{"rewrite":{"id":"r_b42fd8fb390e0308d61978cd","clusterId":"c_04a00f7ad970fadac0e346e3","slug":"adobe-reveals-two-thirds-of-gpus-sat-idle-during-ai-training","model":"deepseek-v4-flash","headline":"Adobe Reveals Two-Thirds of GPUs Sat Idle During AI Training","summary":"At GTC 2026, Adobe CTO Ellie Greenfield said that in early training runs, about two-thirds of thousands of GPUs were idle, wasting roughly $600,000 per $1 million spent. Causes included slow data delivery and uneven workload distribution. Adobe fixed this with a high-performance network and fragmented checkpointing.","whyItMatters":"The disclosure shows that even a major AI developer can lose most of its GPU budget to idle time, and that the fixes were about data plumbing, not model design.","webCardHtml":"\u003cp\u003eAdobe\u0026#39;s CTO Ellie Greenfield detailed at GTC 2026 how the company built custom large-scale generative AI models, and what the profiler output showed. The data pipeline, not the model, was the bottleneck. Adobe replaced standard Ethernet with a high-performance fabric and split checkpoints into fragments saved across multiple locations, cutting the time spent on continuous write operations. The second problem, uneven workloads, was a data loader issue: equal-sized chunks did not mean equal processing time. Backblaze, which explained the presentation, noted the issue is not unique to Adobe.\u003c/p\u003e","blueskyPost":"Adobe CTO says early AI training runs left ~2/3 of GPUs idle, costing $600k per $1M. Fix: faster network, fragmented checkpoints. https://gigazine.net/news/20260826-gpu-idle/","twitterPost":"At GTC 2026, Adobe CTO Ellie Greenfield said ~2/3 of GPUs sat idle in early training runs, wasting $600k per $1M. The fix: a high-performance network and fragmented checkpoints. https://gigazine.net/news/20260826-gpu-idle/","threadsPost":null,"newsletterBlurb":"Adobe's CTO revealed at GTC 2026 that early AI training runs left about two-thirds of GPUs idle, costing roughly $600,000 per $1 million spent. The causes were slow data delivery and uneven workloads. Adobe solved it by upgrading the network and splitting checkpoints into fragments.","attributionJson":"[{\"source\":\"GIGAZINE\",\"url\":\"https://gigazine.net/news/20260826-gpu-idle/\",\"title\":\"Why GPUs Sit Idle During AI Training and How Adobe Addressed It\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":4565,"outputTokens":536,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1788010440,"createdAt":"2026-08-29T13:30:38.000Z","publishedAt":"2026-08-29T13:31:44.000Z","updatedAt":"2026-08-29T13:30:38.000Z"},"cluster":{"id":"c_04a00f7ad970fadac0e346e3","canonicalTitle":"AIトレーニングでGPUが無駄に待機状態になってしまっている理由とAdobeが行った対処とは？","representativeArticleId":"a_cb304ebb27078448ed2a2387","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[],\"studios\":[],\"people\":[\"Eli Greenfield\"],\"type\":\"news\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-08-26T12:00:00.000Z","lastSeenAt":"2026-08-26T12:00:00.000Z","updatedAt":"2026-08-29T13:31:46.000Z"},"attribution":[{"source":"GIGAZINE","url":"https://gigazine.net/news/20260826-gpu-idle/","title":"AIトレーニングでGPUが無駄に待機状態になってしまっている理由とAdobeが行った対処とは？"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":[],"studios":[],"people":["Eli Greenfield"],"type":"news","domain":"other","is_roundup":false},"keyFacts":null}