Nvidia CFO says non-hyperscale cloud already accounts for roughly half of data-center revenue

Nvidia’s data center business, which now accounts for roughly 92% of the company’s total revenue, is no longer just a story about Microsoft, Google, and Amazon writing enormous checks. CFO Colette Kress has pointed to a near-even split between hyperscale cloud customers and what the company calls its ACIE segment, covering AI clouds, industrial clients, and enterprise buyers.

The numbers behind the shift

Nvidia introduced this new revenue segmentation in Q1 of its fiscal year 2027, which wrapped up in April 2026. That quarter, total data center revenue hit $75.2 billion, with hyperscale customers, think Microsoft Azure, AWS, Google Cloud, Meta, and Oracle Cloud Infrastructure, accounting for approximately $38 billion, or about 50% of the total.

By Q2 FY2027, ending in late July, the data center segment had surged to a record $89.0 billion. That’s a 117% year-over-year increase. In that same quarter, hyperscale revenue came in at $48.7 billion while the ACIE category delivered $40.3 billion.

The ACIE cohort includes AI-native startups building their own compute infrastructure, enterprises investing in on-premise AI deployments, regional cloud providers, and sovereign customers (governments and state-backed entities building domestic AI capacity).

Why diversification matters for Nvidia

The hyperscalers are sophisticated buyers with significant bargaining leverage. They negotiate aggressively, develop custom silicon alternatives (Google’s TPUs, Amazon’s Trainium and Inferentia chips, Meta’s MTIA), and have the engineering talent to explore non-Nvidia architectures. Enterprise and AI-native customers, by contrast, tend to be more dependent on Nvidia’s CUDA software ecosystem.

Kress has pointed to strong momentum from AI cloud providers and sovereign customers specifically. The sovereign AI trend, where national governments fund domestic compute infrastructure for security and economic competitiveness reasons, has accelerated over the past two years.

The $3-4 trillion opportunity

Nvidia’s management has estimated that global spending on AI infrastructure could reach between $3 trillion and $4 trillion by the end of the decade. That figure encompasses data centers, networking equipment, cooling systems, power infrastructure, and the GPUs at the heart of it all.

Nvidia doesn’t just sell GPUs. It provides the full-stack platform, from hardware to networking (InfiniBand, now transitioning to NVLink-based fabrics) to software frameworks (CUDA, cuDNN, TensorRT) that most AI workloads are built on.

For investors tracking Nvidia’s trajectory, the customer diversification story may end up being as important as the raw revenue growth. The ACIE segment’s trajectory will be worth watching closely in coming quarters, as non-hyperscale revenue at $40.3 billion versus hyperscale at $48.7 billion in Q2 FY2027 shows the gap between the two categories narrowing.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.