Accelerated computing platforms for AI training and inference. Data Center is 89.7% of FY26 revenue at $193.7B; Gaming, Pro Visualization, Automotive and OEM together come to $22.2B.
Trailing earnings doubled in twelve months while the shares gained 22%, so the multiple has already halved to 33× — the 4th percentile of NVIDIA's own five-year range, and 24× a FY27 consensus still modelled at +82% growth. We are bullish on that arithmetic and below the Street on the reason: inventory is up 112% and receivables 67% against 65% revenue growth. That is what a ramp financed by the balance sheet looks like, and it is the line that breaks first.
NVIDIA sells accelerated computing systems — the processors, the networking that lashes thousands of them into one machine, and the CUDA software layer that makes the result expensive to leave. It manufactures none of the silicon. TSMC does that, which is why capital spending is 2.6% of revenue and $253.5B of trailing sales converts into $119.1B of free cash.
The concentration is the whole story. Data Center was 39.4% of revenue in FY22 and is 89.7% today, and a handful of hyperscalers fund most of it. That is either the best demand signal in the market or the narrowest one, and nothing in the reported financials settles which. The customers answer it, on the dates listed on page 7.
In four years this stopped being a graphics company. Gaming grew from $12.5B to $16.0B — a real business, compounding at 6.5% a year — and went from 46% of revenue to 7.4%, because Data Center grew eighteen-fold underneath it. Everything on the next six pages is a judgment about one dark band.
These are ship-to locations, not end markets. Singapore was 18.1% of FY25 revenue and vanishes from FY26 entirely as invoicing was re-cut; Taiwan absorbs most of it. China fell from 13.1% of revenue to 9.1% while still growing in dollars.
Revenue is eight times what it was four years ago and 60 cents of every dollar reaches the operating line. Performance is not the open question. Whether the cash follows the earnings is — and for five consecutive years it has not.
| FY22 | FY23 | FY24 | FY25 | FY26 | What it says | |
|---|---|---|---|---|---|---|
| Revenue | 26.9 | 27.0 | 60.9 | 130.5 | 215.9 | Eight-fold in four years, off a real base |
| Revenue growth | +61.4% | +0.2% | +125.9% | +114.2% | +65.5% | FY23 was a crypto hangover, not a ceiling |
| Gross profit | 17.5 | 15.4 | 44.3 | 97.9 | 153.5 | Added $55.6B of gross profit in FY26 alone |
| Gross margin | 64.9% | 56.9% | 72.7% | 75.0% | 71.1% | FY26 dip is one quarter's $4.5B H20 write-off |
| Operating income | 10.0 | 4.2 | 33.0 | 81.5 | 130.4 | Larger than the revenue of all but a few peers |
| Operating margin | 37.3% | 15.7% | 54.1% | 62.4% | 60.4% | Peaked in FY25; scale has stopped adding margin |
| Net income | 9.8 | 4.4 | 29.8 | 72.9 | 120.1 | More than the next four AI-semi peers combined |
| Diluted EPS | $0.38 | $0.17 | $1.19 | $2.94 | $4.90 | Split-adjusted; buybacks add about a point a year |
| Free cash flow | 8.1 | 3.8 | 27.0 | 60.9 | 96.7 | Below net income in every one of the five |
Eight beats from eight, in a band between +3.2% and +9.9% and averaging +5.9%. A streak that regular is a guidance policy, not a forecasting edge: the printed quarter is almost never the risk. The guide that follows it is.
Against eight other names in the accelerator supply chain, NVIDIA ranks first or second on every operating measure and sits mid-pack on price. It trades 28% above the peer median on EV/Sales and 51% below it on earnings. Both statements are true. The distance between them is the operating margin.
| Company | Market cap | Rev growth | Gross margin | Op margin | ROIC | EV / Sales | P / E |
|---|---|---|---|---|---|---|---|
| NVIDIA NVDA | $5.29T | +65.5% | 74.1% | 64.0% | 63.0% | 20.8× | 33.4× |
| TSMC TSM | $2.19T | +33.0% | 64.2% | 56.0% | 27.1% | 13.3× | 27.4× |
| Broadcom AVGO | $2.01T | +23.9% | 67.0% | 43.7% | 19.5% | 27.3× | 68.4× |
| Micron MU | $1.02T | +48.9% | 72.6% | 65.8% | 44.0% | 11.1× | 20.2× |
| AMD AMD | $799B | +34.3% | 53.2% | 15.7% | 7.6% | 19.3× | 124.5× |
| Intel INTC | $510B | −0.5% | 38.9% | 0.1% | 0.0% | 9.6× | n/m |
| Arm Holdings ARM | $307B | +22.8% | 95.3% | 17.3% | 7.3% | 59.1× | 296.7× |
| Marvell MRVL | $189B | +42.1% | 50.6% | 16.2% | 5.0% | 21.8× | 73.3× |
| Qualcomm QCOM | $168B | +13.7% | 54.2% | 23.2% | 17.4% | 4.1× | 18.3× |
| Peer median excl. NVDA | $655B | +28.4% | 59.2% | 20.3% | 12.5% | 16.3× | 68.4× |
The two cheapness rows disagree because the denominators do: 64 cents of every sales dollar becomes operating profit.
The regression through this set is almost flat. Arm grows 23% and trades at 59× sales; Micron grows 49% and trades at 11×. Growth explains close to nothing about price here. NVIDIA sits a fraction below a line that barely slopes — the market is not charging it a premium for being the fastest-growing name in the group.
Three questions in order: cheap against its own history, cheap against its industry, and what would have to be true for today's price to be right. The last matters most — it puts the market in the position of making a claim, and lets us judge that instead of ours.
NVIDIA has never been this cheap on earnings. The P/E and EV/EBITDA both sit at the 4th percentile of five years of daily observations — not because the price fell, but because earnings tripled underneath it. The shares are cheaper than the AI-semi peer median and than the NASDAQ technology sector, which has not been true at any point in this cycle.
Accept consensus through FY29 and the price needs only 12% a year afterwards, slower than cloud has compounded for a decade. Reject it and the same price demands 21% for ten straight years. No valuation argument here is not really an argument about capex.
The mean target of $319.48 puts 35.5× on FY27 earnings, a re-rating from today's 24.2× forward. Our $283 asks for less — 22.2× on the FY28 number the market will actually be discounting a year from now.
Every assumption behind the three targets on page one is on this page. Change one and the number moves — that is the point. A target you cannot take apart is a guess with a decimal place.
| Driver | Bear | Base | Bull | Why it moves |
|---|---|---|---|---|
| Revenue CAGR, FY26–28 | +39.5% | +56.7% | +63.9% | Hyperscaler capex, nothing else |
| FY28E revenue | $420B | $530B | $580B | Consensus is $563B |
| FY28E operating margin | 55.0% | 63.0% | 65.0% | Rack mix and HBM cost |
| Diluted shares, B | 24.2 | 23.9 | 23.6 | Buybacks net of stock compensation |
| FY28E EPS | $8.35 | $12.30 | $14.10 | Consensus is $12.75 |
| Exit P / E on FY28E | 18.0× | 23.0× | 28.0× | Trailing multiple is 33.4× today |
| 12-month target | $150 | $283 | $395 | −31.3% / +29.7% / +81.0% |
| FY28E EPS | 17× | 20× | 23× | 26× | 29× |
|---|---|---|---|---|---|
| $9.50 | $162 | $190 | $219 | $247 | $276 |
| $11.00 | $187 | $220 | $253 | $286 | $319 |
| $12.30 | $209 | $246 | $283 | $320 | $357 |
| $13.60 | $231 | $272 | $313 | $354 | $394 |
| $15.00 | $255 | $300 | $345 | $390 | $435 |
The weighted value lands 1.9% below the headline, so the base case is not being carried by the bull tail. Note the shape: the bear costs 31% and the bull pays 81%. That asymmetry, not the point estimate, is the reason for a bullish rating.
Eight events over the next eighteen months, each scored for how much of the thesis it settles. The two heaviest dates are a January capex season NVIDIA does not control, and the February guide that prices off it.
Four companies fund most of Data Center. What they budget for 2027 settles more of this thesis than anything NVIDIA says.
| Aggregate 2027 capex | Reads as | Target |
|---|---|---|
| Up more than 25% | Second wave funded; FY28 consensus is too low | $395 |
| Up 20 – 25% | Digestion without a break, base case intact | $283 |
| Up less than 20% | The bear revenue path becomes the base | $150 |
Six risks ranked by what they would cost rather than how likely they are, each with the observable that tells you it is happening. A risk you cannot monitor is not a risk, it is an anxiety.
| Insider | Role | Filings | Value |
|---|---|---|---|
| Jensen Huang | CEO | 276 | $806M |
| Mark Stevens | Director | 14 | $700M |
| Ajay Puri | EVP Field Ops | 13 | $217M |
| Colette Kress | CFO | 140 | $111M |
| All insiders | 15 people | 513 | $2.10B |
Read this as texture, not signal. The disposals run through pre-arranged plans and $2.10B is four basis points of a $5.3T company. The one fact worth keeping is the zero: across twelve months and fifteen insiders, nobody bought a share on the open market.
Nothing here contradicts the bullish case. Nothing here supports it either.
Each case sets revenue, margin and multiple assumptions independently, converts operating income to EPS at an 88% net-to-operating ratio, then values FY28 estimates at a twelve-month horizon. Probabilities are subjective; the headline is the base case.
Prices, consensus estimates, analyst targets and Form 4 filings from Financial Modeling Prep; fundamentals from NVIDIA's SEC filings. Peer medians are calculated across the nine-name set on page 4. Data as of 6 August 2026.
Generated automatically from structured data. It cannot interview management, verify disclosure quality, or price an event it was never told about. Institutional ownership was unavailable and is therefore not shown rather than estimated.
Every figure above comes from NVIDIA's own SEC filings and live market data, assembled and written up automatically. Ticker Alpha is building this for any listed company, on demand — so the research you need on a Tuesday afternoon takes a minute, not a week.