NVIDIA Q2 FY2027 Earnings: Revenue Scale, Margin Pressure, and Cash Generation Capacity

NVIDIA (NVDA) reported $96 billion in Q2 FY2027 quarterly revenue, more than double year-over-year and the fourth consecutive quarter of accelerating growth. Data center revenue reached $89 billion, up 18% sequentially, driven by continued expansion across hyperscale cloud platforms, enterprise deployments, and sovereign AI programs. The quarterly result extends a trajectory of compounding acceleration that has now run for four consecutive earnings periods.

Data Center Revenue: Hyperscale and ACIE Diverge

Q2 data center revenue split into two segments with distinct growth profiles. Hyperscale customers generated $49 billion in the quarter, up 13% sequentially, as major cloud providers continued building out AI training and inference capacity at scale. The ACIE segment, which covers NeoClouds, enterprise customers, and sovereign AI deployments, produced $40 billion, up 25% sequentially and 138% year-over-year.

ACIE now represents roughly half of NVIDIA's data center business, a structural shift in the revenue mix. Sovereign AI revenue tripled year-over-year as government-sponsored infrastructure programs accelerated across multiple regions. The faster sequential growth rate in ACIE relative to hyperscale indicates that enterprise and sovereign deployments remain in earlier stages of ramp compared to large cloud providers, which have been deploying at scale for longer. This mix shift adds a layer of demand diversification beyond the four largest hyperscalers.

Supply Constraints and the FY2028 Revenue Ceiling

NVIDIA projected approximately 70% full-year FY2028 revenue growth, but characterized this figure as supply-constrained. Actual demand growth is running closer to 100%, meaning manufacturing throughput and supply chain capacity, rather than customer willingness to deploy capital, represent the binding limit on revenue realized in any given period. The distinction between a supply ceiling and a demand ceiling matters for how operating cash flow scales as production capacity expands.

Q3 FY2027 revenue guidance came in at $108 billion, plus or minus 2%, a $12 billion sequential step up from the $96 billion Q2 result. If the supply-constrained dynamic persists into FY2028, the revenue trajectory will continue tracking production ramp rates rather than the size of the order pipeline. Customers seeking capacity ahead of the Vera Rubin ramp face lead times that management noted are extending across multiple quarters. For a deeper look at how these dynamics have unfolded quarter by quarter, NVIDIA earnings episodes at /groups/NVDA trace the sequential progression from Hopper through Blackwell.

Gross Margin Compression: DRAM as the Primary Driver

Near-term gross margin faces pressure from DRAM pricing spikes embedded in new product configurations. Gross margin is guided down to 74% in Q3 and is expected to bottom at 71-72% in Q4 before recovering to 72-73% in the next fiscal year. The compression does not reflect pricing pressure on NVIDIA's GPU outputs or increased competitive intensity on the hardware side.

High-bandwidth memory is a significant cost input across Blackwell and successor architectures, and supplier pricing has moved materially as HBM production has struggled to keep pace with the demand generated by AI infrastructure buildouts. The expected recovery to 72-73% in FY2028 assumes DRAM pricing normalizes as memory manufacturers bring additional capacity online. The margin trough represents a timing artifact of the current commodity cycle intersecting with a new product generation, rather than a structural change in NVIDIA's pricing position.

Revenue Per Gigawatt: The Generational Economics of GPU Infrastructure

Each GPU generation produces more revenue per unit of physical data center capacity, measured in power consumption. Hopper-based deployments generate approximately $18 billion per gigawatt of compute capacity. Blackwell raises that figure to $25 billion per gigawatt, a 39% increase. The Vera Rubin architecture, representing the next generational step, is projected at $40 billion per gigawatt, a further 60% increase over Blackwell.

This progression has a compounding effect on operating leverage across the installed base. As customers cycle through GPU generations, the revenue density of a fixed amount of physical infrastructure, measured in watts of power draw and cooling load, increases without requiring proportional increases in data center footprint. The economics favor customers who secure power commitments early, since the revenue opportunity per megawatt of capacity rises with each successive architecture. For a detailed breakdown of how these generational economics are reflected in NVIDIA's operating results, the NVIDIA Q2 FY2027 earnings breakdown at /podcasts/NVDA_Q2_2027 walks through the sequential build in cash generation capacity.

Q3 Guidance and Forward Demand Visibility

Amazon (AMZN) is deploying an additional 2 million GPUs through fiscal 2029, including Vera CPUs paired with the Rubin GPU architecture. This commitment spans multiple product generations and extends forward demand visibility well beyond the current fiscal year. The Vera Rubin deployment specifically indicates that Amazon is committing to next-generation capacity before the current Blackwell ramp has completed, reflecting confidence in the revenue-per-gigawatt economics of that architecture.

Q3 guidance of $108 billion, combined with the Amazon multi-year deployment timeline, provides a forward demand signal extending into at least FY2028 and FY2029. The interaction of supply-constrained growth projections, DRAM-driven margin pressure in the near term, and the expanding revenue-per-gigawatt progression across GPU generations frames how operating cash flow is likely to evolve across the next several fiscal periods. The supply constraint is the factor most likely to determine whether actual revenue tracks closer to the 70% guided figure or the underlying 100% demand growth rate as manufacturing capacity continues to scale.

Key Metrics at a Glance

• Q2 FY2027 total revenue: $96 billion, more than double year-over-year (fourth consecutive quarter of accelerating growth)
• Data center: $89 billion total; hyperscale $49 billion (+13% sequential), ACIE $40 billion (+25% sequential, +138% year-over-year)
• ACIE now roughly half of data center revenue; sovereign AI revenue tripled year-over-year
• Q3 FY2027 revenue guidance: $108 billion, plus or minus 2%
• FY2028 revenue growth projection: approximately 70%, supply-constrained (underlying demand growth approximately 100%)
• Gross margin path: 74% in Q3, 71-72% floor in Q4, expected recovery to 72-73% in FY2028
• Revenue per gigawatt: Hopper $18 billion, Blackwell $25 billion, Vera Rubin $40 billion
• Amazon: 2 million additional GPUs through fiscal 2029, including Vera CPU and Rubin GPU