← All posts / Industry
English 繁體中文
Industry EN

Nvidia Doubles Revenue Again: Q2 FY2027 Hits $96.2B as AI Demand Stays 'Parabolic'

Nvidia's Q2 FY2027 results blew past estimates with $96.2B revenue (+106% YoY), $89B data center sales and $108B Q3 guidance — but Wall Street's reaction was muted, and supply, not demand, is now the binding constraint.

Nvidia Semiconductors Infrastructure Funding
Nvidia Doubles Revenue Again: Q2 FY2027 Hits $96.2B as AI Demand Stays 'Parabolic'

On August 26, 2026, Nvidia reported second-quarter results for fiscal year 2027 that once again made a mockery of the word “consensus.” Revenue came in at $96.2 billion, up 106% year over year and 18% sequentially, beating analyst estimates of roughly $92.3 billion. Adjusted earnings per share of $2.22 topped expectations by nearly 7%. The quarter closed on July 26, 2026, and the numbers confirm what has become the defining story of this AI buildout cycle: the demand for AI compute remains, in Jensen Huang’s own word, “parabolic” — and the bottleneck is no longer customer appetite but physical supply.

The numbers behind the beat

The headline figure that matters most to the AI industry is data center revenue, and it did not disappoint: $89.0 billion, up 117% year over year, representing 92.7% of total company sales and an 18% jump from the prior quarter. The growth was driven overwhelmingly by the Blackwell Ultra GPU infrastructure ramp, which CEO Jensen Huang cited as the primary engine of the record quarter.

Breaking down the rest of the report:

  • Revenue: $96.2B, up 106% YoY, up 18% QoQ — consensus was ~$92.2B
  • Adjusted EPS: $2.22, a 6.73% beat versus expectations
  • Data center revenue: $89.0B (+117% YoY, +18% QoQ), 92.7% of total revenue
  • Gross margin: 75.0%, a level that would have seemed impossible for a hardware company five years ago
  • Q3 FY2027 guidance: $108.0 billion, plus or minus 2% — another double-digit sequential step up

That guidance number deserves its own sentence. A $108 billion quarter would represent annualized revenue of more than $430 billion for a company that was doing $26.97 billion in the corresponding quarter of fiscal 2025 — a sixteen-fold expansion in two years, with no sign of deceleration in the reported trajectory.

Supply, not demand, is the constraint now

Perhaps the most consequential disclosure of the earnings call was management’s framing of fiscal 2028. Nvidia indicated growth of roughly 70% for FY2028 — and explicitly attributed the cap on that growth to supply constraints, not demand weakness. This is a subtle but critical distinction for anyone modeling the AI infrastructure cycle. The bear case for Nvidia has long rested on the idea that hyperscaler capital expenditure is a bubble waiting to pop, that token prices are deflating faster than volumes rise, and that depreciation schedules will eventually force a digestion period. The Q2 print doesn’t refute those arguments so much as push them out: as long as Nvidia is turning away business because it cannot manufacture chips fast enough, the demand-destruction thesis remains a forecast rather than a fact.

The supply chain picture supports this. Blackwell Ultra production continues ramping, and the constraint chain now spans advanced packaging capacity, HBM memory allocation, and even power infrastructure at the data center level. Multiple analysts noted after the call that “Wall Street’s imagination can’t keep up” with the pace Nvidia is setting — a wry acknowledgment that even the most aggressive spreadsheets keep getting lapped by reality.

The curious non-event: stock reaction

Here is where the story gets interesting for anyone who has watched Nvidia’s earnings as market theater. Despite the magnitude of the beat — revenue $4B above consensus, guidance implying another record — the stock reaction was muted, with shares essentially flat-to-slightly-up in the following sessions (one widely-shared framing: the stock “ticks up” rather than surges). Shares had actually fallen about 7% in the week before the report, to around $208, as investors braced for the print.

There are a few overlapping explanations worth taking seriously:

  1. The law of large numbers. A 106% growth rate on a $46.7B base is spectacular; the same percentage gain on a base approaching $400B annualized requires adding more revenue each quarter than most S&P 500 companies generate in a year. Investors are pricing saturation mechanics, not the current print.
  2. The FY2028 guide. 70% growth is extraordinary by any historical standard — but it is a deceleration from 106%, and the market trades on second derivatives. When management itself flags supply as the limiter, the question becomes whether that supply arrives on schedule.
  3. China remains a hole in the model. The quarter reportedly included essentially zero revenue guidance contribution from China, continuing the export-restriction saga that has seen H20-class products locked out of forecasts. Any normalization would be upside; its continued absence is a drag baked into expectations.
  4. Expectations were already enormous. A 4% revenue beat and 7% EPS beat is, by Nvidia’s recent standards, merely in line. The “whisper number” culture around this stock means beats must be historic to move it.

What it means for the AI stack

Zoom out from the stock chart, and the Q2 FY2027 report carries three broader signals for the AI ecosystem.

First, the compute deficit is real and monetizable. Data center revenue growing 117% in a year, with 92.7% of company revenue now compute infrastructure, tells you where the value in generative AI is currently accruing: not in applications, not yet in models, but in the silicon and systems underneath everything. Every model provider racing to frontier capability — from OpenAI and Anthropic to the open-weight labs in China — is bidding for the same finite pool of accelerators.

Second, the capex cycle has further to run. The $108B Q3 guide is effectively a forward indicator of hyperscaler and sovereign AI spending. When your largest customers are collectively committing hundreds of billions in multi-year data center buildouts, a single quarter’s results become less a snapshot and more a confirmation that contracts are converting to shipments on schedule.

Third, watch the margin line. Gross margin at 75.0% on hardware that is supply-constrained is a flex of pricing power. If that number holds as competition from custom silicon (TPUs, trainium-class accelerators, and the rising wave of domestic Chinese GPUs) intensifies, it will tell you Nvidia’s moat is holding. If it compresses faster than volume grows, that will be the first hard data point that the cycle is turning.

Sources

Sources

Related