NVIDIA Corporation (NVDA) — Company Research

Last Updated: 12 May 2026

NVIDIA Corporation (NASDAQ: NVDA) is the world's dominant supplier of graphics processing units for AI training and inference, data centre compute, gaming, and professional visualisation. Fiscal year 2026 (ended 25 January 2026) was the most profitable year in the company's history: revenue reached $215.9 billion (+65% year-on-year), net income $120.1 billion, and GAAP diluted EPS $4.90. Data Centre — driven by the Hopper H100/H200 and the newly ramped Blackwell B200/GB200 architectures — contributed $197.3 billion of that total, or approximately 91% of group revenue. On 12 May 2026 the stock hit an intraday all-time high of $223.75, valuing the company at roughly $5.4 trillion. Q1 FY2027 earnings are scheduled for 28 May 2026.

1. Company Snapshot

Full name	NVIDIA Corporation
Ticker	NVDA (Nasdaq: NMS)
Sector / Industry	Technology — Semiconductors & Semiconductor Equipment
Founded	1993, Santa Clara, California
CEO	Jensen Huang (co-founder)
Employees	~36,000 (FY2026 10-K)
Exchange	Nasdaq
Market cap (12 May 2026)	~$5.4 trillion
Revenue (FY2026)	$215.9 billion
Net income (FY2026)	$120.1 billion
GAAP EPS diluted (FY2026)	$4.90
Fiscal year end	Last Sunday of January
Website	nvidia.com / investor.nvidia.com

2. Bull Case vs Bear Case

Distilled from the full report below — factual only, no ratings.

Bull Case

• AI compute monopoly: Data Centre revenue reached $197.3bn in FY2026 — 91% of total — driven by Hopper H100/H200 and a full ramp of Blackwell B200/GB200. No competitor ships accelerators at remotely comparable volume or software-ecosystem depth.
• Vera Rubin production confirmed: Jensen Huang confirmed at CES 2026 that Vera Rubin NVL72 entered full production, with shipments to Microsoft, Google, Amazon, Meta, and Oracle expected from July 2026. The company cited $1 trillion in combined Blackwell and Rubin orders through calendar 2027.
• CUDA moat: Over a decade of CUDA developer investment creates extremely high switching costs. The installed base of CUDA-trained engineers and CUDA-optimised models makes re-targeting to AMD or custom silicon deeply disruptive for most customers.
• Gross margin expansion: FY2026 GAAP gross margin of 71.1% shows the business is not competing on price despite rapid volume growth; Blackwell carries similar or better margins than Hopper.
• Automotive compounding: NVIDIA's Drive platform is in production with Mercedes-Benz, BYD, Volvo, and others; automotive and robotics revenue is growing from a small base but adds a second structural growth vector beyond data centre.

Bear Case

• H20 export controls: The U.S. government imposed a licence requirement for H20 exports to China in April 2025, resulting in a $4.5 billion charge in Q1 FY2026 and approximately $8bn in lost annual run-rate revenue. Further export tightening remains a policy risk.
• Customer concentration and in-house silicon: Microsoft, Google, Meta, Amazon, and Oracle collectively represent a very large share of Data Centre revenue. All are developing or deploying custom AI ASICs (Google TPU6, AWS Trainium, Microsoft Maia) that may displace NVIDIA at the margin over time.
• Valuation multiple: At ~44x trailing GAAP PE and ~25x trailing P/S, the stock prices in sustained hyper-growth. Any quarter where revenue misses expectations or guidance disappoints has historically caused sharp drawdowns.
• Supply-chain dependency: NVIDIA designs but does not manufacture its chips; TSMC CoWoS packaging capacity constraints have limited Blackwell ramp speed. Disruption at TSMC (geopolitical or operational) would be immediately material.

3. What Does This Company Actually Do?

NVIDIA designs and licenses graphics processing units and system-on-chip products. It does not manufacture; all production is outsourced to TSMC (primary), Samsung, and packaging specialists. The business generates revenue in two reportable segments: Compute & Networking (data centre AI accelerators, networking, automotive Drive SoCs, CUDA software) and Graphics (GeForce gaming GPUs, professional Quadro/RTX workstation cards, OEM). A third lens is geographic: approximately 57% of FY2026 revenue came from customers in the United States, with the remainder from Taiwan, China (now declining due to export controls), Europe, and Asia.

The core product cycle is: design a GPU architecture (Hopper, Blackwell, Rubin), license manufacturing to TSMC, sell to hyperscalers, cloud providers, enterprises, researchers, and OEM system builders. Software revenue via NVIDIA AI Enterprise and DGX Cloud is growing but remains a small fraction of total.

Segment	% of revenue	What it is
Data Centre (Compute & Networking)	~91% ($197.3bn)	H100/H200/B200/GB200 AI accelerators, NVLink, InfiniBand and Spectrum-X networking, DRIVE automotive SoCs, CUDA software platform. Revenue grew +93% YoY in FY2026. Customers include Microsoft, Google, Amazon AWS, Meta, Oracle, and thousands of enterprise AI deployments.
Gaming (Graphics)	~4% (~$11.4bn est.)	GeForce RTX 40/50-series desktop and laptop GPUs. Consumer market; benefits from AI-rendered super-resolution (DLSS) as a hardware differentiator. Revenue grew but at a slower pace than Data Centre.
Professional Visualisation	~1% (~$2.2bn est.)	Quadro/RTX workstation GPUs for designers, engineers, and AI workstation operators. Q4 FY2026 saw 159% YoY growth as AI workstation demand accelerated.
Automotive & Other	~1% (~$2.5bn est.)	DRIVE Orin/Thor SoCs for autonomous and ADAS vehicles; robotics. Production with Mercedes-Benz, BYD, Volvo, Hyundai, Toyota. Growing structurally as NVIDIA expands physical AI.

4. The Business Model

How NVIDIA makes money. NVIDIA earns revenue primarily from selling accelerator hardware (GPUs, networking cards, rack systems) to hyperscalers and cloud providers, then to enterprise customers and system builders. The Blackwell NVL72 rack — a 72-GPU, 36-CPU Grace Blackwell rack system — sells for approximately $3 million per rack, versus $250,000–$400,000 for an 8-GPU Hopper server. Hardware gross margins have expanded alongside scale, reaching 71.1% GAAP for FY2026 as Blackwell volumes ramped.

Unit economics. NVIDIA's gross margin trajectory is unusual for a hardware company: margins have expanded even as volumes grew exponentially, because NVIDIA holds pricing power that commodity chip vendors do not. The company reinvests heavily in R&D (approximately $8.5bn in FY2026, ~4% of revenue) to maintain the architectural lead that sustains pricing. Operating income was $130.4bn on $215.9bn revenue — a 60% operating margin.

Moat. The CUDA software ecosystem is the primary moat. Over one million developers write CUDA code; virtually every large AI model has been trained or optimised on NVIDIA hardware with CUDA. Switching to AMD ROCm or Google TPUs requires significant re-engineering effort. Secondary moats include: NVLink high-speed GPU interconnects (proprietary, not available on AMD), InfiniBand networking (acquired via Mellanox), and decades of relationships with data-centre architects at hyperscalers.

Subsidy/regulatory credit dependency. NVIDIA does not receive material government subsidies. However, the business is significantly exposed in the opposite direction: U.S. export controls on advanced chips to China represent a material revenue risk. The April 2025 H20 licence requirement cost approximately $4.5bn in Q1 FY2026 alone. NVIDIA has stated the company will not circumvent export rules and is working on alternative China-market products that remain within regulatory bounds.

Manufacturing dependency. NVIDIA is fabless — entirely dependent on TSMC for leading-edge node production (TSMC N4P for Blackwell) and CoWoS advanced packaging. TSMC capacity constraints have at times limited Blackwell shipment ramp speed. This dependency is structural and cannot be quickly resolved.

5. Financial Health

All figures sourced from NVIDIA's official quarterly earnings press releases on nvidianews.nvidia.com and investor.nvidia.com. NVIDIA's fiscal year ends on the last Sunday of January.

Fiscal year	Revenue ($bn)	YoY %	GAAP EPS (diluted)	Non-GAAP EPS (diluted)	Dividend/share	Long-term debt (YE)
FY2022 (ended Jan 2022)	$26.9bn	+61%	$0.38	$0.53	$0.01	$10.9bn
FY2023 (ended Jan 2023)	$27.0bn	+0.4%	$0.17	$0.49	$0.01	$9.7bn
FY2024 (ended Jan 2024)	$60.9bn	+126%	$1.19	$1.45	$0.01	$8.5bn
FY2025 (ended Jan 2025)	$130.5bn	+114%	$2.94	$2.99	$0.01	$8.5bn
FY2026 (ended Jan 2026)	$215.9bn	+65%	$4.90	$4.77	$0.01	~$8.5bn

Note: EPS figures are split-adjusted for the 10-for-1 stock split completed June 2024. FY2022–FY2023 EPS derived from NVIDIA's historical filings at investor.nvidia.com. Non-GAAP EPS excludes stock-based compensation and acquisition-related amortisation. Long-term debt figures from balance sheet as of fiscal year-end.

Quarterly revenue (FY2026, most recent first):

Quarter	Revenue ($bn)	YoY %	Data Centre Revenue ($bn)
Q4 FY2026 (Oct–Jan 2026)	$68.1bn	+73%	$62.3bn
Q3 FY2026 (Jul–Oct 2025)	$57.0bn	+62%	~$52.2bn est.
Q2 FY2026 (Apr–Jul 2025)	$46.7bn	+56%	~$42.0bn est.
Q1 FY2026 (Jan–Apr 2025)	$44.1bn	+69%	$39.1bn
FY2026 Total	$215.9bn	+65%	$197.3bn

Cash and debt: NVIDIA held approximately $43bn of cash, cash equivalents, and short-term investments at end of FY2026, against approximately $8.5bn of long-term debt — a very strong net cash position. Free cash flow for FY2026 was in excess of $100bn. The company returned capital via dividends ($0.01/share quarterly) and a $50bn share repurchase programme authorised in FY2025 of which substantial portions were executed.

Capex: As a fabless company NVIDIA's own capex is modest (~$1.5bn in FY2026 on property and equipment). The heavy capital investment is on TSMC's balance sheet. NVIDIA does invest in leased data centre space for its own inference cloud (DGX Cloud) but this is not material at the group level.

6. Valuation & Market Data

Raw metrics, May 2026. Not opinions on whether the stock is cheap or expensive.

Metric	Value
Stock price (12 May 2026 close)	~$216–224 (intraday high: $223.75)
Market cap	~$5.4 trillion
Enterprise value	~$5.35 trillion (net cash position reduces EV below market cap)
Trailing P/E	~44x (GAAP; $216 / $4.90)
Forward P/E	~28x (consensus estimate based on Q1 FY2027 guidance of ~$45bn revenue; management has not guided EPS)
P/S (TTM)	~25x ($5.4T / $215.9bn)
EV/EBITDA (TTM)	~37x (approximate; operating income $130.4bn + D&A)
P/FCF	~50x (approximate; FCF ~$100bn–$108bn FY2026)
52-week high	$223.75 (12 May 2026)
52-week low	$120.28
Short interest (% of float)	~0.7% (low; MarketBeat May 2026)
Days to cover	~1 day

Note: Market data sourced from public exchange data and MarketBeat, May 2026. Valuation multiples use FY2026 reported figures from NVIDIA's official press releases.

7. What Are They Building / What's Coming?

Vera Rubin platform (2H 2026). Jensen Huang announced at CES 2026 that Vera Rubin NVL72 entered full production. Trial production with ODM partners is expected in June 2026, with shipments to Microsoft, Google, Amazon, Meta, and Oracle commencing in July 2026. The Rubin architecture promises up to 5x training performance improvement over Blackwell in some workloads and 10x lower inference token cost. NVIDIA cited combined Blackwell and Rubin order visibility of approximately $1 trillion through end of calendar 2027 at GTC 2026 in March.

Rubin Ultra and beyond. NVIDIA presented a two-year chip cadence at GTC 2026: Rubin Ultra (2H 2027) and Feynman (2028+). The Feynman GPU is expected to use a next-generation packaging technology and integrate quantum computing co-processors.

DGX Cloud and software revenue. NVIDIA offers DGX Cloud — a managed AI supercomputing service on Azure, Google Cloud, Oracle, and Amazon AWS. AI Enterprise software licensing is growing but remains a small fraction of group revenue. The company is building its software business to provide recurring revenue alongside hardware cycles.

Physical AI and robotics. NVIDIA's Isaac robotics platform and DRIVE Thor automotive SoC are in active deployment. At GTC 2026, Huang described physical AI (robots, autonomous vehicles, industrial automation) as the next major compute wave after cloud AI training. The company partnered with BMW, Foxconn, and BYD on factory automation using Isaac Sim and Omniverse digital-twin software.

NVIDIA and Corning partnership. On 8 May 2026, NVIDIA and Corning Incorporated announced a multi-year commercial and technology partnership to expand U.S.-based manufacturing of advanced optical connectivity solutions for next-generation AI infrastructure — supporting U.S. domestic supply-chain goals.

Board change. Suzanne Nora Johnson (former Goldman Sachs Vice Chairman) was appointed to NVIDIA's board of directors, effective 13 July 2026, as announced 8 May 2026.

8. Competitive Landscape

The AI accelerator market is effectively NVIDIA vs everyone else. Industry estimates put NVIDIA's share of the discrete AI GPU market at 70–90% depending on segment (training vs inference, cloud vs enterprise). The primary challengers are AMD in merchant silicon and the hyperscalers building custom ASICs.

Peer	Market cap (approx, May 2026)	Notable AI compute product	Positioning vs NVIDIA
AMD (AMD)	~$210–230bn	Instinct MI300X / MI350 accelerators; ROCm software stack	Sole credible merchant GPU alternative at scale; gaining share but AI GPU revenue (~$5–8bn estimated FY2025) is roughly 30–40x smaller than NVIDIA's Data Centre segment. ROCm ecosystem remains far behind CUDA.
Intel (INTC)	~$90–100bn	Gaudi 3 accelerator; Falcon Shores roadmap	Has lost significant ground; Gaudi 3 targets efficiency-tier AI workloads rather than frontier training. Foundry troubles add distraction. Not a near-term threat at the frontier.
Broadcom (AVGO)	~$1.0–1.1T	Custom TPU ASICs for Google (TPU6), Apple (M-series AI), Meta MTIA	Different competitive model — custom ASIC for hyperscalers. Directly competes for wallet at Google and Meta specifically. Could cannibalise NVIDIA revenue at those accounts over multi-year periods.
Marvell (MRVL)	~$65–75bn	Custom AI ASIC silicon; co-packaged optics	Similar custom-ASIC niche to Broadcom; Microsoft and Amazon are notable customers. Growing but smaller threat at this stage.
Qualcomm (QCOM)	~$165–180bn	Cloud AI inference chips (on-device AI, edge)	Competes at the edge/device inference tier, not frontier training. A different market segment.

A key structural risk: Microsoft (Azure Maia), Google (TPU), Amazon (Trainium/Inferentia), and Meta (MTIA) are all deploying custom silicon alongside or as a partial replacement for NVIDIA purchases. The direction is clear but the pace of displacement remains slower than bears have modelled — CUDA lock-in and Blackwell/Rubin performance leads keep NVIDIA in the critical path for new frontier model training.

9. Leadership and Ownership

CEO: Jensen Huang. Co-founder, President and CEO since 1993. One of the longest-tenured tech CEOs; deeply technical and widely credited with setting the CUDA strategy in 2006 that created today's AI moat. No succession plans disclosed. Key-person risk is considered material.

Key executives: Colette Kress (CFO, since 2013); Ajay Puri (EVP, Worldwide Field Operations); Debora Shoquist (EVP, Operations); Tim Teter (EVP, General Counsel).

Insider ownership: Jensen Huang holds approximately 3.5% of shares outstanding (post-split, approximately 860 million shares), worth roughly $190bn at current prices. He is the largest individual insider holder by a wide margin.

Major institutional holders: Vanguard Group (~8%), BlackRock (~7%), State Street (~4%), Fidelity (~3%). Extensive institutional ownership typical of a mega-cap index constituent.

Recent insider transactions (Form 4, SEC filings):

Name	Date	Type	Shares	Price	Value	Plan type
Jensen Huang (CEO)	Various, Q1 2026	Sell	Multiple tranches	~$130–160	~$500m+	10b5-1 pre-planned
Colette Kress (CFO)	Various, Q1 2026	Sell (option exercise + sell)	Multiple tranches	~$130–160	~$20m+	10b5-1 pre-planned

Note: Insider selling at NVIDIA is conducted under pre-planned 10b5-1 programmes. Given Jensen Huang's ~$190bn stake, planned portfolio diversification sales are expected and are not discretionary signals of negative outlook. For precise Form 4 transaction data, see SEC EDGAR.

10. Risks and Challenges

• Export controls (Regulatory): The April 2025 H20 export licence requirement cost $4.5bn in Q1 FY2026 and approximately $8bn in annual run-rate revenue. The U.S. government could expand controls to cover other NVIDIA products, further excluding China revenue. China represented roughly 13% of FY2025 revenue before restrictions tightened.
• Hyperscaler in-house silicon (Competitive): Microsoft, Google, Meta, and Amazon are all investing heavily in custom AI accelerators. While these displace NVIDIA at the margin today, the trajectory is toward greater self-sufficiency over a 5–10 year horizon, particularly for inference workloads.
• TSMC concentration (Operational): 100% of leading-edge NVIDIA chips are manufactured by TSMC in Taiwan. A Taiwan Strait geopolitical event, major earthquake, or TSMC operational failure would halt production entirely. No alternative foundry can yet produce NVIDIA's advanced chips at scale.
• Valuation multiple compression (Financial): At ~44x trailing GAAP P/E and ~25x P/S, any deceleration in growth — whether from market saturation, a competitor breakthrough, or macro slowdown — could compress multiples sharply. The stock has previously experienced 60%+ drawdowns.
• Key-person risk (Operational): Jensen Huang is the company's strategic architect and public face. NVIDIA has not disclosed a succession plan. His absence would be materially disruptive.
• Demand concentration (Concentration): A small number of hyperscalers represent a disproportionate share of Data Centre revenue. If one or two major customers simultaneously reduce GPU procurement (due to AI-budget reviews, custom silicon deployment, or macro pressures), NVIDIA's quarterly results would be materially impacted.
• CoWoS packaging bottleneck (Supply-chain): Blackwell production has been constrained at various points by TSMC's CoWoS-S advanced packaging capacity. This is being addressed but remains a risk for Rubin ramp.
• Antitrust scrutiny (Regulatory): NVIDIA's dominant market position in AI accelerators has attracted regulatory attention in the EU and U.S. The company's acquisition of Mellanox and attempted ARM acquisition (abandoned 2022) have drawn scrutiny. Future M&A may face heightened review.

11. Recent Developments

Today (12 May 2026): NVIDIA stock hit an intraday all-time high of $223.75, a gain of approximately 2%, on reports that trial production of the Vera Rubin next-generation AI chip platform is on track for June 2026. Market capitalisation briefly surpassed $5.4 trillion, making NVIDIA the world's most valuable company.

8 May 2026: NVIDIA and Corning Incorporated announced a multi-year strategic partnership to expand U.S. manufacturing of optical connectivity components for AI infrastructure. Separately, Suzanne Nora Johnson (former Goldman Sachs Vice Chairman) was appointed to NVIDIA's board of directors, effective 13 July 2026.

Q4 FY2026 / February 2026: NVIDIA reported Q4 FY2026 revenue of $68.1 billion (+73% YoY) and full-year FY2026 revenue of $215.9 billion. Net income for the year was $120.1 billion. Management guided Q1 FY2027 (Feb–Apr 2026) revenue to approximately $45 billion ±2%. This guidance reflected approximately $8 billion in lost H20 revenue due to export controls, with the remainder offset by Blackwell demand from U.S. and non-China customers.

March 2026 (GTC 2026): Jensen Huang delivered the GTC 2026 keynote and disclosed combined Blackwell and Rubin order visibility of approximately $1 trillion through calendar 2027. He also announced the Rubin Ultra (2H 2027) and Feynman (2028+) roadmap, confirming a two-year GPU upgrade cadence at scale.

January 2026 (CES 2026): Jensen Huang confirmed Vera Rubin NVL72 entered full production, promising up to 5x training and 10x lower inference token costs versus Blackwell. The Rubin platform features the NVL72 rack (72 Rubin GPUs, 36 Vera CPUs) and HGX Rubin NVL8 for enterprise.

Q1 FY2026 (reported May 2025): Revenue was $44.1 billion (+69% YoY) but included a $4.5 billion charge for H20 excess inventory and purchase obligations following the U.S. government export licence mandate for China-bound H20 products. Excluding this charge, non-GAAP EPS would have been $0.96 vs reported $0.81.

12. Key Dates Coming Up

• 28 May 2026 — Q1 FY2027 earnings release (Feb–Apr 2026 quarter). Revenue consensus approximately $78.6 billion (+78% YoY). The first major read on Blackwell sustainability and Vera Rubin pre-order momentum.
• July 2026 — Vera Rubin first shipments expected to hyperscalers (Microsoft, Google, Amazon, Meta, Oracle) per management statements at CES 2026 and GTC 2026.
• 13 July 2026 — Suzanne Nora Johnson joins NVIDIA board of directors.
• 2H 2026 — Vera Rubin NVL72 broader availability; management stated high 2H FY2027 revenue weighting.
• ~August 2026 — Q2 FY2027 earnings (estimated; covers May–Jul 2026 quarter, will include first Vera Rubin shipment revenue).

---

Disclaimer: This research is produced by ChartsView for educational and informational purposes only. It does not constitute financial advice or a recommendation to buy or sell any security. All information is sourced from publicly available company filings, press releases, and official data. ChartsView does not use analyst opinions or third-party ratings. Always conduct your own due diligence and consider your personal financial situation before making investment decisions. Past performance is not indicative of future results.

Related: Live Charts | Economic Calendar | Forum | Blog