Nvidia May 20 earnings will test durability of AI hardware supply squeeze

Nvidia is due to report first-quarter fiscal 2027 results on May 20, and the release has become a test of whether the AI hardware boom is still outrunning supply. The company’s investor relations site lists the earnings event for 2 p.m. Pacific time that day. Nvidia’s most recent reported quarter, the fourth quarter of fiscal 2026, showed revenue of $68.1 billion, up 73% from a year earlier, after earlier quarters had already decelerated from triple-digit annual growth. The question for this report is not whether Nvidia remains large or profitable. The question is whether customers are still ordering AI systems fast enough to keep pressure on chips, networking gear, data-center power and the specialist labor needed to deploy them. Jensen Huang, Nvidia’s chief executive, said in May 2025 that “global demand for NVIDIA’s AI infrastructure is incredibly strong,” and he tied that demand to Blackwell systems and the spread of AI agents. (investor.nvidia.com) ### How much has Nvidia’s growth already slowed from the peak? Nvidia’s reported numbers show a clear slowdown from the 2025 surge, even as revenue kept rising. Fiscal 2025 revenue was $130.5 billion, up 114% from the prior year, according to the company’s annual results. Fiscal 2026 revenue then rose to $215.9 billion, up 65%. Quarterly year-over-year revenue growth moved from 78% in the fourth quarter of fiscal 2025 to 69% in the first quarter of fiscal 2026, 56% in the second quarter and 73% in the fourth quarter of fiscal 2026. (investor.nvidia.com) That sequence matters because investors have spent two years treating Nvidia as the clearest measure of AI infrastructure spending. If the company reports another quarter of very high expansion, it would show that Blackwell shipments and related system sales are still offsetting a much larger revenue base. If growth slows further, the result would still be large in absolute dollars, but it would sharpen attention on how much hyperscalers, enterprises and governments are willing to spend in 2026. (nvidianews.nvidia.com) That reading is an inference from Nvidia’s reported growth path and the company’s position in AI servers. ### Why are China and other geographies part of the earnings question? April 9, 2025, marked a clear break in Nvidia’s China business. The company said the U.S. government informed it that a license would be required for exports of H20 products into China, leading Nvidia to record a $4.5 billion charge tied to excess inventory and purchase obligations. Nvidia said first-quarter fiscal 2026 H20 sales were $4.6 billion before the new licensing rules and that it could not ship an additional $2.5 billion of H20 revenue in that quarter. (investor.nvidia.com) August 27, 2025, showed the effect continuing. Nvidia said there were no H20 sales to China-based customers in the second quarter of fiscal 2026, though it recorded about $650 million in unrestricted H20 sales to a customer outside China and a $180 million release of previously reserved inventory. (investor.nvidia.com) Those disclosures make geography central to the May 20 report. Any update on where demand is being replaced — whether by U.S. cloud providers, enterprise buyers or sovereign AI projects in other regions — would help explain whether lost China revenue is being absorbed elsewhere. Huang said in May 2025 that countries around the world were treating AI as essential infrastructure, a formulation Nvidia has used to describe government-backed demand. (nvidianews.nvidia.com) ### What does Blackwell need to prove in this report? Blackwell is the product line investors are using to judge whether Nvidia can keep expanding at its current scale. Nvidia said in August 2025 that Blackwell data-center revenue grew 17% sequentially, and Huang said then that production of Blackwell Ultra was ramping “at full speed” and that demand was “extraordinary.” In May 2025, he had already said the Blackwell NVL72 system was in full-scale production across system makers and cloud service providers. (investor.nvidia.com) The May 20 report will show whether those comments translated into another step up in reported revenue, margins and guidance. Investors will also look for any detail on networking and rack-scale systems because Nvidia has described NVLink rack-scale computing as a major part of the current buildout. (nvidianews.nvidia.com) ### Why are data-center power and colocation part of the Nvidia story? January 14, 2026, offered a measure of the strain outside Nvidia itself. Data Center Knowledge, citing CBRE, reported that primary market vacancy rates in North American data centers fell to 1.6% in the first half of 2025 and that nearly three-quarters of 5,242 megawatts under development were already pre-leased. The publication also cited CBRE data showing average rates of $184 per kilowatt per month for 250-500 kilowatt deployments, while requirements of 10 megawatts or more saw pricing rise by as much as 19%. (nvidianews.nvidia.com) Andrew Baffoe of Myriad360 told the publication that colocation discussions in 2026 were shifting from rack counts to megawatts, delivery dates and price per kilowatt. That matters because Nvidia’s systems are increasingly deployed in power-dense AI clusters, and sustained demand for those clusters would keep pressure on the same colocation and networking markets used by other compute-heavy industries. (datacenterknowledge.com) May 20 is the next concrete checkpoint. Nvidia’s investor site says the company will release first-quarter fiscal 2027 results and hold a webcast at 2 p.m. Pacific, and investors will be parsing the filing, prepared remarks and Huang’s comments for updates on Blackwell, China exposure and the pace of AI infrastructure orders. (investor.nvidia.com) (datacenterknowledge.com)