NVDA — NVIDIA Corporation

NVIDIA Corporation (NVDA) — deep-value analysis, REFRESH

Date: September 4, 2026 · Reference price: … · Market cap: … · Exchange: NASDAQ Regime: REFRESH over the thesis from 08.03.2026 (verdict then: MONITOR, GBL score …)

New primary sources (not present at the prior analysis): 10-Q Q2 FY2027 (quarter ended 26.07.2026, filed 26.08.2026, accession 0001045810-26-000075) — downloaded manually, since edgar_10k_downloader.py limits 10-Qs to the fiscal year covered by the last 10-K and hadn’t caught the FY2027 filings; 10-Q Q1 FY2027 (26.04.2026, filed 20.05.2026); 8-K 26.08.2026 with the release and CFO commentary (Q2 FY2027 + Q3 outlook); 8-K 17.08.2026 (SB Energy / PORTS). Inherited sources: 10-K FY2026, 10-K FY2025, the FY2026 10-Qs. Support: deterministic data pack 04.09.2026, research brief 04.09.2026 (treated as a deposition — see corrections in the “Red flags” chapter), delta pack delta-NVDA-20260903.md, GBL tracker.

What I explicitly inherit from the 08.03.2026 report, because it hasn’t changed: the three-layer moat architecture (CUDA / networking / rack-scale system) and its durability hierarchy; the competitive map (AMD merchant, Broadcom on Ethernet, hyperscalers’ captive ASICs concentrated on inference); Jensen Huang’s capital-allocation history (Mellanox 2020, ARM abandoned 2022); the structure of scenarios 2 and 3 in the pre-mortem. What I fully re-derive, because the new filing moved the facts: all quarterly figures, the FCF bridge, the capital structure (… bond issuance), off-balance-sheet commitments (… of guarantees), the valuation and the Monte Carlo simulation — the old one had a units error (see “Triangulated valuation”).

Executive summary

The Q2 FY2027 quarter (ended 26.07.2026) confirmed, fact by fact, the operational part of the old thesis and broke the financial part. Revenue was …mn, … YoY and … QoQ (10-Q, Statements of Income); gross margin 75.0%, the highest in the last eight quarters; operating profit …mn (66.2% of revenue). The old thesis’s third falsifier — “growth is accelerating, not decelerating” — had remained unverifiable at triage and now has an answer: … (Q4 FY2026), … (Q1 FY2027), … (Q2 FY2027). Three consecutive quarters of acceleration, confirmed from the filings. Guidance for Q3 FY2027 — … ±2%, GAAP gross margin 74.0% ±50bp, with zero China compute revenue contribution (8-K from 26.08.2026, EX-99.1) — implies … YoY, i.e. the first deceleration, and the first guided decline in gross margin this cycle.

The break is in the balance sheet and cash flow. Three figures sum it up. First: the quarter’s operating cash flow was …mn on net profit of …mn — a 40% conversion, versus 86% the prior quarter. Of the … gap, … is the growth in receivables in a single quarter, and … is unrealized, non-cash gains from the equity portfolio. The company explicitly explains the receivables: “extended payment terms on large multi-quarter agreements with certain investment-grade customers,” with terms “ranging from 90 days up to one year” (10-Q, Liquidity and Note 7). NVIDIA has started financing its customers.

Second: in June 2026 NVIDIA issued … of senior unsecured notes in seven tranches (Note 9), raising gross debt from … to …. In the same quarter it paid … of buybacks plus … of dividends — …, more than the quarter’s CFO. This is the first time in the company’s recent history that shareholder returns plus portfolio investments exceed operating cash flow, with the difference covered by debt.

Third, the largest: off-balance-sheet guarantees went from … to …. In August 2026 NVIDIA signed guarantees capped at … for the lease obligations of an OpenAI affiliate at the SB Energy PORTS campus in Pike County, Ohio — ~4.25 GW of IT load, nine phases, the first operational in FY2029, 20-year leases, with an option for another ~3.8 GW (Note 10 and MD&A). That’s nearly twice the company’s current annual net profit, committed to a single unlisted counterparty, and it appears in no leverage metric in the tracker.

Refined thesis: NVIDIA remains the best operating business in the universe I analyze — CUDA moat intact, 75% gross margin, 66% operating margin, 117% ROE — but this quarter the business model shifted from an iron-balance-sheet component seller into a financier of its own demand: … of ecosystem investments, … of additional commitments, … of mutual cloud commitments, … of supply commitments (up from … a quarter earlier), … of guarantees, … of new debt. Each of these elements is a lever that amplifies both the good and the bad scenario, and none of them enters the P/E.

Estimated value: …–…/share depending on the model, the median of the five triangulated models at … (MOS …). Monte Carlo over 20,000 scenarios gives a median intrinsic value of … and an undervaluation probability of …%.

Verdict: MONITOR, with an explicit entry threshold lower than before. This isn’t a buy at … Nor a short: …% of scenarios give a positive MOS, and the right tail is real. The re-evaluation trigger remains a correction toward the …–180 zone (where even the conservative base scenario leaves margin) OR two consecutive quarters where DSO stabilizes below 55 days and CFO/net-profit conversion returns above 75%.