NVDA Q2 FY27 — Revenue Up 18%, Free Cash Flow Down 56%. Read the 15 Extra Days.

The only tension that matters

Every headline number in this print is excellent, and I want to state that before I complicate it, because the complication is not a bear case and it should not be read as one. Revenue was $96,221M, up 18% sequentially and 106% from a year ago, against a guide of approximately $91.0B plus or minus 2%, roughly 5.7% above the midpoint (computed). Data Center was a record $89,023M, up 117%. GAAP and non-GAAP gross margins were both 75.0%. Non-GAAP EPS was $2.22, up 120%. The third quarter is guided to $108.0B, which would be another 12.2% sequential step (computed) on a base that has already doubled. The stock closed the following session at $227.98 against a prior close of $209.66 — up 8.74% on 263.7M shares versus a 30-day average near 118.8M (unverified — market data, not filing figures). The tape read the guide and bought it.

So the interesting question is not whether demand is real. On this site the standing prior is that AI infrastructure demand is uncontested and the debate is always the funding, and nothing in this quarter disturbs the first half of that sentence. The interesting question is the second half, and it is answerable this time, because NVIDIA answered it in writing. Jensen Huang's line was "AI has reached its inflection point. It's doing useful work. Its tokens are productive and profitable. Now, compute is revenue." Decode the last clause literally and it is the whole finding. Compute did become revenue this quarter. Revenue then became a receivable — $63.1 billion of it, sitting sixty days out.

Two numbers that are the same number

Take the cash flow statement first, because it looks alarming in isolation and the alarm needs sizing before it can be judged. Cash flow from operating activities was $24,077M, against $50,344M the prior quarter — down 52.2% while revenue rose 17.9% (computed). Free cash flow was $21,341M against $48,554M, down 56.0% (computed), and well under a consensus that sat near $43B (unverified — estimate data). Operating cash flow as a share of revenue went from 61.7% to 25.0% (computed). GAAP net income of $59,688M exceeded operating cash flow by $35.6B in a single quarter (computed).

Now the second number, which is the first number wearing different clothes. The CFO commentary states it plainly: "Accounts receivable was $63.1 billion with 60 days sales outstanding (DSO), up from 45 days sequentially, due to extended payment terms on large, multi-quarter agreements with certain investment-grade customers." Fifteen days of additional credit, granted deliberately, disclosed without prompting. At $96.2B of quarterly revenue, fifteen days is on the order of $15.9B of revenue recognized and not yet collected (computed) — which is most of the gap between the two cash-flow quarters. The receivable and the cash-flow decline are not two findings. They are one decision, reported twice.

The company's own explanation for the cash line is thinner than its explanation for the receivable: "the sequential decrease was driven by higher working capital adjustments and cash taxes." That sentence is accurate and it is the entire narrative treatment of a $26.3B sequential swing. Note which disclosure carries the mechanism and which carries the summary. The receivable paragraph tells you the terms changed and who got them; the cash-flow paragraph tells you working capital moved. This is the house edge — the gap between the narrative and what the data supports — and here the gap is not that management hid anything. It is that the filing contains a sharper sentence about the cause than the section where the effect lands.

Where the growth actually came from

The Data Center split is where the receivable stops being a treasury footnote and starts being a thesis. NVIDIA breaks the $89,023M into two: Hyperscale at $48,710M, up 13% sequentially and 102% year over year, and AI Clouds, Industrial & Enterprise — ACIE — at $40,313M, up 25% sequentially and 138% year over year. The faster line is the second one, and it is nearly half the segment. Growth is migrating from the cohort with trillion-dollar balance sheets toward the cohort described in NVIDIA's own commitments section: "AI clouds and model makers are seeing extraordinary demand for AI infrastructure, yet many are growing faster than their balance sheets and long-term credit profiles can support."

Read that sentence twice. It is a vendor writing down, in a public filing, that the fastest-growing half of its largest segment cannot fund itself. Then read what the company does about it, in the same document: it guarantees $3.5B of AI cloud partners' data center lease obligations against default; it commits $105B of phased credit support behind roughly 4.25GW at SB Energy's PORTS-Pike campus in Ohio, exclusively hosting NVIDIA infrastructure under twenty-year leases to OpenAI; it carries $25B of equity-investment commitments "focused on AI model makers, infrastructure financiers, and other private companies"; and it extends fifteen extra days of payment terms to the customers buying the chips. Four instruments, one direction. This is the circularity loop — the vendor investing in, lending to, and now underwriting the same customers whose orders validate its forecast — and the point has never been that the loop is fraudulent. The point is that its durability is only as good as the weakest balance sheet inside it, and NVIDIA has just told you those balance sheets are the binding constraint.

The question this site keeps asking of every AI-infrastructure print is who ultimately pays for the capex. For six quarters the honest answer was some blend of hyperscaler cash flow, private credit, and neocloud equity. This filing supplies a new answer for the marginal dollar, and the answer is NVIDIA — not as the seller, but as the guarantor, the lender of working capital, and the equity holder. That is a materially different risk profile than the one the picks-and-shovels frame assumes, which is precisely why it belongs above the paywall rather than buried next to the trade.