At 5:00 PM ET today, NVIDIA will release the single most-watched number in technology: Q1 FY2027 revenue. Wall Street consensus sits at roughly $78 billion — 77% growth year over year. Data center alone is expected to clock in near $73 billion. Non-GAAP gross margin: 74%-plus. EPS: around $1.77.
If those numbers come in at consensus, NVIDIA will have done in a single quarter what Intel did in an entire year at its peak. If they miss, every AI valuation on the planet gets repriced overnight. If they beat, the AI capex cycle gets another twelve months of fuel.
This is not a normal earnings call. It is the closest thing the AI economy has to a Federal Reserve meeting.
The Numbers in Context
Here is what the print needs to clear to keep the AI bull case intact:
| Metric | Q1 FY26 (a year ago) | Q4 FY26 (last quarter) | Q1 FY27 Consensus | YoY Growth Implied |
|---|---|---|---|---|
| Total revenue | $44.1B | $65.8B | ~$78B | +77% |
| Data center revenue | $35.5B | $57.0B | ~$73B | +106% |
| Non-GAAP gross margin | 71% | 73% | 74%+ | +300 bps |
| Non-GAAP EPS | $0.96 | $1.41 | ~$1.77 | +84% |
For perspective: NVIDIA's expected data center quarter is larger than the entire annual revenue of Coca-Cola. It is bigger than the GDP of Iceland. And it is being generated by selling rectangles of silicon to roughly 50 customers — about ten of whom account for the majority of the business.
That last point is where the risk lives. When the buyers are this concentrated and this aggressive, a single guidance cut from Microsoft or Meta can move NVIDIA's stock 8% in an afternoon. Q1 FY27 is the quarter where investors find out whether the hyperscaler capex commitments from February's earnings season have actually shown up as orders.
Why $73 Billion in Data Center Is the Only Number That Matters
Forget gaming. Forget automotive. Forget professional visualization. NVIDIA is now a data center company that has a few side businesses, and the data center segment is what investors will move the stock on tonight.
The $73 billion consensus implies sequential growth of roughly 28% from Q4 FY26. That is not a normal quarter-over-quarter increase for a mature business — it is the kind of growth you usually see in a pre-revenue startup. NVIDIA is doing it on a $250 billion annualized base.
The data center number breaks down into roughly three buckets:
- Hyperscalers (Microsoft Azure, AWS, Google Cloud, Meta, Oracle) — roughly 50% of data center revenue. These are the customers paying for OpenAI's training runs, Anthropic's Claude clusters, and every B2B AI service running at scale.
- Sovereign AI (UK, Saudi Arabia, UAE, Japan, Korea, France) — the fastest-growing segment. Countries are building national AI stacks because they have decided AI infrastructure is now critical national infrastructure, like power grids and telecom.
- Enterprise and Tier-2 (banks, pharma, defense contractors, neoclouds like CoreWeave and Lambda) — the long tail that is becoming less of a long tail every quarter.
If sovereign AI keeps growing at triple-digit rates, NVIDIA has another 18-24 months of insulation from any single hyperscaler pulling back. If sovereign growth stalls tonight, the entire bull case rests on five US companies continuing to outspend each other.
The Margin Question Nobody Wants to Ask
A 74%-plus gross margin on hardware is a phenomenon that essentially does not exist anywhere else in industrial history. Apple, in its most aggressive iPhone years, ran hardware margins around 38%. Cisco at peak was 65%. Intel at peak was 60%. NVIDIA is running 14 points higher than Intel ever did, on a product category that historically commoditizes inside three years.
The margin holds because three things are true at once:
- Blackwell and the next-generation Rubin platform have no real competitor. AMD's MI400 series is shipping but not stealing material share at the high end. Custom silicon from Google (TPU v7) and Amazon (Trainium 3) chips away at the edges but does not displace NVIDIA in any data center where the customer wants the broadest software compatibility.
- CUDA is the actual moat. Hardware competitors can catch up on transistors. They cannot catch up on twenty years of developer mindshare, libraries, optimizations, and the fact that every PhD in machine learning since 2012 learned on NVIDIA hardware.
- Demand still outstrips supply. When you cannot make enough product to meet orders, you do not discount.
If tonight's margin number prints below 74%, that is the market telling you the third pillar is starting to crack. Watch the guidance for next quarter even more carefully than the current quarter — guidance is where any softness shows up first.
The China Wildcard
Last quarter, NVIDIA disclosed it could not ship roughly $2.5 billion of H20 revenue to China because of US export controls. That was a real number that left a hole in the data center line.
The picture has shifted. The H200 — one tier above the H20 — is now cleared for sale to China under the latest export framework. NVIDIA has restarted H200 manufacturing for Chinese customers and CEO Jensen Huang confirmed at GTC 2026 that purchase orders are already on the books. Blackwell-class chips (B100, B200, and the GB200 systems) remain restricted.
What this means for tonight's print:
- If China revenue contributed meaningfully this quarter, it will be the upside surprise that pushes total revenue past $80 billion.
- If China revenue was minimal because the manufacturing restart only kicked in late in the quarter, the upside lands in Q2 guidance instead.
Either way, the China line item is the variable most likely to move the stock outside the ±$3 billion range of consensus. Pay attention to the geographic revenue disclosure in the earnings supplement, not just the headline number.
What It Means for OpenAI, Anthropic, and xAI
Every dollar NVIDIA reports tonight is a dollar that the frontier model companies have committed to spending on compute. The relationship is recursive — and getting more so.
Earlier this month I covered NVIDIA's $40 billion in equity bets on AI companies, including the $30 billion OpenAI investment that is the largest corporate venture round in history. That money flows back as GPU orders. Anthropic is doing the same dance with AWS Trainium plus NVIDIA hybrid clusters. xAI's Colossus 2 cluster in Memphis is reportedly the largest single deployment of Blackwell systems in the world.
The compute lock-in is not theoretical. OpenAI's $13B+ ARR depends on GPU supply that only NVIDIA can deliver at scale. Anthropic at $5B+ ARR is in the same boat. If NVIDIA's guidance tonight signals a supply ceiling for the next two quarters, model providers will be the first to feel it — and the trillion-dollar AI race becomes a contest of who pre-ordered earlier.
What It Means for AI Startup Valuations
Series A and Series B AI valuations have been running at 40-80x ARR through 2025 and the first half of 2026. That premium exists almost entirely because investors believe AI infrastructure spending is a multi-trillion-dollar wave that lifts every boat in the category.
NVIDIA's print tonight is the closest thing to a leading indicator that wave gets. If the data center number beats and Q2 guidance is above $80 billion, expect multiples to hold and possibly expand for the rest of the year. If guidance comes in below $78 billion, every AI startup raising in the next 90 days will see the multiple compression that public market investors are already pricing in.
I wrote a full breakdown of how VCs are valuing AI companies in this cycle — the short version is that gross margin structure and net revenue retention now matter more than top-line growth. Tonight's NVIDIA print does not change which metrics matter, but it changes the multiple investors are willing to pay on top of them.
What It Means for Your Enterprise AI Bill
If you run a company that uses Claude, ChatGPT Enterprise, or any major AI platform, NVIDIA's earnings are not just a stock story. They are the upstream cost driver for everything you buy.
When NVIDIA holds 74% gross margins on the chip layer, every model provider has to either eat that cost or pass it through to you. Most have started passing it through. I covered this dynamic in detail in AI cost pass-through across enterprise software — usage-based pricing on AI features is climbing across the board, and the underlying reason is that the inference cost of running a frontier model has not dropped at the rate everyone predicted.
If NVIDIA delivers a clean beat tonight and guides higher, expect Microsoft, OpenAI, and Anthropic to push another round of price increases inside 90 days. The AI cloud capacity crunch is not getting better — it is getting more expensive while the wait list grows.
What to Watch When the Call Drops
When the earnings supplement hits at 5:00 PM ET and the call kicks off at 5:30 PM, three numbers matter more than the headline:
- Q2 FY27 revenue guidance. Consensus is roughly $84 billion. Anything below $82 billion is a stock-mover to the downside. Anything above $87 billion triggers a melt-up.
- Sovereign AI as a percentage of data center revenue. Last quarter Jensen disclosed it crossed 15%. If that number is now 20%+, the diversification story is real and the hyperscaler concentration risk shrinks.
- Blackwell shipment commentary. The market needs to hear that Blackwell rack-scale systems (GB200 NVL72 and the new NVL144) are shipping at volume. Any commentary about yield issues or thermal challenges hits the stock immediately.
Skip the macro commentary in the prepared remarks. Skip the AI total addressable market slide. Watch the segment breakdown in the supplement and the geographic disclosure in the 10-Q tomorrow. That is where the actual story is.
The Trillion-Dollar Lens
NVIDIA's market cap is roughly $4.5 trillion as of yesterday's close. That makes it larger than the entire German stock market. Larger than the GDP of India. Larger than every public oil company on Earth combined.
The bear case is that a single quarter of weak guidance compresses that valuation by 20-30% — wiping out a trillion dollars of market cap inside a week. The bull case is that AI infrastructure spending is still in inning three of a nine-inning game and the data center number tonight is on its way to $150 billion per quarter by FY29.
Both cases are defensible from where we sit on May 20, 2026. The print tonight is the data point that tells us which one is closer to true. The companies building on top of NVIDIA — OpenAI, Anthropic, xAI, every sovereign AI program, every enterprise customer running production AI workloads — are all making their 2027 plans based on what gets said in the next four hours.
Whatever happens at 5 PM, the verdict gets delivered before tomorrow's open. And the entire AI economy gets repriced before lunch on Thursday.
Tools for action
Turn this insight into execution
Use the calculator, stack selector, and playbooks to estimate value and launch faster.
