Revenue doubled to $96.2B in the July quarter, gross margin holds at 75%, and the stock just printed an all-time high — yet it trades at only ~14× next fiscal year’s consensus EPS, because the market is pricing one question: is the AI build-out a self-funding utility, or a financing loop Nvidia helps pay for? Six analyst lenses, three scenarios, four time horizons.
Gray line = NVDA’s actual price, built only from dated, sourced prints between Dec 31 2024 and Oct 9 2026 (a 22-month window; months between anchors are not interpolated, and the history side uses its own month scale while the forward side uses year marks). The Dec 31 2024 point (≈ $134.29) is derived from a reported +40.45% year-to-date move to $188.61 on Dec 24 2025; the Apr 8 ’25 and most other anchors are closes, while the Oct 29 ’25 peak, Mar 30 ’26 low and Oct 6 ’26 high are intraday. Colored paths = synthesized scenario midpoints (EPS × exit multiple), probability-weighted bear 30% · base 45% · bull 25% — bear-tilted because the shares sit within 6% of an all-time high, beta is ~2.2, roughly $50B of Nvidia’s own capital is committed to customers, and one Street out-year series shows FY30 revenue falling. Linear price axis, mid-year marks. Open circle = Street 12-month mean target ≈ $329 (S&P Global, 62 analysts; range $180–$515).
Those probabilities are a judgment call — so make them yours. Drag to set how likely the bear and bull cases are (base takes the remainder); the blended target below, the dotted line on the chart, and the prob-weighted row of the scenario cards all update live.
The same fundamentals support wildly different conclusions depending on which framework you trust. Each lens below is a synthesized expert perspective with its own 12-month target; the six average about $266 against a Street mean near $329.
Revenue of $96.2B grew 106% (+18% sequentially), Q3 is guided to $108.0B with no China compute in it, and the CFO has pointed to roughly 70% growth in fiscal 2028. Supply commitments of $279B and management’s comment that supply covers only about 70% of demand say the order book, not the factory, is the constraint. Management framed revenue per gigawatt as stepping from ~$25B (Grace Blackwell) toward ~$40B (Vera Rubin), as reported from the call.
Trailing free cash flow of $126.9B is a thin 2.3% yield on a $5.5T cap, but the balance sheet is net cash even after a $25B bond, gross margin is 75%, and the $235B buyback capacity is ~4% of market value. The blemish is conversion: Q2 free cash flow was $21.3B against $54.0B of non-GAAP net income as receivables rose $22.3B.
Nvidia has committed close to $50B to frontier labs and carries $26B of cloud-compute service agreements (Seaport, Dec ’25), so some demand is demand it helped pay for. Custom-ASIC shipments are growing ~45% against ~16% for merchant GPUs, receivables are up from $38.5B to $63.1B in six months, and one Street out-year series already has fiscal 2030 revenue falling. Peak margins and a digestion year compress the multiple to a cyclical 11×.
The moat is the rack, not the die: NVLink, networking, CUDA and software sell as one system, and Nvidia holds roughly 60% of TSMC CoWoS packaging allocation. Bloomberg Intelligence expects a 70–75% accelerator share through 2030 — down from the mid-80s, but on a market growing far faster than any rival can capture. Custom silicon takes the margin at the edges, not the center.
A new all-time closing high of $239.24 on Oct 6 and a 20-week high say trend is intact; beta near 2.2 says it will not be smooth. The multiple has compressed even as price rose: ~18× forward in August (per Tom Lee) to ~14× FY28E today. After falling following each of its prior four prints, the stock rose 8.7% after Aug 26 — a regime change worth watching on Nov 17.
Morgan Stanley sees a 34% net US data-center power shortfall (32 GW) through 2028, which delays deployments for everyone, though it judges Nvidia and Broadcom largely insulated. Credit default swaps on Oracle and Broadcom hit records this week while Nvidia’s sit at 80–87bp. Export policy is case-by-case for H200-class chips and the guide assumes zero China.
What the sell-side expects over the next year. Bars are sorted low to high; the dashed line is today’s $229.28 — every target but the lone Sell sits well above it, which is the crowded-consensus risk in one picture.
Targets as last published or reiterated: Cantor Fitzgerald (Oct 1 ’26), Morgan Stanley (Oct 2), Piper Sandler (initiation, Sep 10), Rosenblatt (Sep 4), and Aug 27 ’26 post-earnings notes for Bernstein, JPMorgan, Citigroup, Mizuho and KeyBanc. S&P Global tally (62 analysts): 59 Buy / Strong Buy · 2 Hold · 1 Sell. Aggregators differ on the mean — $324 (TipRanks), $329 (S&P Global, Chartmill), $343–344 (Benzinga, Light Horse); the $329 figure is used throughout. *Seaport (Sell) $140 is the last figure sourced (Apr 5 ’26) while S&P Global now shows a Street-low of $180, so read that bar as the bear’s old anchor, not a live print. No Hold-rated firm target was individually sourced, so none is charted. Next earnings: Nov 17 ’26 (consensus ~$2.52–2.53 EPS on ~$111B revenue vs. the $108.0B guide).
Synthesized scenario midpoints (mid-year), each = forward fiscal-year EPS × an exit multiple. Returns are vs. today’s $229.28. These are illustrative frameworks, not forecasts — five-year outcomes hinge on whether the buildout is self-funding or financed.
Where the money actually goes. The bull and the bear theses both live in the gap between these four bars — and in a line that is not on the chart at all.
An asset-light engine: revenue is up about 5× in three years and free cash flow about 4.7×, while capex (property, equipment and intangibles) stays near 2% of sales. The trailing-twelve-month bars sum Q3 FY26 through Q2 FY27 from the company’s filings (revenue $303.0B, free cash flow $126.9B, capex $7.4B). Free cash flow uses Nvidia’s own definition (operating cash flow less capex and principal payments): FY26 $96.6B and FY25 $60.7B, where aggregators print $96.68B and $60.85B; FY24 $27.0B is operating cash flow less capex. Total debt jumped from $8.5B to $33.4B after a $24.9B issuance in Q2 — funded, not forced, since cash and debt securities are $56.6B. The line that is not here: strategic capital. Nvidia bought $42.4B of equity securities in the first half of FY27 (against $4.4B of capex) and paid $13.0B plus $2.9B for Groq — the bear’s “circular financing” question lives in that spend, not in capex.
Every scenario price is an EPS estimate times an exit multiple. Here is the earnings ladder the scenarios are built on — and what the shares cost on each rung.
From Q1 FY27 Nvidia’s non-GAAP measures include stock-based compensation, so history is restated: FY25 $2.92 and FY26 $4.64 (old basis: $2.99 and $4.77). Gray = reported; solid olive = Street consensus (FY27E ≈ $9.40, between Yahoo’s $9.31 and WallStreetZen’s $9.47; FY28E ≈ $15.90, between $15.70 and $16.09); outlined oat = the base-case extension for FY29E–FY32E, built from revenue × margin × share count rather than consensus because out-year Street series conflict (one shows FY29 EPS of $16.4, another a FY29 revenue of $929B that implies about $21). The base case’s $22.50 of FY32 EPS at ~16× ≈ the $360 five-year base target. Reported GAAP EPS ($7.91 TTM) swings on equity-stake revaluations, so every multiple below is struck on the clean figure.
| Valuation, Oct 9 ’26 | NVDA | Basis |
|---|---|---|
| P/E trailing — headline GAAP | 29.0× | $7.91 TTM GAAP EPS, lifted by ≈ $30.6B of pre-tax equity-security gains |
| P/E trailing — clean non-GAAP | 33.0× | $6.95 TTM = $1.27 + $1.59 + $1.87 + $2.22 (stock comp included) |
| P/E forward — FY27E | 24.4× | Consensus EPS ≈ $9.40 (fiscal year ends Jan 2027) |
| P/E forward — FY28E | 14.4× | Consensus EPS ≈ $15.90 (fiscal year ends Jan 2028) |
| EV / EBITDA (TTM) | ≈ 27× | EV ≈ $5.51T (cap less $56.6B cash and debt securities, plus $33.4B debt); EBITDA ≈ $202B |
| Price / sales (TTM) | 18.3× | $5.54T ÷ $303.0B |
| FCF yield | 2.3% | $126.9B ÷ $5.54T market cap |
| PEG | ≈ 0.8× | FY27E P/E 24.4× ÷ FY27–29E EPS CAGR ≈ 32% (base-case path) |
| Peers (Morningstar) | NVDA | AVGO | AMD |
|---|---|---|---|
| P/E (normalized) | 32.7 | 37.0 | 105.6 |
| Price / sales | 18.5 | 19.8 | 24.3 |
| Price / cash flow | 32.0 | 34.0 | 85.1 |
| Price / book | 24.2 | 17.3 | 14.8 |
Against its own history: trailing P/E was about 37× in early April (Motley Fool, Apr 5 ’26) and about 27.5× on a GAAP basis in late August (Tom Lee), with the forward multiple falling from ~18× then to ~14–15× now — the price rose, the multiple fell, because estimates rose faster. Peer figures are Morningstar’s Oct 9 ’26 quote-page values.
Q2 FY27, year-over-year — read these against a stock sitting within 6% of its all-time high at ~14× next year’s EPS.
Q2 FY27 (quarter ended Jul 26 ’26) vs. Q2 FY26, non-GAAP unless noted, from the Aug 26 ’26 release; free cash flow $21.3B vs $13.5B; the Q3 bar compares the $108.0B guide midpoint with $57.0B a year earlier. Every line is green and the shares are near their high — so the disconnect runs the other way from most reports: growth is intact, and it is the multiple (~14× FY28E) that says the market doubts how long it lasts. Free cash flow lagging revenue is the one visible crack: receivables absorbed $22.3B in the quarter.
The entire valuation argument compresses into one disagreement: are Nvidia’s customers earning their money back — or is the build-out being paid for, in part, by the company selling into it?
Where each risk sits over a 3–5 year horizon, not just how big it is. The hot corner — likely and high-impact — holds a single risk, a capex-digestion year; the financing risks that define the bear case sit one row down, in “possible.”
Hyperscalers and labs pause after a multi-year build; growth stalls for a year or more and the shares trade as a cyclical. One Street series already models FY30 revenue −12%.
TPUs, Trainium, MTIA and Broadcom/Marvell ASICs take inference and a rising slice of training; share slides toward the 70% low end and pricing power softens.
HBM and DRAM price spikes and allocation fights pull gross margin below the 74% guide.
Receivables and inventory keep outrunning revenue, holding free-cash-flow conversion well under net income.
If stakes in labs and the $500B financing platforms read as vendor financing, the quality of demand — and the multiple — is questioned.
A borrower buying Nvidia systems on debt is refused refinancing; orders slip or are cancelled.
A 32 GW US power shortfall delays site energization and pushes revenue right even if Nvidia’s own allocations hold.
Licenses stay case-by-case and capped; the guide already assumes zero, so the loss is upside, not base.
A blockade, quake or advanced-packaging outage at TSMC halts CoWoS output and the whole product line; low odds, immediate repricing.
Hover the dotted terms in the metrics, or scan the desk’s working definitions here.