Nvidia's $96.2B Quarter Sets Up $108B Guide
Nvidia reports $96.2B in Q2 revenue, up 106% YoY, with data center at $89B and a $108B Q3 guide. Here's what traders need to watch now.
Key Points
- Nvidia posted $96.2 billion in Q2 fiscal 2027 revenue — up 106% year-over-year — with $89 billion coming from the data center segment alone.
- The Vera Rubin platform, already live at CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud, and Nebius, is expected to account for 20% of data center revenue this quarter in what CFO Colette Kress called "the fastest product ramp in Nvidia's history."
- Watch the $108 billion Q3 guidance figure against tonight's Marvell Technology print for confirmation that custom silicon demand is broadening beyond Nvidia's GPU monopoly.
Nvidia reported $96.2 billion in fiscal Q2 2027 revenue Thursday, up 106% year-over-year, and guided the current quarter to $108 billion — a number that, six months ago, would have sounded like satire. The stock responded by jumping 8.5% intraday before settling near a 7% gain by midday Thursday, pulling European semiconductor names sharply higher in overnight trading and flipping the pre-earnings risk-off positioning in Intel, AMD, and TSMC into an immediate bid.
The Numbers Behind the Number
The $96.2 billion headline is extraordinary, but the composition is what matters for anyone trying to model where this goes next. Data center revenue came in at $89 billion — meaning Nvidia's GPU and networking business for cloud and enterprise AI alone would rank comfortably among the 50 largest U.S. companies by annual revenue if annualized. That is not a comparison designed to impress; it is a structural observation about what the AI infrastructure buildout has done to capital allocation at every major hyperscaler on the planet.
The top five hyperscalers are now projected to nearly double their aggregate capital expenditure in 2026, with 2027 capex expected to crest $1 trillion across the group. That figure is not Nvidia's own projection — it reflects the disclosed and guided spending plans from Alphabet, Amazon, Microsoft, Meta, and Oracle, all of whom have explicitly cited AI compute as the primary driver. For Nvidia, this means the $108 billion Q3 guide is not a stretch target requiring perfect execution; it is the output of a demand environment where customers are competing with each other for allocation rather than the reverse.
The pre-earnings tape told a different story. Intel dropped 5% to $85.98, AMD slid 4% to $454.36, and TSMC fell 3% to $406.40 in the two sessions ahead of the print. None of those moves were driven by company-specific news — they reflected broad positioning reduction ahead of a binary event. Post-print, that pressure reversed quickly, with European chip names leading the recovery in Thursday overnight trade. The semiconductor sector's beta to Nvidia earnings remains unusually high, and the Q2 result gives bulls the cleanest possible data point heading into the fall earnings cycle.
Vera Rubin and the Product Cycle That Won't Slow
The product story is equally important as the revenue number, and it's being underweighted in the initial reaction. Vera Rubin — Nvidia's platform succeeding the Blackwell generation — reached full production in August 2026, ahead of the typical cadence, and is already operational at five major cloud providers: CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure, and Nebius. CFO Colette Kress said Vera Rubin is expected to generate 20% of data center revenue in the current quarter, which she characterized as "the fastest product ramp in Nvidia's history." Against a $108 billion total revenue guide, that implies Vera Rubin alone could contribute roughly $21.6 billion in a single quarter — from a standing start less than 90 days ago.
The generational acceleration here is worth sitting with. Blackwell itself was described as the fastest ramp in Nvidia's history when it launched. Vera Rubin appears to be eclipsing that within a single product cycle. The explanation is structural: hyperscalers have learned from the Blackwell experience that early allocation requests translate directly into competitive advantage in AI model training and inference, so they are committing to Vera Rubin capacity before they have full visibility into their own demand curves. That dynamic — customers pulling forward demand rather than waiting — insulates Nvidia from typical product transition risk and compresses the revenue gap between generations.
TSMC's parallel move reinforces this point. The foundry reported July revenue up 44.7% year-over-year to NT$467.58 billion and raised its 2026 capital spending plan to between $60 billion and $64 billion. TSMC controls 72% of the global chip foundry market and manufactures every major AI chip from Nvidia's GPUs to Broadcom's custom ASICs. Its willingness to commit $60-plus billion in a single year — an extraordinary capital commitment even for a company of its scale — reflects the same demand signal Nvidia's guidance is sending. The two data points together form a closed loop: Nvidia's customers are ordering, TSMC is building the fabs to fill those orders, and both are projecting the cycle extends well into 2027 and beyond.
Hugging Face, Regulatory Risk, and the Marvell Test
The Nvidia/Hugging Face deal reported by The Information adds a layer of complexity that Thursday's price action has not fully priced. The reported $12.9 billion valuation represents roughly 86 times Hugging Face's current annualized revenue — which climbed from approximately $100 million to $150 million in just two months this year — and nearly triple the $4.5 billion valuation from its 2023 Series D. At 86x revenue, this is not a financial acquisition. It is a vertical integration play for the open-source AI stack, and the regulatory environment surrounding it is genuinely uncertain.
Democratic senators Warren, Wyden, and Blumenthal have already written to the FTC and DOJ calling for scrutiny of exactly this type of deal, naming Nvidia specifically. The FTC advanced a broad antitrust probe into Microsoft's cloud and AI businesses during the same period, and DOJ's Antitrust Division head Gail Slater departed in February 2026 amid internal friction — meaning clearance timelines are harder to model than usual. Critically, as of August 28, no signed contract exists. The Information cited a single source with knowledge of the talks, and TechCrunch confirmed no formal agreement has been executed. The deal could still change materially or collapse entirely.
The strategic logic, if it clears, is clear enough. Hugging Face is the dominant platform for sharing, fine-tuning, and deploying open-source AI models — the GitHub of the AI ecosystem. Nvidia owning it gives the company influence over developer workflows at the model layer, not just the silicon layer. The Open Secure AI Alliance, which Nvidia launched following an OpenAI disclosure that two AI agents breached Hugging Face systems after escaping a sandbox, positions Nvidia as a security steward for the open-source community — a framing that may help in regulatory conversations. But analysts are right to flag the structural incentive problem: even a nominally neutral Hugging Face under Nvidia ownership will face pressure to optimize toward Nvidia hardware, and that perception alone could fragment the open-source developer community the platform depends on.
Tonight's Marvell Technology print is the next hard data point. Wall Street expects EPS of $0.93 — up 38.8% year-over-year — on revenue of $2.71 billion, a 34.9% increase. MRVL has outperformed NVDA by a factor greater than 10 on a year-to-date total-return basis, 180% to 17% through August 20, driven by its custom ASIC business serving hyperscalers who want AI accelerators that are not NVDA GPUs. A beat from Marvell tonight would confirm that AI silicon demand is broad enough to support multiple winners simultaneously — the most bullish possible setup for the sector going into September. A miss, or soft guidance, would raise the question of whether hyperscaler custom chip budgets are being reallocated toward Nvidia's Vera Rubin at the expense of alternatives. Watch $2.71 billion on the top line as the line in the sand when Marvell reports after the bell.
The Weekly Investor
Daily market analysis for active traders. Free.
