nvidia has declined after three of its last four earnings reports. The average one-day post-earnings move over that stretch is negative 1.5%. When expectations are already priced for excellence, even excellence is not enough.
This is the environment Broadcom faces when it reports next month. So does every AI infrastructure stock trading at a premium. The quality of the business is not the issue. The multiple you pay at these yield levels is.
Two weeks ago, Stanley Druckenmiller's 13F showed he sold all of Alphabet, slashed Amazon by 99%, and opened a fresh position in Broadcom. His thesis: rotate from the consumers of AI to the infrastructure providers before the crowd does. Wednesday night did not prove him right, but it did not contradict him either.
What to Watch From Here
The $91 billion Q2 guide resets the AI trade at full strength. Any talk of a chip demand slowdown is dead for at least one more quarter. The bear case, including Michael Burry's semiconductor short thesis, takes a significant hit.
But the stock's reaction tells you the easy money in mega-cap AI may already be made. The next leg of returns could come from infrastructure layer names like Broadcom, Micron, and Intel, the three stocks Druckenmiller bought in Q1, rather than from Nvidia itself.
OpenAI is preparing to file its own IPO paperwork as soon as Friday at an $850 billion-plus valuation. That filing, combined with the SpaceX S-1 this week and Cerebras already trading, creates the most concentrated AI IPO pipeline in market history. Capital is flowing into AI from every direction. The question is not whether AI spending continues. It is which entry point still has upside.
