Financial Performance and Outlook

FY27Q1 Financial Performance

• Total revenue: $81.6B, up 85% YoY and 20% sequentially, above the prior guidance midpoint of $78B.
• Data Center revenue: $75.2B, up 92% YoY and 21% sequentially. Under the new reporting framework: Hyperscale $37.9B, ACIE (AI Clouds, Industrial & Enterprise) $37.4B.
• Under the prior sub-market framework: Data Center compute $60.4B (up 77% YoY, up 18% QoQ); Data Center networking $14.8B (up 199% YoY, up 35% QoQ).
• Edge Computing revenue: $6.4B, up 29% YoY and 10% QoQ.
• GAAP gross margin: 74.9%; non-GAAP gross margin: 75.0%, approximately flat sequentially.
• GAAP diluted EPS: $2.39 (up 214% YoY); non-GAAP diluted EPS: $1.87 (up 140% YoY). Note: starting Q1 FY27, non-GAAP figures now include stock-based compensation, making them not directly comparable to prior non-GAAP figures.
• Free cash flow: $48.6B, up from $34.9B in FY26Q4 and $26.1B in FY26Q1.

FY27Q2 Guidance

• Revenue: $91.0B ± 2%. No China data center compute revenue assumed.
• GAAP gross margin: ~74.9%; non-GAAP gross margin: ~75.0%, ± 50 bps.
• GAAP operating expenses: ~$8.5B; non-GAAP operating expenses: ~$8.3B.
• Full-year FY27 non-GAAP operating expense growth: upper 40s% YoY (increased from prior guidance of low 40s%).
• Full-year FY27 tax rate: 16.0–18.0% (tightened from prior 17.0–19.0% due to favorable geographic mix).

Vera Rubin Ramp Outlook

• Production shipments commence in Q3 FY27; ramp expected to continue into Q4 and accelerate into Q1 FY28.
• Management stated confidence in $1T in Blackwell and Rubin revenue from 2025 through calendar 2027 (not including standalone Vera CPU revenue).

Capital Returns

• Quarterly dividend increased from $0.01 to $0.25 per share.
• New $80B share repurchase authorization added (in addition to $38.5B remaining).
• NVIDIA plans to return roughly 50% of free cash flow to shareholders in FY27.

Business Drivers

Agentic AI as the Demand Inflection Point

• Jensen Huang characterized the quarter as an arrival of agentic AI at scale: "Agentic AI has arrived. AI can now do productive and valuable work. Tokens are now profitable, so model makers are in a race to produce more."
• Revenue from AI-native clouds, sovereign customers, enterprises, and frontier model builders collectively now equals hyperscale revenue at approximately $37B each. This diversification is a meaningful structural shift from a business previously dependent on a handful of hyperscalers.
• Sovereign AI revenue increased more than 80% YoY in Q1 FY27, with NVIDIA AI infrastructure now deployed across nearly 40 countries representing $50T in GDP.
• The number of partner data centers exceeding 10 MW has nearly doubled in one year, surpassing 80 sites.

Networking Acceleration

• Data Center networking revenue of $14.8B grew 199% YoY and 35% QoQ in FY27Q1, accelerating sharply.
• InfiniBand grew more than 4x YoY, driven by next-generation XDR technology deployments.
• Spectrum-X Ethernet is now described as larger than all Ethernet networking peers combined.
• NVLink scale-up fabric revenue continues to compound as GB300 NVL72 rack deployments expand.

Vera CPU: A New TAM

• Management disclosed ~$20B in standalone Vera CPU revenue visibility for FY27, from a TAM NVIDIA has never previously addressed (~$200B per Jensen Huang).
• Vera is designed for agentic AI orchestration: as agent populations scale, the CPU harness running orchestration, tool use, and I/O becomes a meaningful workload distinct from GPU inference. Jensen Huang estimated billions of agents would eventually each require the equivalent of a PC for tool use.
• Vera is used in four configurations: paired with Rubin GPUs in VeraRubin racks, standalone CPU, Vera + CX9 for storage, and Vera + CX9 for confidential computing.
• Every major hyperscaler and system maker is partnering on Vera deployment according to management.

Frontier Model Coverage Expanding

• The addition of Anthropic as a deep technical partner is significant: Anthropic was largely absent from NVIDIA's infrastructure prior to FY26Q4, and its Claude platform is among the fastest-growing AI services. Management expects to bring substantial compute capacity online for Anthropic across AWS, Azure, CoreWeave, and others.
• OpenAI launched GPT-5.5 on Blackwell, currently at the top of Artificial Analysis leaderboards. Microsoft's Farweave, described as the world's most powerful AI data center, is now live on hundreds of thousands of Blackwell GPUs, ahead of schedule.
• AWS will add more than 1 million Blackwell and Rubin GPUs and is collaborating on Spectrum networking.

Inference Share Gains

• Management stated NVIDIA is growing share of inference rapidly, driven by frontier model proliferation (adding Cursor, Perplexity, TML, Reflection) and the Anthropic partnership.
• Jensen Huang described the ACIE segment as serving customers who "do not want to design, do not want to build" and need a fully integrated platform—a market almost exclusively served by NVIDIA.
• On the competitive question around custom ASICs and LPX-style decode accelerators, Jensen Huang reiterated that LPX remains a niche product for high-token-rate, premium-priced services, and that context processing and breadth of model support favor full-rack NVIDIA systems for most workloads.

H100 Cloud Pricing as a Demand Signal

• Management noted the rental price of H100 instances has risen 20% year-to-date, and A100 cloud pricing is up nearly 15%—cited as evidence that even 6-year-old Ampere GPUs remain fully utilized, validating the structural demand argument.

OpEx Acceleration

• Full-year FY27 non-GAAP operating expense growth guidance was raised to the upper 40s% (from low 40s%), driven by higher R&D spending and an acceleration in AI tool usage to enhance productivity. This is a meaningful increase and will bear watching relative to revenue growth.

One-Offs

Non-GAAP Definition Change: Stock-Based Compensation Now Included

• Starting Q1 FY27, NVIDIA's non-GAAP financial measures now include stock-based compensation (SBC). Historical figures have been restated. Prior period non-GAAP EPS comparisons in the Q1 FY27 filing are stated on the new (SBC-included) basis, which is why non-GAAP diluted EPS of $1.87 in Q1 FY27 is not directly comparable to older non-GAAP EPS figures reported before the restatement. Investors using sell-side models built on pre-change non-GAAP figures should rebase.

Equity Securities Gains: $15.9B GAAP Windfall in Q1 FY27

• GAAP net income of $58.3B (vs. non-GAAP net income of $45.5B) was inflated by $15.9B in unrealized gains from publicly held and non-marketable equity securities—reflecting appreciation of NVIDIA's investments in AI model companies (OpenAI, Anthropic, xAI, etc.). These gains are excluded from non-GAAP results but make GAAP EPS of $2.39 substantially higher than operating earnings.

China Revenue: Zero in Q1 FY27, Zero Assumed in Q2

• No China data center compute revenue was recognized in Q1 FY27. NVIDIA states it has "yet to generate any revenue" from H200 licenses approved by the U.S. government and is "uncertain whether any imports will be allowed into the country." This compares to $4.6B of H20 revenue in Q1 FY26 (before the export ban). Year-over-year growth rates throughout FY27 benefit from this zero-baseline comparison on the China line.

DSO Timing Benefit in Q1; Expected to Normalize

• Days sales outstanding was 45 days in Q1 FY27, below the typical mid-50s range, due to favorable timing of cash collections. Management guided DSO to return to the mid-50s in Q2, which will reduce cash flow from operations relative to Q1.

Supply-Related Commitments at $119B

• Total supply-related commitments (inventory + purchase obligations) reached $119B at end of Q1 FY27, up from $95B at FY26Q4 end. This extends further in time than usual, reflecting both Blackwell demand and Rubin ramp commitments. A demand deceleration while commitments are this elevated would create write-down risk similar to the $4.5B H20 charge in FY26Q1.

Thesis Updates

ACIE Segment Validates Demand Diversification

• The introduction of the new reporting framework reveals that non-hyperscale data center revenue (ACIE: AI clouds, industrial, enterprise, sovereign) at $37.4B in Q1 FY27 is roughly equal to hyperscale revenue at $37.9B. ACIE grew 31% QoQ vs. 12% QoQ for hyperscale—a meaningful divergence suggesting the non-hyperscale demand is accelerating faster.
• This partially addresses the bear concern about customer concentration: a broader and more fragmented customer base reduces the risk that any single customer's spending decisions create a sharp revenue discontinuity.
• However, ACIE includes AI-native clouds like CoreWeave, which are themselves heavily financed. Management's commentary on the growing ecosystem of partner data centers exceeding 10 MW (now 80+ sites) is consistent with broad-based adoption, but the credit quality of some ACIE customers is less certain than the hyperscalers.

Vera CPU Changes the Long-Term Revenue Framework

• The $20B standalone CPU revenue target for FY27 is a material new disclosure. This was explicitly excluded from the prior $1T Blackwell/Rubin revenue framework, so it represents upside to prior expectations.
• If agentic AI scales as management envisions—billions of agents each requiring CPU-based orchestration—Vera addresses a recurring infrastructure demand that would compound alongside GPU demand. Investors who modeled NVIDIA purely as a GPU company need to incorporate this.
• The $200B CPU TAM claim is large. NVIDIA's credibility in CPUs is unproven at scale, though management stated that every major hyperscaler and system maker is partnering on Vera deployment.

Vera Rubin Ramp: Management's Track Record Matters

• Production shipments of Vera Rubin begin Q3 FY27, with a meaningful Q4 ramp expected. Colette Kress was careful to note timing uncertainty, stating "it's hard to say at this point what will be a faster ramp" than GB300. GB300 was described as the fastest ramp in company history—so Rubin faces a high bar.
• Architecture transitions have historically created gross margin compression (see Q1 FY26 with Blackwell). If Rubin ramps aggressively in H2 FY27, margins could face pressure. Management reiterated a mid-70s full-year gross margin target but noted input cost pressures.

Hyperscaler CapEx Sustainability

• Analyst Ben Reitzes pressed management on whether NVIDIA can grow faster than hyperscaler CapEx. Jensen Huang's argument rests on the ACIE segment growing faster than hyperscale—and the Q1 data supports this narrative with ACIE at $37.4B vs. hyperscale at $37.9B and growing at a faster rate.
• Management forecasts hyperscaler CapEx exceeding $1T by 2027 and reaffirmed the $3–4T annual AI infrastructure target by end of the decade. These are management projections, not consensus estimates, and investors should treat them as directional rather than commitments.

ASIC and Custom Silicon Competition

• Jensen Huang addressed the custom ASIC question directly, arguing that LPX-style decode accelerators are niche products limited by SRAM-based context constraints and low throughput. The core argument—that NVIDIA runs every frontier model across every phase of the AI pipeline—remains intact as long as model architectures continue to evolve rapidly.
• The more important competitive question is whether the ACIE segment, which Jensen describes as requiring a fully integrated platform that "they do not want to design or build," is genuinely NVIDIA-exclusive, or whether alternative platforms can compete over time as the segment matures. No competitive defections were disclosed this quarter.