NVIDIA Fiscal 2027 Q2: Data Center Revenue Hits $89B, Q3 Guided to $108B

NVIDIA Corporation (NVDA) reported Q2 fiscal 2027 total revenue of $96 billion, more than double the year-ago period. Data center revenue reached $89 billion, up 18% sequentially, accounting for the overwhelming majority of the company's top line. The quarter extended a pattern of supply-limited output: underlying demand, by management's account, continues to outpace what the supply chain can deliver.

Two Engines Inside the Data Center

Data center revenue in Q2 FY2027 broke into two distinct streams. The hyperscale segment, covering major cloud platforms building out AI infrastructure, contributed $49 billion, up 13% sequentially. Sustained deployment paces at the largest cloud providers kept the hyperscale line on a steady growth path despite already operating at significant scale.

The Accelerated Computing and Infrastructure (ACIE) segment, which covers NeoClouds, enterprise AI deployments, and sovereign AI programs, delivered $40 billion, up 25% sequentially and 138% year-over-year. ACIE now represents roughly half of NVIDIA's data center business, according to CFO Colette Kress. That share reflects a broadening of the customer base beyond the handful of companies that dominated early AI infrastructure build-outs. Sovereign AI revenue, capturing government-directed AI programs, tripled year-over-year.

The divergence in growth rates between the two segments is worth noting. Hyperscale's 13% sequential gain is substantial in absolute terms given the $49 billion base, but ACIE's 25% sequential move on a $40 billion base signals that the non-hyperscale tier is compressing the gap quickly. If ACIE's relative growth rate holds, the two segments will continue to converge in size, reshaping NVIDIA's customer concentration profile over the coming fiscal year.

Q3 Guidance of $108B and the Supply-Demand Gap

NVIDIA guided Q3 FY2027 revenue to $108 billion, plus or minus 2%. The midpoint represents a sequential step-up of roughly 12.5% from Q2's $96 billion. For full fiscal year 2028, the company projects approximately 70% revenue growth.

CEO Jensen Huang provided important context around that projection. Actual demand, he stated, is running closer to 100% annual growth. The gap between the stated demand rate and the projected revenue rate reflects supply constraints on the company's ability to produce and ship product, not a moderation in customer orders. Backlog and purchase order volumes continue to exceed near-term production capacity.

That distinction matters for reading the guidance range. The 70% FY2028 projection is a supply-limited number. The 100% figure is a demand-side signal. Whether production ramps can close the gap during the fiscal year will be a key variable in quarterly results through FY2028.

DRAM Pricing Introduces Margin Pressure

Gross margin headwinds represent the key financial tension in NVIDIA's near-term story. The company guided Q3 gross margin to 74%, down from higher levels in recent quarters. The primary driver is DRAM pricing: higher memory costs have increased the bill of materials across the data center product stack, compressing the spread between revenue and cost of goods sold.

Management communicated a clear trajectory for the pressure. Q4 gross margin is expected to trough in the 71-72% range as DRAM costs remain elevated through the remainder of the fiscal year. Recovery begins in the following fiscal year, with gross margin projected to return to the 72-73% range as memory pricing normalizes.

The compression does not alter the revenue growth trajectory, but it means that a larger fraction of incremental top-line dollars flows to cost of goods sold over the next two quarters. At the Q4 trough, gross margin would sit roughly 300-400 basis points below recent peaks, a meaningful shift for a company generating $96 billion-plus per quarter.

Vera Rubin and the Revenue-per-Gigawatt Shift

NVIDIA's Vera Rubin architecture introduces a structural change in how much revenue the company can generate from a fixed amount of data center power capacity. The Vera Rubin GPU delivers 30x higher throughput per megawatt versus Blackwell Ultra and a 35x reduction in token cost. Those figures reflect both advances in chip design and changes in the surrounding system architecture, including memory bandwidth, interconnect, and power delivery.

The revenue-per-gigawatt progression across GPU generations illustrates the compounding economics. Hopper-equipped data centers represent approximately $18 billion of revenue opportunity per gigawatt of capacity. Blackwell raises that figure to $25 billion per gigawatt. Vera Rubin extends it to $40 billion per gigawatt, a 2.2x increase over Hopper and a 60% increase over Blackwell. Each architectural generation has not simply added performance; it has raised the revenue ceiling per unit of installed power.

Production shipments of Vera Rubin have begun, and NVIDIA reported purchase orders from every major hyperscaler, cloud provider, and original equipment manufacturer. As customers build new capacity around Vera Rubin rather than prior-generation silicon, the addressable revenue per unit of power consumption rises materially for both NVIDIA and the data center operators sizing their infrastructure.

Amazon and the Hyperscaler Deployment Pipeline

Amazon Web Services (AWS) provided one of the more concrete demand signals from the quarter. Amazon is deploying an additional 2 million GPUs through fiscal 2029, with some of those paired with the new Vera CPU. The commitment spans both the Rubin GPU and Vera CPU architectures, consistent with NVIDIA's expansion beyond standalone accelerator hardware into CPUs and data center networking fabric.

The Amazon deployment is one data point in a broader pattern across the hyperscaler tier. Multi-year infrastructure commitments tied to AI compute expansion have lengthened capital expenditure cycles at the major cloud providers, extending the visibility of NVIDIA's data center revenue pipeline well beyond a single fiscal year. Full details from the Q2 FY2027 earnings call are available at /podcasts/NVDA_Q2_2027, with prior quarters and context accessible through the NVIDIA podcast feed at /groups/NVDA.

The Q2 results and Q3 guidance point to a company operating at the edge of its near-term production capacity. Gross margin compression toward the 71-72% range introduces a financial complexity over the next two quarters. The Vera Rubin architecture and the ACIE segment's 138% year-over-year growth provide the structural context for how NVIDIA's data center revenue base continues to expand as the customer mix broadens and revenue per gigawatt rises across each product generation.

• Q2 FY2027 total revenue: $96 billion, more than double year-over-year
• Data center revenue: $89 billion, up 18% sequentially
• Hyperscale segment: $49 billion, up 13% sequentially
• ACIE segment: $40 billion, up 138% year-over-year, now roughly half of data center
• Sovereign AI revenue: tripled year-over-year
• Q3 FY2027 revenue guidance: $108 billion, plus or minus 2%
• FY2028 projected growth: approximately 70%, versus stated demand closer to 100%
• Q3 gross margin guidance: 74%; trough expected at 71-72% in Q4
• Vera Rubin: $40 billion revenue opportunity per gigawatt versus Hopper's $18 billion
• Amazon: 2 million additional GPUs deploying through fiscal 2029