NVDA Q1 Earnings: AI Infrastructure Boom Drives 69% Revenue Surge, but China Ban Clouds Margin Outlook
wealthvista.top Editorial · May 12, 2026 · 11 min read
Executive Summary
NVIDIA reported Q1 FY26 revenue of $44.1 billion, up 69% year-over-year and 12% sequentially, beating consensus estimates by roughly $900 million. The headline numbers look solid until you scratch the surface: a $4.5 billion charge related to the H20 export ban crushed gross margins, pushing GAAP gross margin down to 60.5% from 73.0% in Q4 FY25. Strip out that one-time hit, and EPS would have been $0.96 rather than $0.81 reported. Data Center continues to drive the business, delivering $39.1 billion in revenue, up 73% year-over-year. Q2 guidance of $45.0 billion comes with an asterisk — roughly $8 billion in H20 revenue is now written off due to China export restrictions. The bull case is straightforward: Blackwell is in full-scale production, AI inference demand is accelerating, and sovereign AI spending is a genuine multi-year tailwind. The bear case is equally valid: China exposure is material, margins face permanent compression from export controls, and the valuation leaves no room for error.
1. Quarter Highlights vs. Expectations
| Metric | Q1 FY26 Actual | Q4 FY25 | Q1 FY25 | QoQ | YoY | Consensus Estimate |
|---|---|---|---|---|---|---|
| Revenue | $44.062B | $39.331B | $26.044B | +12% | +69% | ~$43.2B (beat) |
| GAAP Gross Margin | 60.5% | 73.0% | 78.4% | -12.5 pts | -17.9 pts | ~73% (miss) |
| Non-GAAP Gross Margin | 61.0% | 73.5% | 78.9% | -12.5 pts | -17.9 pts | ~73.5% (miss) |
| GAAP EPS | $0.76 | $0.89 | $0.60 | -15% | +27% | ~$0.75 (beat) |
| Non-GAAP EPS | $0.81 | $0.89 | $0.61 | -9% | +33% | ~$0.75 (beat by 8%) |
| Non-GAAP EPS ex-H20 | $0.96 | — | — | — | — | — |
The EPS beat was 8%, though it would have been a larger beat had the Street been modeling H20 headwinds more aggressively. The real story is the margin hit from the H20 charge. GAAP gross margin of 60.5% was the lowest since the crypto downturn, down 12.5 percentage points sequentially.
Revenue Breakdown
Data Center drove the lion’s share at $39.1 billion, up 10% from Q4 and up 73% from a year ago. This segment now represents 89% of total revenue. The 73% year-over-year growth is remarkable given the size. Blackwell NVL72 systems are in full production across hyperscalers and cloud providers, and the jump from Hopper to Blackwell architecture is driving ASP upside.
Gaming hit a record $3.8 billion, up 48% sequentially and 42% year-over-year, powered by Blackwell GPU rollouts (RTX 5070, 5060) and the Nintendo Switch 2 launch. The sequential jump of 48% from Q4 to Q1 is unusual seasonal strength, suggesting Blackwell demand is pulling forward upgrade cycles.
Professional Visualization was $509 million, flat sequentially but up 19% year-over-year, a solid but unspectacular performer.
Automotive came in at $567 million, down 1% sequentially but up 72% year-over-year, driven by DRIVE AGX Thor shipments and robotics momentum. The sequential dip is a small concern given the humanoid robotics thesis.
Margin Performance
The H20 charge dominated the margin story this quarter. GAAP gross margin of 60.5% and non-GAAP of 61.0% versus 73.0% and 73.5% in Q4. Excluding the $4.5 billion H20 charge, non-GAAP gross margin would have been 71.3% — still below Q4, likely reflecting early Blackwell ramp costs.
Operating leverage shows through despite the margin compression. Non-GAAP operating income of $23.3 billion grew 29% year-over-year, and GAAP net income of $18.8 billion grew 26% year-over-year. The company is still printing money even after absorbing a $4.5 billion one-time charge.
2. Business Segment Analysis
Data Center — The Growth Engine
Data Center brought in $39.1 billion, up 73% year-over-year. That’s a $17 billion revenue jump in four quarters. The driver is Blackwell, which entered full-scale production this quarter. Jensen Huang calls it a “thinking machine” for AI reasoning workloads. The Hopper-to-Blackwell jump delivers up to 30x higher inference throughput per MLPerf benchmarks — a genuine step change.
Key customer and partnership developments include:
- Sovereign AI factories in Saudi Arabia (with HUMAIN), UAE (Stargate UAE with G42/OpenAI/Oracle/SoftBank), France, Germany, Italy, Spain, and the U.K. — countries treating AI infrastructure as national security assets
- Hyperscaler availability: Blackwell cloud instances now live on AWS, Google Cloud, Microsoft Azure, and Oracle Cloud Infrastructure
- Enterprise expansion: RTX PRO servers targeting IT infrastructure upgrades, NVIDIA DGX Spark/Station for personal AI supercomputers
- Open AI ecosystem: Llama Nemotron family of open reasoning models, integration with OpenAI’s GPT-oss models
The data center business faces one major risk: export controls. The H20 ban cost NVIDIA $4.5 billion in inventory charges and $2.5 billion in deferred revenue in Q1. For Q2, NVIDIA has not assumed any H20 shipments to China, creating an $8 billion revenue hole.
Gaming — Record Quarter Driven by Blackwell
Gaming at $3.8 billion is a record, up 42% year-over-year. The Nintendo Switch 2 launch (powered by an NVIDIA processor with DLSS) adds a new revenue stream with console-like recurring royalties. Blackwell desktop GPUs (RTX 5070 at $299, RTX 5060) are driving PC upgrade cycles. DLSS 4 adoption in 125+ games (Black Myth Wukong, DOOM: The Dark Ages, Marvel Rivals) is expanding the performance-per-dollar value proposition for gamers. The 48% sequential jump is unusual — typically gaming is flat-to-slightly-down Q4 to Q1 — suggesting Blackwell pre-builds and new product launches pulled forward demand.
Automotive and Robotics
At $567 million, up 72% year-over-year, this remains a small but growing piece of the puzzle. The humanoid robotics thesis is real: Isaac GR00T N1 (open humanoid robot foundation model) launched, followed by GR00T N1.5. NVIDIA’s Cosmos world foundation models and physical AI data tools position the company for the next wave of robotics. DRIVE AGX Thor is now in production shipment. The 1% sequential decline is worth watching — this segment may be supply-constrained.
3. Management Guidance vs. Street Expectations
Q2 FY26 Guidance (quarter ending July 2025):
- Revenue: $45.0 billion ± 2%, reflecting approximately $8 billion in lost H20 revenue from export controls
- GAAP gross margin: 71.8% ± 50 bps; Non-GAAP: 72.0% ± 50 bps
- GAAP operating expenses: ~$5.7 billion; Non-GAAP: ~$4.0 billion
- Full-year FY26 operating expense growth: mid-30% range (GAAP), mid-to-high 30s (non-GAAP)
The guidance is in line or slightly ahead of Street expectations on revenue. The gross margin recovery to 72% from 61% in Q1 is a sharp rebound. NVIDIA expects to exit FY26 with non-GAAP gross margins in the mid-70s, meaning Q3 and Q4 margins should continue climbing as the H20 inventory situation clears and Blackwell yields improve.
The $8 billion H20 revenue loss for Q2 is a large number — essentially the entire China data center opportunity going to zero. But Blackwell demand from everywhere else nearly offsets it. The $45 billion Q2 guidance implies roughly $53 billion in normalized revenue, minus $8 billion from China controls.
Q3 FY26 guidance from Q2 results (for context):
- Revenue: $54.0 billion ± 2% — this was guided in Q2 earnings call
- Gross margins continue to recover
Analyst consensus (as of May 2026):
- 54 analysts covering, consensus Buy rating
- Average 12-month price target: $275–278
- Forward PE: 26x FY2026 earnings, 19x FY2027 earnings
- FY2027 revenue consensus: ~$375 billion
4. Balance Sheet and Cash Flow Health
NVIDIA’s balance sheet remains strong, though the H20 charge weighed on reported Q1 numbers.
Based on Q1 FY26 data (ended April 27, 2025):
- The company generated strong operating cash flow, well above net income
- Q1 operating income of $21.6 billion GAAP / $23.3 billion non-GAAP on $44 billion revenue implies roughly 50% operating margins excluding the H20 charge
- Capital returns: In H1 FY26, NVIDIA returned $24.3 billion to shareholders via buybacks and dividends
- Share repurchase authorization expanded: An additional $60.0 billion authorized in Q2 FY26, with $14.7 billion remaining at quarter-end
- Cash dividend: $0.01 per share quarterly — modest, but the buyback program is the primary capital return tool
The buyback expansion to $60 billion signals management conviction in long-term value creation. At ~$222 per share, this authorization can retire roughly 270 million shares — real EPS accretion over time.
The H20 charge of $4.5 billion is a one-time hit to inventory, not a cash flow item in the traditional sense, though it does tie up working capital. The company indicated $2.5 billion in H20 revenue that could not be recognized. The $8 billion revenue loss guidance for Q2 reflects the complete cessation of China H20 shipments.
5. Valuation Assessment
At approximately $222 per share (May 18, 2026 close) and a market cap near $540 billion, NVIDIA trades at:
| Metric | Value | Commentary |
|---|---|---|
| Forward P/E (FY2026, ~$4.90 EPS) | ~45x | Elevated, reflecting AI premium |
| Forward P/E (FY2027, ~$8.47 EPS) | ~26x | More reasonable for a growth company |
| EV/EBITDA (est.) | ~25–30x | Premium to peers but justified by growth |
| P/S (FY2026, ~$216B revenue) | ~2.5x | Lower than peak multiples |
The stock has compressed from its 2024 peak P/E of 60x+ as earnings outpaced the stock price. The FY2027 multiple of 26x is reasonable for a company growing earnings at 72% annually. But this is not a stock for the faint-hearted: a single earnings miss or a slowdown in hyperscaler capex could send it down 20–30%.
AMD trades at roughly 20x forward earnings, though at a fraction of NVIDIA’s AI chip revenue. Broadcom sits at 15–18x forward but lacks pure-play AI exposure. Intel is a restructuring story and not directly comparable. The AI semiconductor market is effectively NVIDIA plus AMD for training and inference GPUs, and NVIDIA’s CUDA ecosystem is a moat that competitors cannot easily replicate.
6. Competitive Positioning and Catalysts
Competitive Moat
NVIDIA’s moat is not just silicon — it’s the entire software stack. CUDA, TensorRT, cuDNN, and NIM microservices create switching costs that customers cannot justify. When a cloud provider spends $10 billion on NVIDIA GPUs, rewriting training pipelines for AMD ROCm is not a realistic option. This software moat may be worth more than the hardware advantage itself.
Blackwell’s 30x inference throughput improvement versus Hopper extends the performance gap. Even if AMD’s MI300X/MI350 is competitive on paper, the software ecosystem means NVIDIA retains 80%+ market share in AI training.
Key Catalysts
- Blackwell Ultra production ramp in H2 FY2026 — higher ASPs drive margin expansion
- Sovereign AI factory deals — Saudi Arabia, UAE, European nations treating AI infrastructure as a national priority
- Enterprise AI factory adoption beyond hyperscalers — corporations building their own AI infrastructure
- Humanoid robotics commercialization — Tesla Optimus, Figure, 1X deploying GR00T models
- AI agent proliferation — reasoning models require more inference compute, NVIDIA benefits directly
- GB200 NVL72 system deployments — each rack-scale system represents $10+ million in revenue
Competitive Risks
AMD’s MI350 and Intel’s Gaudi 3 are real alternatives for inference workloads where performance-per-dollar matters more than raw throughput. Amazon’s Trainium and Google’s TPU v5 are increasingly used for in-house workloads, which trims the addressable market for NVIDIA’s cloud customers. That said, no competitor offers the full stack — GPU plus networking via NVLink/NVSwitch plus software — that NVIDIA provides.
7. Key Risks
Risk 1: China Export Controls — Permanent Revenue Loss
The H20 export ban is not a temporary headwind. NVIDIA built a China-specific product to comply with earlier rules, then the U.S. government shifted mid-stream. If H20 had shipped freely, Q1 revenue would have been roughly $51 billion instead of $44 billion. The $8 billion quarterly run-rate loss in Q2 represents about 15–18% of revenue. If rules tighten further — potentially affecting future architectures — the impact could be larger. This is a geopolitical risk that operational excellence cannot offset.
Risk 2: Customer Concentration and Capex Cycle
The hyperscalers — Microsoft/Azure, Amazon/AWS, Google, Meta, Oracle — represent 40–50% of Data Center revenue. If any of them meaningfully cuts AI capex, NVIDIA’s revenue would face a sharp correction. The generative AI investment cycle has been exceptional, but extrapolating 70%+ growth indefinitely is risky. A 20–30% reduction in hyperscaler AI spend would be material.
Risk 3: Valuation Risk — No Room for Error
At 26x FY2027 earnings, NVIDIA is priced for near-perfection. Any earnings miss, guidance cut, or macro headwind could compress the multiple quickly. The stock historically trades in a wide band (15x to 60x earnings) depending on growth expectations. The current 26x forward multiple leaves limited downside protection if the AI narrative cools.
8. Investment Conclusion
BUY with a 12-month price target of $275–285, roughly 25% upside from current levels, based on 26–28x FY2027 earnings.
The bull case: NVIDIA is the essential infrastructure provider for the AI revolution. Blackwell demand is strong, sovereign AI spending adds a new customer category, and the inference market (AI agents, reasoning models) is in its early innings. Q3 FY26 revenue guidance of $54 billion shows the company is firing on all cylinders.
The bear case: China export controls create a permanent 15–18% revenue headwind, hyperscaler capex concentration introduces cyclical risk, and 26x FY2027 earnings leaves no margin of safety if growth slows even slightly. Blackwell supply constraints could also delay revenue recognition.
The H20 charge obscured what was actually a stellar quarter. Strip out the one-time impact and EPS would have been $0.96 versus $0.81 reported — a much bigger beat. The margin recovery trajectory (60.5% → 71.8% in Q2 → mid-70s by year-end) is the key number to watch. If gross margins reach 75% by Q4 FY26 as guided, the earnings power of this business is even more extraordinary than the headline numbers suggest.
Sources: NVIDIA Q1 FY26 Earnings Press Release — NVIDIA News · NVIDIA Q2 FY26 Earnings Press Release — NVIDIA News · StockAnalysis.com — NVDA Analyst Forecasts · MarketBeat — NVDA Price Targets