A number that shouldn’t be possible yet
One year ago, NVIDIA posted $46.7 billion in quarterly revenue. On August 26, 2026, the company reported $96.2 billion for Q2 fiscal 2027 — more than double, in twelve months. That’s not a rounding error. That’s a structural shift in where enterprise capital is going.
So let’s work through what the numbers actually say, what they don’t say, and what you should be watching if AI infrastructure decisions land anywhere near your desk.
The headline figures from Q2 FY2027
NVIDIA’s Q2 FY2027 results, for the quarter ended July 26, 2026, broke down like this:
NVIDIA posted $96.2B in Q2 FY2027 revenue — a 106% year-over-year leap that marks the fourth consecutive quarter of accelerating growth.
- Total revenue: $96.2 billion — up 18% from Q1 FY2027 and up 106% year over year
- Data Center revenue: $89.0 billion — up 18% quarter over quarter and 117% year over year
- GAAP gross margin: 75.0% — up 2.6 percentage points from a year ago
- GAAP operating income: $63.7 billion — up 124% year over year
- GAAP net income: $59.7 billion — up 126% year over year
- Diluted EPS (GAAP): $2.46 — up 128% year over year
Growth accelerated for the fourth consecutive quarter. That detail matters more than the absolute number. Deceleration is the thing analysts watch for. It hasn’t arrived yet.
Data Center is the whole story — almost
$89 billion of a $96 billion quarter came from Data Center. That’s 92% of total revenue from a single segment. Within that segment, two sub-lines are worth separating out.
Hyperscale contributed $49B while Enterprise ACIE surged 138% YoY to $40B — the fastest-growing sub-segment and a sign enterprise adoption is closing the gap.
Hyperscale revenue reached $49 billion, up 13% sequentially. But the faster-growing line was ACIE — AI, cloud, and infrastructure enterprise — which hit $40 billion, up 25% sequentially and 138% year over year. Enterprise adoption is no longer trailing hyperscaler spend. It’s closing the gap fast.
Networking also hit a record. Spectrum-X Ethernet grew 2.6 times year over year. NVIDIA’s networking business is increasingly a revenue line in its own right, not just a complement to GPU sales.
The Vera Rubin ramp — and what inventory signals
NVIDIA’s inventory rose to $32 billion during the quarter. That number sounds alarming in isolation. In context, management attributed it directly to preparation for the Vera Rubin platform launch — described on the earnings call as the fastest product ramp in NVIDIA’s history, with purchase orders already in from all major customers.
Vera Rubin is the next-generation compute platform succeeding Blackwell. Production has started. If the ramp executes, that inventory converts to revenue in coming quarters. If it stumbles, it’s a risk to watch in Q3 and Q4 results.
Days Sales Outstanding also rose to 60 days. Management linked this to extended payment terms on large purchases — typical for deals at this scale — rather than collection problems. Worth monitoring, but not a red flag on current evidence.
The China caveat embedded in Q3 guidance
Q3 FY2027 guidance came in at $108 billion, plus or minus 2%. That’s another 12% sequential step up. But one line in the guidance deserves attention: NVIDIA stated it is not assuming any Data Center compute revenue from China in that outlook.
That’s a meaningful disclosure. China has historically been a significant market. Removing it from the forecast entirely — and still guiding to $108 billion — tells you something about the strength of demand everywhere else. It also tells you that any policy shift on China export controls could represent upside, not just risk.
Gross margin guidance for Q3 is approximately 74%, down roughly one point from Q2. Management flagged Q4 as the likely trough, in the 71–72% range, with fiscal 2028 expected to settle at 72–73%. The Vera Rubin ramp carries higher initial costs. That’s the explanation, and it’s consistent with what we’ve seen on prior platform transitions.
Sovereign AI and the geography of demand
Sovereign AI revenue grew 35% sequentially and more than tripled year over year in Q2. Governments and national cloud programs are now a distinct demand category — not a rounding line inside enterprise. This matters for operators thinking about where the next wave of infrastructure build-out lands geographically.
AWS expanded its partnership with NVIDIA during the quarter, deploying an additional 2 million GPUs and adopting NVIDIA’s full physical AI stack for warehouse robotics. That partnership signal points to AI moving from pure compute workloads into physical infrastructure — a different procurement cycle and a different buyer inside most organizations.
Capital return: $26 billion back to shareholders in one quarter
NVIDIA returned approximately $26 billion to shareholders in Q2 — $20 billion through share repurchases and $6 billion through dividends. The company had roughly $99 billion remaining under its buyback authorization at quarter end.
NVIDIA's capital return program hit a new scale in Q2 FY2027, with $20B in buybacks, a $6B dividend payout, and $99B still remaining under buyback authorization.
The next quarterly dividend of $0.25 per share is scheduled for October 1, 2026, to shareholders of record on September 10, 2026. At this cash generation rate, the buyback authorization gives NVIDIA significant runway — and signals management’s view of where the stock sits relative to intrinsic value.
What this means if you’re buying, building, or budgeting AI infrastructure
Three observations worth carrying into your next planning conversation:
First, the demand signal is not slowing. Four consecutive quarters of accelerating growth, with a $108 billion guide for Q3, suggests the hyperscaler and enterprise capex cycle has further to run. If your organization is still in “wait and see” mode on AI infrastructure, the window for cheap optionality is narrowing.
Second, enterprise is catching up to hyperscale. ACIE revenue growing 138% year over year — faster than the hyperscale line — means the tools, the use cases, and the willingness to spend are now present outside the top five cloud providers. In our experience, that’s typically when procurement cycles accelerate inside mid-market and large enterprise buyers.
Third, watch the gross margin trajectory. A guided trough of 71–72% in Q4 is not a crisis — it’s a platform transition cost. But if margins compress further than guided, or if the Vera Rubin ramp encounters supply issues, that’s the leading indicator to watch, not the revenue line.
The single action worth taking this week: pull your organization’s AI infrastructure roadmap and check whether your planning assumptions still reflect a world where NVIDIA’s Data Center revenue was $38 billion a year. It isn’t that world anymore.
Eagentix helps growth-focused enterprises transform manual, time-consuming business processes into fast, dependable automated operations. By combining executive strategy with tailored smart automation, we empower companies across Southeast Asia to scale productivity, ensure regulatory compliance, and reduce operational costs by up to 70%.