Nvidia Q2 FY27: $96.2B Beat, Guides Q3 to $108B ±2%

Nvidia delivered a fiscal Q2 that ran meaningfully hotter than the number it printed just three months ago. Revenue landed at $96.2 billion, up 106% year over year and 18% sequentially, roughly $5 billion above the company’s own $91 billion guide. Data center revenue alone reached $89.0 billion, up 117% year over year. And the Q3 outlook came in at $108 billion at the midpoint, implying another double-digit sequential step higher.

The three-line summary: demand is not slowing, mix continues to shift toward accelerated computing, and the guide keeps racing the print.

The numbers

Metric Q2 FY27 YoY QoQ
Total revenue $96.2B +106% +18%
Data Center revenue $89.0B +117% n/d
GAAP EPS (diluted) $2.46 n/d n/d
Non-GAAP EPS (diluted) $2.22 n/d n/d
Gross margin (GAAP & non-GAAP) 75.0% n/d n/d
Q3 FY27 revenue guide (midpoint ±2%) $108.0B +12% seq.
Source: Nvidia press release, Aug. 26, 2026. “n/d” = not directly disclosed in this excerpt.

The Q3 guide is the number that moved the tape

Guides matter more than prints for a story stock like Nvidia, and this one landed above where the sell side had been drifting. The company set Q3 FY27 revenue at $108 billion at the midpoint, plus or minus 2%, implying a range of roughly $105.6 billion to $110.4 billion. Off a $96.2 billion base, that is another ~$12 billion of sequential revenue in a single quarter — a step function that only a handful of companies in market history have ever pulled off.

What the guide is really saying: hyperscaler capex plans are still being revised up, not down. Between Google, Microsoft, Meta, Amazon, and the emerging tier of “neocloud” buyers, the addressable pipeline through calendar 2027 keeps expanding faster than Nvidia can build to fill it.

Nvidia quarterly revenue with Q3 FY27 guide Bar chart of Nvidia quarterly total revenue: $46.7B in Q2 FY26, $81.5B in Q1 FY27, $96.2B in Q2 FY27, and a $108B midpoint guide for Q3 FY27. $120B $90B $60B $30B $0 Q2 FY26 $46.7B Q1 FY27 $81.5B Q2 FY27 $96.2B Q3 FY27E $108B Nvidia total revenue by quarter (with Q3 FY27 guide)
Source: Nvidia Q2 FY27 release, Aug. 26, 2026. Q1 FY27 back-implied from the disclosed 18% QoQ growth; Q2 FY26 back-implied from the disclosed 106% YoY growth.

Data center: 92% of the mix

Data center revenue was $89.0 billion, or roughly 92% of the total. That share was 88% a year ago, so the mix keeps tilting — every incremental dollar of Nvidia revenue is now overwhelmingly a rack of AI accelerators, networking, and system software, not a gaming GPU or a Quadro card.

Two things stand out inside that number. First, the year-over-year growth rate of 117% actually reaccelerated versus the 114% pace Nvidia reported for data center in Q1 FY27, per its Q1 press release. Off a nearly $90 billion base, that is unusual. Second, gross margin held at 75.0% on both GAAP and non-GAAP measures, which suggests the Blackwell-generation product ramp is not hurting unit economics the way some sell-side models had feared.

What Jensen said

CEO Jensen Huang framed the print in language that has become his standard: “AI has reached its inflection point. It’s doing useful work…and demand is accelerating. This time last year, one lab alone was driving the buildout; today, we have a golden age of new AI labs and startups.” He described the infrastructure buildout as “at full steam.”

Under that rhetoric, three concrete details matter for anyone modeling the story: Nvidia says frontier-lab deployments are broadening beyond a handful of accounts, sovereign AI programs continue to layer in as an incremental customer set, and the pipeline of enterprise and neocloud commitments now stretches meaningfully into calendar 2027.

Peer read-through

An Nvidia guide of this magnitude has never been contained to Nvidia itself. The obvious beneficiaries: system integrators (Dell, Supermicro) that ship the racks, memory (Micron, SK Hynix, Samsung) whose HBM is the binding constraint on Blackwell throughput, foundry (TSMC) whose CoWoS packaging capacity gates supply, networking (Arista, Credo, Marvell, Broadcom’s custom-silicon business), and the tier of independent “neocloud” GPU landlords (CoreWeave, Nebius, Nscale) whose entire model is renting Nvidia iron.

The counter-read is worth keeping in view: hyperscaler capital intensity keeps climbing, which pressures free cash flow at the customers even as it lifts the supplier. AWS, Google Cloud, Azure, and Meta have collectively guided calendar 2026 capex materially higher than 2025, and every incremental Nvidia guide raise is another line item those CFOs have to defend to their own shareholders.

What to watch next

  • The 10-Q filing — segment detail on Gaming, Professional Visualization, and Automotive isn’t in the press-release summary; the SEC filing will have it, along with China-specific revenue commentary.
  • Blackwell Ultra and Rubin timing — Nvidia’s product cadence has compressed to roughly one architecture per year. The next platform ramps into the Q3-guide window.
  • Hyperscaler responses — watch the next round of Microsoft, Google, Amazon, and Meta earnings for whether calendar 2026 capex plans are revised again to accommodate Nvidia’s supply.
  • Export-control friction — US restrictions on advanced accelerators shipped to China remain a real variable and can move a quarter’s guide by a few billion dollars.

Sources

Disclosure: This article is for informational purposes only and is not investment advice.

Leave a Comment