NVIDIA CFO flags sustained GPU demand

NVIDIA said on May 20 that demand for AI computing remained strong enough to push first-quarter fiscal 2027 revenue to a record $81.6 billion, with data-center sales reaching $75.2 billion. The results put fresh attention on comments by Chief Financial Officer Colette Kress that hyperscalers now account for more than half of data-center revenue and that AI infrastructure demand continues to broaden beyond the largest cloud companies. (cnbc.com) Yahoo Finance reported on May 25 that Kress also pointed to rising rental prices for NVIDIA’s H100 and A100 GPUs this year, describing a market where supply and pricing have not followed the usual semiconductor cycle. (investor.nvidia.com) That comes as NVIDIA continues to report record data-center revenue and as Jensen Huang said in February that computing demand was growing exponentially at what he called the “agentic AI inflection point.” ### Why are investors focusing on Colette Kress’s comments? Colette Kress drew attention because her remarks came alongside another quarter of record numbers. (investor.nvidia.com) NVIDIA said first-quarter revenue rose 85% from a year earlier and 20% from the prior quarter, while data-center revenue climbed 92% year over year. Yahoo Finance said Kress cited higher GPU rental prices in 2026, with H100 rental prices up 20% and older A100 rental prices up 15%. That detail matters because it points to continued scarcity in the market for high-end AI compute, even as NVIDIA ships more systems. (finance.yahoo.com) ### Where is the demand coming from? NVIDIA said hyperscalers generated more than half of data-center revenue in the quarter. CNBC reported Kress said that contribution reached $38 billion, up 12% from the prior quarter. (investor.nvidia.com) Bisnow reported on May 22 that demand is also coming from “neocloud” providers and startups that rent GPU capacity to other companies. CIO Dive separately reported that NVIDIA’s customer base is diversifying, with roughly half of data-center revenue tied to hyperscalers and the rest linked to enterprises and other buyers. (finance.yahoo.com) ### How does this connect to training and inference demand? Jensen Huang said in NVIDIA’s February earnings release that Grace Blackwell with NVLink was leading inference on cost per token, and he tied demand to the spread of agentic AI. (cnbc.com) That language matters because NVIDIA is selling into both model training and the growing market for inference, where customers run models continuously after they are built. The company’s May 20 results reinforced that demand pattern. (bisnow.com) NVIDIA guided for about $91 billion in second-quarter revenue, according to earnings coverage and call summaries published after the report. ### Why does sustained GPU scarcity matter beyond NVIDIA’s sales? Data-center operators face planning constraints when GPU supply stays tight. Bisnow reported that NVIDIA’s latest earnings showed AI demand extending beyond the biggest cloud platforms, increasing pressure on capacity and new buildouts. (investor.nvidia.com) Security and operations teams can also face tradeoffs when high-value GPU clusters are hard to replace or take offline. The broader infrastructure briefing tied that scarcity to longer patch windows, tighter workload consolidation and harder isolation choices in shared environments, an effect that follows from constrained capacity rather than from any new NVIDIA disclosure. (seekingalpha.com) ### What comes next? NVIDIA said its 2026 annual meeting of stockholders is scheduled for June 24 at 9:00 a.m. Pacific Time. Before then, investors will be watching whether second-quarter revenue reaches the roughly $91 billion target set after the May 20 earnings report and whether management provides further detail on hyperscaler spending, inference demand and supply conditions. (bisnow.com) (investor.nvidia.com 1) (investor.nvidia.com 2)