Needham, DA Davidson, JP Morgan, Morgan Stanley, Tigress Financial, Cantor Fitzgerald, Jefferies, Wedbush, Evercore — all Buy/Overweight/Strong Buy maintained Jun 2026
NVIDIA Corporation designs and sells accelerated computing platforms spanning graphics processing units (GPUs), networking, and software. The company operates two segments: Compute & Networking (data-center AI accelerators — Hopper H100/H200, Blackwell B100/B200/GB200 — plus Mellanox InfiniBand/Ethernet interconnect, DRIVE automotive AI, Jetson robotics, and NVIDIA AI Enterprise software) and Graphics (GeForce gaming GPUs, RTX professional visualization, and GeForce NOW cloud gaming).
NVIDIA's fiscal year ends in late January. FY2026 (ended Jan 26, 2026) was a defining year: revenue surged to $215.9B (+65.6% YoY) as hyperscalers accelerated Blackwell GPU deployments for generative AI training and inference. Data-center GPU demand continues to outpace supply, and NVIDIA's networking and software attach rates are growing alongside chip shipments.
The company executed a 10-for-1 stock split in June 2024 (all per-share figures herein are post-split). Jensen Huang has led the company since co-founding it in 1993 and retains approximately 3.5% economic ownership, aligning long-term interests with shareholders.
Data-center AI dominance. NVIDIA captures an estimated 70–80% of AI GPU market share. Hyperscalers (Microsoft Azure, Google Cloud, Amazon AWS, Meta) and sovereign AI programs are committing multi-year, multi-billion-dollar orders — Jensen Huang has cited a $3–4T data-center capex opportunity by 2030.
Blackwell ramp & Vera Rubin roadmap. The Blackwell architecture (B100/B200/GB200 NVL72 rack-scale systems) is in full production and sold out well into FY2027. The next-generation Vera Rubin platform is slated for late FY2027/FY2028, sustaining the annual cadence of GPU generations that competitors cannot match.
CUDA software moat. The CUDA parallel-computing ecosystem, built over 17+ years with 4M+ developers and 1,000+ GPU-optimized libraries, creates durable switching costs that go far beyond silicon. cuDNN, TensorRT, and NCCL integration in PyTorch/JAX/TensorFlow means customers optimize workloads on NVIDIA hardware — making competitive substitution far more costly than chip-vs-chip comparisons suggest.
Networking monetization. Mellanox InfiniBand and Spectrum-X Ethernet attach rates within AI clusters are growing, adding a recurring high-ASP revenue layer on top of GPU sales. NVLink switching for rack-scale NVL72 clusters further deepens per-system economics and raises the switching cost for the entire cluster.
| Metric | FY2022 | FY2023 | FY2024 | FY2025 | FY2026 |
|---|---|---|---|---|---|
| Revenue | $26,914 | $26,974 | $60,922 | $130,497 | $215,938 |
| Revenue Growth YoY | — | +0.2% | +125.8% | +114.2% | +65.5% |
| Gross Profit | $17,475 | $15,356 | $44,301 | $97,858 | $153,463 |
| Gross Margin | 64.9% | 56.9% | 72.7% | 75.0% | 71.1% |
| Operating Income | $10,041 | $4,224 | $32,972 | $81,453 | $130,387 |
| Operating Margin | 37.3% | 15.7% | 54.1% | 62.4% | 60.4% |
| EBITDA | $11,351 | $5,986 | $35,583 | $86,137 | $144,552 |
| EBITDA Margin | 42.2% | 22.2% | 58.4% | 66.0% | 66.9% |
| Net Income | $9,752 | $4,368 | $29,760 | $72,880 | $120,067 |
| Net Margin | 36.2% | 16.2% | 48.8% | 55.8% | 55.6% |
| EPS (Diluted, post-split) | $0.38 | $0.17 | $1.19 | $2.94 | $4.90 |
| R&D Expense | $5,268 | $7,339 | $8,675 | $12,914 | $18,497 |
| Interest Expense | $236 | $262 | $257 | $247 | $259 |
| Metric | FY2022 | FY2023 | FY2024 | FY2025 | FY2026 |
|---|---|---|---|---|---|
| Cash & ST Investments | $21,208 | $13,296 | $25,984 | $43,210 | $62,556 |
| Total Current Assets | $28,829 | $23,073 | $44,345 | $80,126 | $125,605 |
| Total Assets | $44,187 | $41,182 | $65,728 | $111,601 | $206,803 |
| Total Debt | $11,831 | $12,031 | $11,056 | $10,270 | $11,412 |
| Net Debt / (Net Cash) | $9,841 | $8,642 | $3,776 | $1,681 | $(807) |
| Total Liabilities | $17,575 | $19,081 | $22,750 | $32,274 | $49,510 |
| Total Stockholders' Equity | $26,612 | $22,101 | $42,978 | $79,327 | $157,293 |
| Debt / Equity | 0.44x | 0.54x | 0.26x | 0.13x | 0.07x |
| Current Ratio | 6.65x | 3.52x | 4.17x | 4.44x | 3.91x |
| Goodwill & Intangibles | $6,688 | $6,048 | $5,542 | $5,995 | $24,138 |
| Metric | FY2022 | FY2023 | FY2024 | FY2025 | FY2026 |
|---|---|---|---|---|---|
| Operating Cash Flow | $9,108 | $5,641 | $28,090 | $64,089 | $102,718 |
| Capital Expenditures | $(976) | $(1,833) | $(1,069) | $(3,236) | $(6,042) |
| Free Cash Flow | $8,132 | $3,808 | $27,021 | $60,853 | $96,676 |
| FCF Margin | 30.2% | 14.1% | 44.4% | 46.6% | 44.8% |
| Stock Buybacks | $0 | $(10,039) | $(9,533) | $(33,706) | $(40,086) |
| Dividends Paid | $(399) | $(398) | $(395) | $(834) | $(974) |
| Net Investing CF | $(9,830) | $7,375 | $(10,566) | $(20,421) | $(52,228) |
| Stock-Based Comp | $2,004 | $2,709 | $3,549 | $4,737 | $6,386 |
| D&A | $1,174 | $1,544 | $1,508 | $1,864 | $2,843 |
NVIDIA is effectively net-cash as of FY2026 (net debt $(807M)), having generated $96.7B in FCF during the year while returning $40.1B in buybacks and $974M in dividends. Long-term debt of $7.5B is dwarfed by $62.6B in cash and ST investments. Capital structure is pristine; the balance sheet is not a constraint.
NVDA's margins expanded dramatically from FY2023 trough (gross 56.9%, net 16.2%) to FY2025 peak (gross 75.0%, net 55.8%) as data-center GPUs achieved scale. FY2026 gross margin dipped to 71.1% reflecting Blackwell CoWoS packaging ramp costs and supply chain investments — management guided this to normalize toward 73–74% as Blackwell matures. Net margin of 55.6% is extraordinary for a hardware company of this scale.
| Multiple | FY2022 | FY2023 | FY2024 | FY2025 | FY2026 |
|---|---|---|---|---|---|
| P/E (trailing) | 62.7x | 109.1x | 51.8x | 39.9x | 43.0x |
| EV/EBITDA | 54.7x | 81.0x | 43.5x | 33.8x | 35.0x |
| P/S (Price/Sales) | 22.7x | 17.7x | 25.3x | 22.3x | 23.6x |
| P/B (Price/Book) | 22.9x | 21.6x | 35.9x | 36.7x | 32.9x |
| P/FCF | 75.2x | 125.1x | 57.1x | 47.8x | 52.8x |
| EV/Sales | 23.1x | 18.0x | 25.4x | 22.3x | 23.4x |
All multiples recomputed at current price $210.69 using FY2026 actuals (ended Jan 26, 2026; filed Feb 25, 2026). EV computed as: market cap $5.103T + total debt $11.4B − cash & ST investments $62.6B = ~$5.051T. EV/EBITDA = $5.051T / $144.6B = 35.0x. Historical multiples at $210.69 are presented for trend reference, not as actual historical market prices.
| Metric | FY2022 | FY2023 | FY2024 | FY2025 | FY2026 |
|---|---|---|---|---|---|
| Return on Equity (ROE) | 36.6% | 19.8% | 69.2% | 91.9% | 76.3% |
| Return on Assets (ROA) | 22.1% | 10.6% | 45.3% | 65.3% | 58.1% |
| Return on Invested Capital (ROIC) | 24.6% | 11.7% | 51.3% | 75.3% | 62.9% |
| Asset Turnover | 0.61x | 0.65x | 0.93x | 1.17x | 1.04x |
| Inventory Turnover | 3.6x | 2.3x | 3.1x | 3.2x | 2.9x |
| Days Sales Outstanding | 63.1 | 51.8 | 59.9 | 64.5 | 65.0 |
| FCF Conversion (FCF/Net Income) | 83.4% | 87.2% | 90.8% | 83.5% | 80.5% |
| Interest Coverage (EBIT/Int Exp) | 42.5x | 16.1x | 128.3x | 329.8x | 503.4x |
| R&D as % of Revenue | 19.6% | 27.2% | 14.2% | 9.9% | 8.6% |
| Metric | FY2026A | FY2027E | FY2028E | FY2029E |
|---|---|---|---|---|
| Revenue (avg) | $215.9B | ~$330B | $556.0B | $672.9B |
| Revenue (low / high) | — | — | $387.9B / $638.2B | $672.0B / $673.8B |
| EBITDA (avg) | $144.6B | ~$205B | $284.3B | $344.1B |
| Net Income (avg) | $120.1B | ~$196B | $304.8B | $359.2B |
| EPS (diluted, avg) | $4.90A | ~$8.00E | $12.50E | $14.69E |
| Fwd P/E (at $210.69) | 43.0x | ~26.3x | ~16.9x | ~14.3x |
| EPS Growth YoY | +66.7% | ~+63% | ~+56% | ~+18% |
| PEG Ratio (Fwd P/E / EPS Gth) | — | 0.41x | 0.30x | 0.78x |
FY2027E revenue/EPS are street consensus approximations (FMP long-range consensus begins at FY2028 ending Jan 2028, with 41 analysts for revenue and 30 for EPS). FY2028/FY2029 from FMP consensus data. PEG = Fwd P/E ÷ forward EPS growth %. All at current price $210.69.
NVDA completed a 10-for-1 stock split in June 2024. All share counts above are post-split equivalent. Weighted average diluted shares declined from ~25.35B (FY2022) to ~24.51B (FY2026) despite SBC issuance, reflecting aggressive buyback execution ($40.1B in FY2026, $33.7B in FY2025). SBC remains elevated at ~3.0% of revenue ($6.4B in FY2026), but net share reduction demonstrates capital returns discipline.
| Filing Date | Name | Title | Type | Shares | Avg Price | Value |
|---|---|---|---|---|---|---|
| 2026-06-18 | Jensen Huang | President & CEO | Gift | 400,000 | $0.00 | Philanthropic |
| 2026-06-10 | Scott Gawel | Principal Acctg. Officer | Award | 59,509 | $0.00 | RSU Award |
| 2026-06-05 | Stephen C. Neal | Director | Sale | 15,500 | $215.73 | $3.34M |
| 2026-06-04 | Mark A. Stevens | Director | Sale | 1,000,000 | $220.14 | $220.1M |
| 2026-06-04 | Mark A. Stevens | Director | Gift | 307,500 | $0.00 | Philanthropic |
| 2026-05-29 | John Dabiri | Director | Sale | 625 | $214.00 | $133.8K |
Recent insider activity is dominated by planned dispositions and philanthropic gifts — no open-market purchases by any insider in the 60-day window. Director Mark Stevens sold approximately $220M through June 2–4 filings (likely pre-arranged 10b5-1 plan). Jensen Huang's 400K share gift is philanthropic. Scott Gawel received RSU awards as routine new-officer compensation. Overall pattern is consistent with estate/tax planning at a $5T market cap company, not a directional sell signal.
AI capex cycle sustains into FY2028+. Hyperscalers have committed $200B+ in 2025 capex with NVIDIA capturing a disproportionate share. If infrastructure spend follows the $3–4T by 2030 trajectory Jensen Huang cites, NVDA could sustain 30–50% annual revenue growth through FY2028, driving consensus-beating FY2027 EPS of $9–10 and FY2028 EPS of $14–16.
Blackwell sells out; Vera Rubin expands TAM. GB200 NVL72 rack systems command $2–3M ASPs — 3–5x prior H100 servers — and carry higher gross margins as CoWoS ramp costs normalize. Vera Rubin (late FY2027/FY2028) extends the cadence and maintains GPU leadership over AMD MI350/MI400 and Google TPU v6.
Software/NIM stack becomes recurring revenue. NVIDIA AI Enterprise, NIM microservices, and DGX Cloud usage-based revenue add an annuity-like layer reducing cyclical sensitivity and supporting higher sustained multiples.
Sovereign AI & enterprise inference expand TAM. Government-sponsored AI clusters (Saudi Arabia, UAE, Japan, France) and enterprise inference deployments represent demand vectors entirely separate from hyperscaler training — and with less price sensitivity.
China export controls tighten further. BIS restrictions have already barred H100/A100/H20 sales into China. Additional controls on successor chips or HBM content could eliminate a revenue segment that represented 15–20% of data-center sales pre-restrictions. Geopolitical escalation creates binary, non-hedgeable event risk.
Hyperscaler capex digestion / demand cliff. Microsoft, Google, Amazon, and Meta collectively guided $300B+ in 2025 capex. If AI model efficiency improves faster than expected (DeepSeek-style inference breakthroughs) or enterprise AI ROI disappoints, capex cycles can reverse — as seen with crypto GPU demand in 2022.
Custom silicon displacement (ASICs). Google TPU v6, Amazon Trainium 2, Microsoft Maia, Meta MTIA, and Broadcom/Marvell ASIC programs could take 10–20% of inference capacity by 2027. The largest customers have the deepest engineering resources and the strongest ROI incentive to reduce GPU dependency.
Multiple compression at $5T. At 43x trailing P/E and 35x EV/EBITDA, even modest growth disappointments compress multiples sharply. A re-rating from 43x to 30x P/E with flat EPS implies ~30% downside. Beta 2.20 amplifies any market-wide drawdown. The 2022 analog (−66% peak-to-trough) is a live precedent for NVDA multiple reset.
BIS A100/H100/H20 restrictions eliminated China as a meaningful data-center customer. Further tightening covering HBM content, advanced packaging, or successor chips would materially reduce addressable revenue with no near-term offset. Geopolitical escalation is binary and non-hedgeable.
Microsoft, Google, Amazon, and Meta represent an estimated 40–50% of data-center GPU revenue. Any single hyperscaler reducing orders — via capex cuts, in-house ASIC substitution, or cycle digestion — creates immediate revenue risk. NVDA visibility beyond 2–3 quarter order books is limited.
AMD MI350/MI400 is competitive on inference efficiency; ROCm is improving. Google TPU v6, Amazon Trainium 2, and Broadcom/Marvell ASICs target highest-volume inference at hyperscalers. CUDA moat erodes faster if PyTorch/JAX backends gain broader ASIC portability and hyperscalers fund the software integration.
At 43x trailing and 26x forward P/E, NVDA prices in sustained >50% EPS growth. Any macro shock, rate re-pricing, or earnings miss compresses the multiple and amplifies through beta=2.20. NVDA fell 66% peak-to-trough in 2022 despite no fundamental impairment — the re-rating risk is not theoretical.
NVIDIA depends 100% on TSMC for N4P/N3B node manufacturing and SK Hynix/Micron for HBM3e. Geopolitical Taiwan risk, CoWoS capacity constraints, or HBM supply disruptions (yield failure, fire, earthquake) could limit shipments during peak demand — as H100 supply constraints demonstrated in 2023–2024.
Enterprise AI ROI remains unproven at scale. If inference cost-per-query fails to decline fast enough for profitable AI products, hyperscalers may pause or defer capex in 2026–2027 — analogous to the 2001 telecom/fiber cycle. DeepSeek-efficiency gains reducing GPU count per workload accelerate this risk scenario.
| Scenario | FY2027E EPS | Target Fwd P/E | EV/Fwd EBITDA | PEG Check | Implied Price | vs. $210.69 |
|---|---|---|---|---|---|---|
| Bull | $9.50 | 35x | ~44x | 0.36x (97% EPS gth) | $332 | +57.6% |
| Base | $8.00 | 28x | ~34x | 0.43x (65% EPS gth) | $224 | +6.3% |
| Bear | $6.00 | 20x | ~24x | 0.91x (22% EPS gth) | $120 | −43.0% |
Bull case ($332, +58%): Blackwell shipments materially outperform through H1 FY2027, Vera Rubin sees early customer commitments, and China restrictions do not tighten. FY2027 EPS hits $9.50+ as data-center revenue approaches $280B. The market re-rates to 35x forward P/E — in line with the stock's FY2026 exit multiple. PEG of 0.36 remains well below 1.0, validating the growth premium. Price target ~$332, consistent with the current 1-quarter analyst consensus of $322.
Base case ($224, +6%): Consensus FY2027E EPS of ~$8.00 with modest multiple compression to 28x as the stock consolidates its 2025 run-up and the market awaits Vera Rubin validation. EV/Fwd-EBITDA compresses to ~34x. PEG remains below 0.5 on FY2028 estimates, suggesting continued undervaluation on a growth-adjusted basis. A 28x fwd P/E on $8.00 implies ~$224. This is the grind-and-wait scenario where NVDA keeps executing but the stock digests its premium until the next earnings catalyst.
Bear case ($120, -43%): China export escalation plus hyperscaler capex pause (DeepSeek-efficiency thesis accelerates) plus macro multiple compression. FY2027 EPS misses at $6.00 and the market re-rates to 20x — still a premium to S&P but reflecting China-impaired, slowing growth. EV/Fwd-EBITDA compresses to ~24x. The 2022 analog (66% peak-to-trough) confirms asymmetric downside is real at 43x trailing; beta 2.20 amplifies any broader market selloff. Bear case is a tail scenario, not a base — but the magnitude demands position sizing discipline.
Scenarios framed on forward P/E × FY2027E consensus EPS, cross-checked against EV/Fwd-EBITDA (EV assumes net-cash balance sheet stays roughly constant). PEG = Fwd P/E ÷ forward EPS growth YoY. NVDA's secular growth and fabless model make P/B inappropriate as primary lens; mid-cycle EV/EBITDA is directional but secondary to Fwd P/E + PEG for a high-growth secular compounder. All prices at current $210.69. Not investment advice.
This report was generated using FMP financial data as of 2026-06-20. This is for informational purposes only and does not constitute investment advice. Past performance is not indicative of future results.