NVIDIA Corporation (NVDA) | The Buildout — AI Infrastructure
The Verdict
NVIDIA is a full-stack AI infrastructure company. It designs the accelerators, networking, and software that underpin modern AI factories, and it has expanded into CPUs to orchestrate agentic workloads. Its hardware and CUDA ecosystem are the default platform for training and serving frontier models, and the company sits at the center of an immense buildout of data center capacity aimed at turning computation into a utility.
| Market Cap | — |
| Revenue (TTM) | $253.5B |
| Revenue Growth | +70.7% |
| EBITDA Margin (TTM) | 65.3% |
| Net Cash | $67.8B |
| Earnings Beats | 7 of 7 |
| P/E (TTM) | — |
| EV/EBITDA (TTM) | — |
What We Like
- Revenue hit $81.6 billion in Q1 FY2027, an 85% year-over-year increase and the third consecutive quarter of accelerating YoY growth, with 14 straight quarters of sequential gains.
- A new standalone CPU business (Vera) opened a $200 billion TAM, and management disclosed visibility to nearly $20 billion in CPU revenue this fiscal year — additive to GPU sales.
- The ACIE segment (AI clouds, enterprise, industrial, sovereign) generated $37 billion in Q1, growing 31% sequentially, and is expected by management to eventually exceed hyperscale.
- Total supply commitments reached $145 billion, underpinned by multi-year partnerships with TSMC, SK hynix, and Coherent, locking in capacity for multiple product generations.
- Free cash flow hit a record $48.6 billion, enabling a $0.25 quarterly dividend, an $80 billion buyback authorization, and a plan to return roughly 50% of FCF to shareholders.
What We’re Watching
- Vera Rubin production is set to begin in Q3 FY2027, but the platform integrates 7 custom chips across 5 rack systems and management acknowledged the ramp slope is hard to predict.
- Customer concentration intensified: three direct customers represented 21%, 17%, and 16% of total revenue in Q1, meaning 54% of the business is tied to three accounts.
- Custom silicon at AWS (Trainium at a $20B+ run rate), Google (TPU), and Meta (Broadcom program) could erode NVIDIA's share of hyperscale inference over time.
- China data center compute revenue remains zero and is excluded from the outlook, despite approved H200 licenses, with the possibility of further export restrictions.
The thesis around NVIDIA as the prime beneficiary of the AI buildout has strengthened. Revenue growth is accelerating, the customer base is broadening beyond hyperscale, and a new CPU platform offers a material growth vector that was not priced in a year ago. Supply visibility and massive capital returns add further structural support. The core open question is whether the Vera Rubin transition — the most complex system the company has ever attempted — will execute on time and whether the rising parallel economy of custom silicon among the largest customers will eventually cap NVIDIA's addressable inference market.
Earnings Beat
NVIDIA reported Q1 FY2027 revenue of $81.6 billion, up 85% year-over-year and 20% sequentially. GAAP gross margin held at 74.9%, and Data Center revenue surged 92% to $75 billion as the Blackwell platform became the fastest product ramp in company history. Record free cash flow of $48.6 billion supported a new capital return framework.
| Metric | Q1 FY2027 | Q4 FY2026 | Q1 FY2026 | YoY |
|---|---|---|---|---|
| Revenue | $81.6B | $68.1B | $44.1B | +85.2% |
| Gross margin | 74.9% | 75.0% | 60.5% | +1440bps |
| EBITDA | $54.5B | $45.1B | $22.2B | +145.1% |
| EPS | $2.39 | $1.76 | $0.76 | +213.4% |
| Free cash flow | $48.6B | $34.9B | $26.2B | +86% |
We have visibility to nearly $20 billion in total CPU revenue this year.— Colette Kress, CFO, May 20, 2026
Management tone: Management's tone was confident and expansive. Jensen Huang described demand as 'gone parabolic' and the team volunteered new quantified targets for the CPU business and multi-year revenue visibility. They were candid about the niche nature of LPX and the complexity of the Vera Rubin ramp, reinforcing credibility.
Management Guidance
Management guided Q2 FY2027 revenue to $91 billion ±2%, with GAAP gross margin of 74.9% plus or minus 50 basis points. For the full year, gross margin was reaffirmed in the mid-70s percent range, operating expense growth was raised to the upper-40s percent year-over-year, and the effective tax rate was lowered to 16-to-18 percent. Management also stated it has visibility to nearly $20 billion in standalone Vera CPU revenue this fiscal year and 'full confidence' in $1 trillion of cumulative Blackwell and Rubin revenue from 2025 through calendar 2027.
Trajectory
Revenue growth has accelerated for three straight quarters, with Q1 FY2027 reaching $81.6 billion — an 85% year-over-year increase. GAAP gross margin, which dipped to 60.5% in Q1 FY2026, has stabilized in the mid-70s range for the last three quarters as the Blackwell ramp drove record volumes without diluting profitability. The demand drivers have broadened: the newly reported ACIE segment grew 31% sequentially, outpacing hyperscale, and networking revenue nearly tripled year-over-year as Spectrum-X Ethernet became the dominant AI fabric.
The Model
The model projects FY+1 revenue of $402,000 million and EBITDA of $272,556 million (67.8% margin). For FY+2, revenue of $590,000 million and EBITDA of $399,430 million (67.7% margin). Near-term is anchored by the Blackwell ramp and the contribution from the new Vera CPU, while FY+2 assumes Vera Rubin volume shipments and continued expansion of the ACIE segment beyond hyperscale.
| Metric | FY2026 | Next FY (E) | Following FY (E) |
|---|---|---|---|
| Revenue | $215.9B | $402.0B | $590.0B |
| YoY Growth | — | +86.2% | +46.8% |
| EBITDA | $133.2B | $272.6B | $399.4B |
| EBITDA Margin | 61.7% | 67.8% | 67.7% |
Projections are the median of 5 independent model runs. The model’s revenue sits 5.3% above analyst consensus.
Management guided Q2 FY2027 revenue to $91 billion ±2%, with GAAP gross margin of 74.9% plus or minus 50 basis points. For the full year, gross margin was reaffirmed in the mid-70s percent range, operating expense growth was raised to the upper-40s percent year-over-year, and the effective tax rate was lowered to 16-to-18 percent. Management also stated it has visibility to nearly $20 billion in standalone Vera CPU revenue this fiscal year and 'full confidence' in $1 trillion of cumulative Blackwell and Rubin revenue from 2025 through calendar 2027.
What Could Go Right — and Wrong
- Vera CPU revenue materially exceeds $20 billion, proving NVIDIA can capture a large piece of the $200 billion CPU TAM beyond GPUs.
- The ACIE segment overtakes Hyperscale in absolute revenue and sustains 30%+ sequential growth, demonstrating diversification.
- Vera Rubin ramps on schedule and every frontier model company adopts it from the start, pushing NVIDIA's share of AI compute higher.
- Physical AI revenue scales beyond $20 billion annually as robotaxi fleets and industrial robotics deploy.
- Gross margins remain above 70% even with the shift toward systems, as software and networking offset hardware mix.
- Vera Rubin production is delayed or encounters yield issues, pushing revenue out and allowing competitors time to gain traction.
- Hyperscaler capital spending slows and ACIE growth decelerates sharply, leaving a gap in the growth trajectory.
- Custom silicon at AWS, Google, and Meta takes a material share of inference at the largest accounts, shrinking NVIDIA's addressable market.
- Gross margins slide below 70% as pricing power weakens and the mix shifts toward lower-margin system hardware.
- The financing flywheel that supports AI infrastructure investment breaks if token prices collapse.
Looking Ahead
The next twelve months pivot on the Vera Rubin production ramp starting in Q3 FY2026, the scaling of the standalone Vera CPU business, and whether the ACIE segment's growth can sustain its pace. Key signposts will be quarterly revenue reports, customer announcements of Rubin cluster deployments, and any breakout of standalone CPU revenue that validates management's $20 billion visibility target.
- ~Aug 2026Q2 FY2027 earnings — Tests the $91B revenue guide and whether ACIE growth sustains 30%+ QoQ.
- Q3 FY2027Vera Rubin production shipments — First revenue from next-gen platform; tests execution on the complex 7-chip system.
- FY2027Vera CPU standalone ramp — Measures progress toward the ~$20B standalone CPU revenue target.
- Starting 2026AWS >1M GPU deployment — Confirms hyperscale demand and the scale of Rubin adoption at the largest accounts.
- Q4 CY2026DGX Station for Windows — Expands AI compute into Edge and workstation markets.
- 2026-2027Anthropic multi-cloud capacity — Validates the share-gain narrative as Anthropic moves from near-zero to significant NVIDIA usage.
Financials
Annual Summary
| Metric | FY2026 | TTM |
|---|---|---|
| Revenue | $215.9B | $253.5B |
| Gross Margin | 70.3% | 74.1% |
| EBITDA | $133.2B | $253.8B |
| EBITDA Margin | 61.7% | 65.3% |
| Net Income | $120.1B | $159.6B |
| Free Cash Flow | $96.7B | $191.1B |
| Net Cash | — | — |
Key Ratios (Trailing)
- P/E TTM—
- EV/EBITDA TTM—
- EV/Revenue TTM—
- Price/FCF TTM—
- Gross Margin (TTM)74.1%
- EBITDA Margin (TTM)65.3%
- Net Margin (TTM)63.0%
- ROIC100.4%
- FCF Conversion71.9%
- SBC / Revenue2.7%
The Company
NVIDIA designs the accelerators, networking, and software that form the backbone of modern AI infrastructure. Its GPUs and systems power the training and inference of frontier models, and its CUDA software ecosystem is the default platform across every major cloud and AI lab. The company has recently expanded into standalone CPUs (Vera) and integrated CPU-GPU platforms (Vera Rubin), transforming it from a GPU company into a full-stack data center platform provider.
NVIDIA operates a fabless manufacturing model, outsourcing wafer fabrication primarily to TSMC, with Samsung as an alternate. It relies on third-party foundries, memory suppliers (SK hynix, Micron, Samsung), and assembly and test providers (Hon Hai, Wistron, Fabrinet). The company has deepened its supply partnerships through multi-year agreements, equity investments (Coherent, Lumentum), and $145 billion in total supply commitments, securing capacity for its multi-generation roadmap.
Business Segments
Competitive Landscape
NVIDIA faces a diverse set of competitors across GPUs, networking, and custom silicon. However, its full-stack integration — combining silicon, networking, and CUDA software — creates a moat that is not matched by any single competitor today, particularly in the fragmented non-hyperscale market where customers buy integrated AI factories rather than individual components.
- Named in 10-K as a competitor for discrete GPUs and accelerators.
- Named in 10-K for competing in CPUs and accelerators; not discussed in detail.
- AWSInternal Trainium chips; evidence pack notes $20B+ run rate and expectation of CapEx savings, posing a potential inference share threat over time.
- GoogleTPU development; planning up to 960,000 Rubin GPUs for cloud but simultaneously building internal chips for its own workloads.
- BroadcomCo-designs networking ASICs; also supplies NVIDIA for NVLink Fusion, but listed as a competitor in 10-K.
Supply Chain
NVIDIA sits at the center of a deep supply chain, acting as a dominant customer to its foundry and memory partners and a critical platform provider to its cloud and enterprise customers. No neighbor in the supply chain mentioned NVIDIA anonymously.