NVIDIA Q1 FY2027 Earnings:
$81.6B Record, Stock Falls Anyway

NVIDIA reported Q1 FY2027 earnings on May 20 that, by almost any historical standard, were extraordinary: total revenue of $81.6 billion (up roughly 85% year over year), Data Center revenue of $75.2 billion, a non-GAAP gross margin of 75%, free cash flow of $48.6 billion, and guidance for the next quarter of $91 billion — comfortably above the $86.8 billion Wall Street consensus. The company raised its dividend twenty-five-fold and authorized an additional $80 billion in buybacks.

Then the stock fell. Shares drifted lower in the days after the report, marking the fourth consecutive post-earnings decline for a company that keeps beating its own guidance. For investors trying to reconcile a record quarter with a falling share price, the interesting question is not “was the quarter good?” — it plainly was — but rather what the market is now pricing, what changed in the forward setup, and whether the AI infrastructure thesis that built a multi-trillion-dollar company is intact, fully valued, or something in between. If you want a framework for reading past the headline numbers before diving in, our guide to the earnings-report red flags that predict stock drops is a useful companion to this analysis.

Q1 FY2027 Results: The Numbers Behind a Record Quarter

NVIDIA's Q1 FY2027 produced the kind of figures that would have triggered an unambiguous rally in almost any other era. Total revenue reached $81.6 billion, up 85.2% year over year and 19.8% sequentially. Data Center — the engine that now defines the company — generated $75.2 billion, up 92.4% year over year and accounting for 92% of total revenue. Non-GAAP earnings per share of $1.87 beat the consensus estimate of roughly $1.77, and the company converted the quarter into $48.6 billion of free cash flow.

One structural change is worth flagging up front: beginning this quarter, NVIDIA collapsed its legacy reporting segments (Gaming, Professional Visualization, Automotive, and OEM/Other) into a single Edge Computingplatform alongside Data Center. As a result, the company no longer discloses standalone gaming or automotive revenue. Edge Computing contributed $6.4 billion, or about 8% of total revenue — but the granular sub-segment breakdowns that appeared in prior quarters are no longer reported.

A note on GAAP net income, which came in at $58.3 billion (up 211% year over year): that figure includes approximately $15.9 billion of unrealized gainson NVIDIA's equity investment portfolio — paper gains on stakes in AI companies — not operating profit. For judging the operating health of the business, operating income of $53.5 billion (a 65.6% operating margin) and non-GAAP results are the more meaningful reference points. We flag this because comparing the headline GAAP net income to revenue would materially overstate the company's operating margin.

Why NVIDIA Stock Fell on a Record Beat

Here is the paradox at the center of this report: NVIDIA beat on revenue, beat on EPS, beat on guidance, raised its dividend, and expanded its buyback — and the stock still declined. This was the fourth consecutive quarter in which NVDA fell after reporting a beat. Understanding why requires separating the quality of the business from the expectations already embedded in the price.

The mechanics are straightforward. NVIDIA trades at a forward earnings multiple in the high-20s — depending on which forward EPS estimate is used, roughly 26x to 29xbased on available analyst figures (a directional range, not a precise figure, given the spread in published estimates). At that valuation, a beat is not enough; the beat has to be large enough to exceed the “whisper” expectations that sit above published consensus. When a company has compounded at 85%+ revenue growth, the bar for a positive surprise rises every quarter. A $91 billion guide that beats consensus by roughly $4 billion is impressive in absolute terms, but for a stock priced for sustained acceleration, “merely excellent” can read as deceleration on the margin.

There is also the simpler behavioral dynamic that traders summarize as “buy the rumor, sell the news.” NVDA had climbed more than 11% in May heading into the print. When a stock rallies into an event that everyone expects to be good, the event itself can become a liquidity moment for profit-taking even when the results validate the thesis. None of this means the quarter was weak. It means the quarter was priced in.

Data Center: A $75 Billion Quarter and the Concentration Question

Data Center is no longer a segment of NVIDIA's business — it essentially is the business. At $75.2 billion, it represented 92% of total revenue and grew 92% year over year. The demand is driven by hyperscale cloud providers building AI training and inference capacity, enterprises standing up AI infrastructure, and — increasingly — governments. CEO Jensen Huang framed the demand environment in characteristically expansive terms on the earnings call:

"

Demand has gone parabolic. The reason is simple. Agentic AI has arrived.

— Jensen Huang, CEO, NVIDIA (Q1 FY2027 Earnings Call (May 20, 2026))

The 92% concentration is simultaneously NVIDIA's greatest strength and its most-debated risk. On the bull side, it reflects total dominance of the most important computing market of the decade. On the bear side, it means the company's fortunes are tied to a single demand driver — AI infrastructure capital expenditure — that is itself concentrated among a handful of hyperscale buyers. CFO Colette Kress framed the capex trajectory directly:

"

With analysts now forecasting hyperscale CapEx to exceed $1 trillion by 2027, AI infrastructure spending is on track to reach $3 trillion to $4 trillion annually by the end of this decade.

— Colette Kress, CFO, NVIDIA (Q1 FY2027 Earnings Call (May 20, 2026))

(This is a management projection based on third-party industry forecasting, not a guaranteed outcome.) That is the entire bull case in one statement: if hyperscale AI capex is on a path from $1 trillion toward $3-4 trillion, NVIDIA's Data Center revenue has years of runway. The bear rebuttal is that capex cycles are cycles — they normalize, digest, and occasionally reverse — and a business deriving 92% of revenue from one capex theme carries more cyclical risk than its margins currently imply. Both observations can be true at different time horizons.

Blackwell Is Ramping — and Vera Rubin Is Already in View

The product story underneath the Data Center number is the Blackwell architecture ramp. Blackwell (and the GB-series rack-scale systems built around it) is now the volume platform, and management describes the Grace Blackwell NVLink configuration as the leading inference engine in production. The significance for investors is twofold: Blackwell sustains the revenue growth, and its product mix is a key input into the gross-margin trajectory.

What is striking is that NVIDIA is already directing attention to the next architecture. On the call, CFO Colette Kress previewed Vera Rubin in terms designed to extend the roadmap narrative:

"

Vera Rubin will deliver up to 35x higher inference throughput and up to 10x greater AI factory revenue compared with Blackwell.

— Colette Kress, CFO, NVIDIA (Q1 FY2027 Earnings Call (May 20, 2026))

Forward-looking performance claims like these are management projections, not guaranteed outcomes, and should be weighed as such. But the strategic point stands: NVIDIA's roadmap cadence — Hopper to Blackwell to Blackwell Ultra to Vera Rubin — is designed to keep the upgrade cycle continuous, which is precisely what sustains pricing power and recurring demand. The risk embedded in this cadence is execution: each transition requires yield ramps, supply-chain coordination (advanced packaging, high-bandwidth memory), and customer migration. A stumble in any transition would pressure both revenue and margins.

Gross Margin: The 75% Plateau That Matters More Than the Beat

NVIDIA's non-GAAP gross margin held at 75.0%in Q1 FY2027, with GAAP at 74.9%. For context, the year-ago quarter (Q1 FY2026) reported a GAAP gross margin of just 60.5% — but that figure was depressed by a $4.5 billion inventory charge tied to H20 export restrictions. Excluding that one-time charge, the year-ago non-GAAP gross margin would have been around 71%. So the year-over-year margin expansion is partly genuine operating leverage and partly recovery from a one-time hit; intellectually honest analysis separates the two.

Looking forward, management guided Q2 FY2027 gross margin to roughly 75% (±50 basis points) — essentially flat. That flat trajectory is the number to watch. The bull case assumes NVIDIA holds the mid-70s as Blackwell scales and eventually pushes higher; the bear case warns that next-generation ramps (Blackwell Ultra, then Vera Rubin) typically carry early-stage yield and cost headwinds that can pressure margins during transitions. A company guiding flat margins at 75% is not a problem — but it is a signal that the easy margin expansion of the recovery phase is behind, and incremental gains must now be earned through mix and efficiency.

Sovereign AI: The Demand Pillar Diversifying Away From Hyperscalers

One of the more important and least-discussed stories in the quarter is sovereign AI — governments building national AI infrastructure. CFO Colette Kress quantified the momentum on the call:

"

Sovereign revenue increased more than 80% year over year. NVIDIA AI infrastructure is now deployed across nearly 40 countries, representing $50 trillion in GDP.

— Colette Kress, CFO, NVIDIA (Q1 FY2027 Earnings Call (May 20, 2026))

This matters for a specific reason: it diversifies NVIDIA's demand base away from the handful of US hyperscalers that dominate Data Center spending. Sovereign demand is driven by national strategy rather than corporate capex budgets, which makes it somewhat less correlated with the hyperscaler capex cycle — and largely insulated from the US-China export friction that constrains the China opportunity. For investors who track hyperscaler concentration risk, sovereign AI is a demand-diversification development worth monitoring over the next several quarters. (Note: the 80% figure is management commentary from the earnings call, not a separately audited line item in the financial statements.)

NVIDIA's networking franchise reinforces the same diversification point. Kress noted that NVIDIA's AI-purpose-built Ethernet platform has scaled rapidly:

"

Spectrum-X, our end-to-end Ethernet platform purpose-built for AI, is now larger than all Ethernet network peers combined.

— Colette Kress, CFO, NVIDIA (Q1 FY2027 Earnings Call (May 20, 2026))

Networking — Spectrum-X Ethernet plus NVLink/NVSwitch — is the underappreciated moat. As AI clusters scale toward 100,000+ GPU configurations, the interconnect layer becomes a strategic chokepoint, and it is a key reason competitors matching NVIDIA on raw chip performance do not necessarily match it on full-system economics.

China, H20, and the Revenue That Isn't There (Yet)

NVIDIA recognized essentially zeroData Center revenue from China in Q1 FY2027. Following 2025 US export controls, NVIDIA's China shipments were frozen — the year-ago quarter absorbed a $4.5 billion charge tied to its China-compliant H20 inventory and purchase obligations. As of this report, the US government has since approved export licenses for shipments to certain China-based customers, but — critically — no revenue has yet been recognized. CFO Colette Kress was precise about this on the call:

"

While the U.S. government has approved licenses for H200 to be shipped to China-based customers, we have yet to generate any revenue.

— Colette Kress, CFO, NVIDIA (Q1 FY2027 Earnings Call (May 20, 2026))

(Financial coverage has at times used “H20” and “H200” interchangeably in describing China-bound shipments; we quote management's wording verbatim and note that the specific approved product designation may be clarified in subsequent filings.)

For investors, the way to think about China is as a dormant optionthat is not embedded in guidance. NVIDIA's $91 billion Q2 guide assumes zero China Data Center contribution. If licenses convert into shipments and revenue later in the year, that becomes upside to numbers that are already strong. If the policy environment tightens again, there is nothing to lose in the model because nothing is in the model. That asymmetry is favorable — but it is an option, not a certainty, and the timing depends on a regulatory process outside the company's control.

Valuation Reset: What the Market Is Actually Paying For

At a share price hovering around $220-225 in the days after the report, NVIDIA's market capitalization sits near $5.4 trillion (based on roughly 24.4 billion diluted shares). On forward earnings, the stock trades at approximately 26x to 29x— a range rather than a precise figure, because published forward-EPS estimates vary. For context, NVDA has traded across a wide historical multiple range — well above 40x during peak AI-enthusiasm phases and below 20x during consolidation periods — which makes the current high-20s a middle-of-the-range observation rather than either extreme. For a company growing revenue 85% year over year with 75% gross margins and 65% operating margins, a high-20s forward multiple is not, on its face, expensive relative to the growth rate.

The valuation debate is really a debate about durability. A 28x multiple on a business growing 85% looks cheap if that growth persists for years; it looks expensive if growth normalizes toward 20-30% sooner than bulls expect, because the multiple would then compress against decelerating earnings. These are illustrative scenarios reflecting the two ends of the valuation debate, not price predictions. This is why the flat margin guide and the forward indicators matter more than the backward-looking beat. If you want to pressure-test what you'd pay for a given earnings stream, our P/E ratio framework walks through trailing versus forward multiples and why the denominator assumption drives everything.

Analyst price targets cluster well above the current price. Post-earnings, the verifiable targets included Goldman Sachs at $285 (raised from $250), JPMorgan at $280 (raised from $265), Morgan Stanley around $288, and Bank of America at $350, with a published consensus in the roughly $285-$303 range (depending on the aggregator and snapshot date) across a large majority of covering analysts and no sell ratings reported by major aggregators as of late May 2026. Price targets are opinions, not facts — but the near-unanimous positive analyst posture is itself a data point about how the institutional community frames the risk/reward.

Bull Case vs. Bear Case

The honest synthesis: this is a business of exceptional quality trading at a price that already assumes exceptional quality. The bull and bear cases do not disagree much about the company— they disagree about what multiple to pay and how long the growth persists.

The Competitive Landscape: Who Is Actually Catching Up?

NVIDIA controls an estimated 80%+ of the data-center AI accelerator market by revenue, but the competitive set is more active than a year ago. AMD's MI400 series targets the next upgrade cycle with competitive memory capacity and bandwidth, and analysts project meaningful AMD data-center GPU growth in 2026 — though off a far smaller base (per publicly available AMD commentary and analyst projections as of Q1 2026; specific market-share data for the period is not yet published). Broadcom is the key enabler of custom hyperscaler accelerators (Google's TPU, Meta's MTIA) and competes directly in AI cluster networking, making it the most credible long-term challenger to NVIDIA's networking moat. Marvell plays a similar custom-silicon role with other hyperscalers, and Intel's Gaudi line continues to struggle for traction against the entrenched CUDA software ecosystem.

The throughline: NVIDIA's durable advantage is not any single chip but the full-stack combination of silicon, NVLink/Spectrum-X networking, and the CUDA software ecosystem that carries more than a decade of developer lock-in. Competitors can match a FLOP count; matching the system is far harder. But “harder” is not “impossible,” and the most important competitive risk is not a rival GPU — it is hyperscalers designing their own accelerators to reduce dependence on a single supplier.

Catalysts to Watch: Q2 FY2027 and Beyond

The single most important variable between now and the next print is whether hyperscaler capex commentary stays on its upward path. NVIDIA's Data Center revenue is, functionally, a derivative of hyperscaler and sovereign AI budgets. As long as those budgets keep rising, the demand side of the thesis holds; the day they flatten is the day the valuation debate gets decided.

Frequently Asked Questions

What did NVIDIA report in Q1 FY2027?

NVIDIA reported Q1 FY2027 revenue of $81.6 billion (up about 85% year over year), driven by Data Center revenue of $75.2 billion. Non-GAAP EPS was $1.87, beating the roughly $1.77 consensus, and the company guided Q2 FY2027 revenue to approximately $91 billion (±2%), above the $86.8 billion consensus.

Why did NVIDIA stock fall after a record quarter?

The stock fell modestly because expectations were already extremely high. NVDA trades at a high-20s forward earnings multiple and had rallied into the report, so a beat that wasn't dramatically larger than “whisper” expectations triggered profit-taking — the fourth consecutive post-earnings decline. The decline reflects valuation and positioning, not weakness in the quarter.

How much of NVIDIA's revenue comes from Data Center?

Data Center generated $75.2 billion of the $81.6 billion total, or about 92% of revenue. This concentration is both NVIDIA's greatest strength (dominance of the AI compute market) and its most-debated risk (dependence on a single capex-driven demand theme).

Is NVIDIA still making money in China?

Essentially no Data Center revenue was recognized from China in Q1 FY2027. US export licenses for compliant chips have been approved for certain China-based customers, but no revenue had been recognized as of the report. China is best viewed as potential upside not included in current guidance.

What is sovereign AI and why does it matter for NVIDIA?

Sovereign AI refers to governments building their own national AI infrastructure. NVIDIA's CFO said sovereign revenue grew more than 80% year over year, with infrastructure deployed across nearly 40 countries. It matters because it diversifies demand away from US hyperscalers and is largely insulated from US-China export restrictions.

When does NVIDIA report Q2 FY2027 earnings?

NVIDIA typically reports its fiscal Q2 results in late August. The exact date is set by the company closer to the period end; the Q2 FY2027 print will be the key test of whether the $91 billion guidance proves conservative or stretched.

The Bottom Line

NVIDIA's Q1 FY2027 was, by the numbers, a record quarter: $81.6 billion in revenue, 75% gross margins, $48.6 billion in free cash flow, a $91 billion forward guide, a 25x dividend increase, and an $80 billion buyback. The decline in the stock afterward was not a verdict on the business — it was a verdict on the price the market had already paid for it.

The investment debate going forward is not about whether NVIDIA is a great company; the financials settle that. It is about durability and valuation: how long 85% growth and 75% margins persist, whether hyperscaler and sovereign capex keep climbing, how the next-generation product transitions affect margins, and whether the China option ever converts to revenue. Investors who hold conviction that the AI infrastructure build-out is a multi-year secular cycle will read this quarter as confirmation that the runway is long. Investors who believe AI capex is closer to a cyclical peak will see a high-quality business at a full price.

Both views rest on the same facts — they differ on the assumptions about what comes next. What is not in dispute is that the operating quality of the business, by every metric reported this quarter, has not weakened. What has changed is how much the market is willing to pay for it.