MarketsNVDA -1.59%AUG 26, 2026

Nvidia's $96.2B Print Sends a Cost Warning to Every Small Business Buying AI Tools

Q2 revenue doubled to $96.2B and CFO Colette Kress guided hyperscaler capex to $1.3T in 2027, a compounding cost base that flows downstream into every SaaS renewal.

BY HENLEY MARRAST · MARKETS DESK

Nvidia reported fiscal Q2 2027 revenue of $96.2 billion after Wednesday’s close, up 106% year-over-year and 18% sequentially, blowing past LSEG’s $92.2 billion consensus. Adjusted EPS came in at $2.22 against the $2.10 estimate. For small-business owners renewing AI tool contracts this fall, the more consequential number is buried further down: hyperscaler capex is on track to hit $1.3 trillion in 2027, up from $800 billion in 2026.

Data center revenue was $89 billion, up 117% year-over-year, with $48.7 billion of that from the hyperscalers. The AI Clouds, Industrial and Enterprise segment grew 138% to $40.3 billion. CFO Colette Kress guided Q3 to $108 billion (±2%), ahead of the Street’s $104.2 billion, and modeled fiscal 2028 growth at 70% versus the Street’s 44%. That’s the $92B question we flagged before the print answered decisively.

CEO Jensen Huang’s release language did the framing work Nvidia wanted done. “AI has reached its inflection point.” And then: “Now, compute is revenue.” Read as elite-psychology signal, that second line is the more revealing one. Huang is telling every downstream buyer that GPU access is no longer a research expense; it’s the input cost of the product itself.

The cost passthrough is already visible. Nvidia’s own supply commitments more than doubled to $279 billion, largely for memory, which Kress described as scarcity that “is being driven in large part by the AI buildout itself.” Trendforce projects server DRAM prices up 260% in 2026. Louis Navellier, cited in Kiplinger, notes Nvidia raised prices roughly 15% for its largest customers. Those customers are the same hyperscalers whose margins power every SaaS tool a small business pays for, echoing what AWS’s Q2 print showed about compute repricing.

For a 5-to-30-person operator, the operational read is straightforward. AI line items aren’t fixed costs anymore. They’re variable inputs indexed to a supply chain that’s tightening quarter over quarter. Tools that can prove revenue lift will survive 2027 renewals. Tools sold on time-saved narratives won’t. Salesforce’s read on downstream AI demand already priced this in before the print landed.

The inflection point Huang described isn’t ahead. It’s the invoice.

Sources

About the author

Henley Marrast

MARKETS DESK

Henley Marrast covers AI-equity flow, accelerator demand, and earnings prints for AI Sheet Report. She leads coverage of the public AI complex from the New York markets desk, with a focus on the daily tape and quarterly results. She has been writing about technology markets for several years.