GPUs priced 29% higher, availability falls to 9%

An X post on May 21 said GPU prices had risen 29% while availability had fallen from about 80% to roughly 9%, describing the market as a “seller’s market.” The post, published by user pepe_maltese, included screenshots of retail listings and tied the squeeze to companies building AI cloud capacity, including Nebius Group and IREN. The figures in the thread could not be independently reproduced from the underlying screenshots alone, but broader market reporting and company disclosures show tight supply and aggressive procurement across AI infrastructure. TechSpot reported on February 16 that GPU prices had already moved higher across 10 regions, using in-stock listings from regional price-comparison sites as a snapshot of street pricing. In that survey, the RTX 5050 was up 9% on average and the RTX 5060 was up 36% in India, one of the steeper regional increases cited in the report. ### Where did the 29% price jump and 9% availability claim come from? The May 21 X thread came from a market commentator rather than a retailer, distributor or chipmaker. The post presented screenshots of listings and comments, then argued that buyers in the middle of the GPU market were facing higher costs this week. The screenshots appear to describe spot conditions rather than an official industry index. (techspot.com) Without a published methodology, the post’s exact 29% and 9% figures should be treated as a market snapshot from the author’s feed rather than a verified benchmark. The broader direction, though, matches independent reporting that retail GPU prices have remained elevated in multiple regions. ### Are there signs that GPU supply is still tight beyond one social-media thread? TechSpot’s February survey showed that the cheapest in-stock gaming GPUs had become more expensive in several countries, with increases varying by model and market. The publication said it tracked minimum available prices across 10 regions and found that cards that had been near MSRP in November 2025 were selling higher in early 2026. (techspot.com) IREN said on March 4 that it had entered purchase agreements for more than 50,000 NVIDIA B300 GPUs, taking its planned fleet to 150,000 GPUs. The company said those deployments would arrive in phases through the second half of 2026 and described “speed and certainty of deployment” as increasingly important in the AI cloud market. (techspot.com) ### Why were NBIS and IREN mentioned in the post? Nebius Group said on May 13 that it had secured up to 1.2 gigawatts of power and land for a new AI factory site in Pennsylvania. The company reported first-quarter revenue of $399 million, up 684% from a year earlier, as it expanded AI cloud capacity. (iren.com) NVIDIA said on March 11 that it would invest $2 billion in Nebius as part of a partnership to scale hyperscale AI cloud infrastructure, with a plan to help Nebius deploy more than 5 gigawatts of NVIDIA systems by the end of 2030. NVIDIA said the partnership was designed to meet “rapidly growing global demand for high-performance compute.” (assets.nebius.com) IREN said on May 7 that it had signed a five-year AI infrastructure cloud services contract with NVIDIA. In separate disclosures this month, IREN said it was targeting $3.7 billion in AI cloud annualized run-rate revenue by the end of 2026 and had expanded funding to support additional hardware and data-center buildout. (investor.nvidia.com) ### Who is likely to feel the squeeze first? Mid-market buyers are the group most directly described in the May 21 thread, because they typically buy from the same channel inventory that reflects current street prices rather than long-term contracted supply. TechSpot’s regional survey also focused on the cheapest in-stock models, which is the part of the market where retail scarcity shows up fastest. (irisenergy.gcs-web.com) Large AI cloud operators, by contrast, are disclosing forward purchase agreements, financing packages and multi-quarter deployment schedules. That does not mean supply is easy, but it does mean companies such as Nebius and IREN are trying to lock in capacity before it reaches spot channels. (techspot.com) ### What are the next concrete markers to watch? IREN said its latest GPU deployments are scheduled through the second half of 2026, with capacity added in British Columbia and Texas. Nebius said on May 13 that management would discuss first-quarter results on its investor webcast, and its investor hub lists a June 9 AI forum in San Francisco. (iren.com)