[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-the-cost-of-ai-intelligence-is-falling-faster-than-compute-ever-did":10,"sections":40},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":22,"persona_id":22,"persona_name":22,"section":30,"tags":31,"sources":35,"feedback":39,"feedback_at":22,"cost_usd":39,"total_tokens":39},8770,"the-cost-of-ai-intelligence-is-falling-faster-than-compute-ever-did","The Cost of AI Intelligence Is Falling Faster Than Compute Ever Did","A new analysis finds AI model costs dropping roughly 50% every quarter, outpacing compute, DNA sequencing and batteries, though the pace is already slowing.","AI model costs are falling faster than any other technology on record, according to a new analysis.\n\nResearch firm Epoch AI tracked the price of getting an AI model to answer a tough graduate-level science question correctly, measured against the GPQA Diamond benchmark, and found costs dropping almost 50% every quarter - about 13 times cheaper per year. That outpaces the long-run cost curves for computer chips, DNA sequencing and lithium-ion batteries, each of which took one to six decades to see comparable declines. Epoch points to OpenAI's o3 model, which answered a GPQA Diamond question for $0.30 in January 2025, and a newer OpenAI model released roughly a year and a half later that matched that score for a fraction of a cent - one of the sharper drops in a pattern Epoch describes as lumpy rather than steady, with long plateaus followed by sudden plunges. Across five benchmarks, the average quarterly cost decline has actually slowed, from 66% two years ago to about 32% now.\n\nThat trend undercuts the business case for sticking with any one AI vendor: whatever edge a frontier model offers today may be available for pennies within months. It is an awkward backdrop for OpenAI, Anthropic and other labs pouring hundreds of billions into infrastructure, since cheaper open-weight competitors can erode that advantage almost as fast as it appears.\n\nEpoch flags its own weak spot, too - the results lean heavily on one benchmark family that is vulnerable to models being tuned to ace the test rather than genuinely getting smarter, so the comparison to Moore's Law is a useful headline, not a settled fact.","[\"ai\",\"ai pricing\",\"openai\",\"benchmarks\"]","2026-09-30T12:40:00.000Z","2026-10-01T01:24:52.892Z","2026-10-01T01:24:58.187Z","published",null,[24],{"id":25,"reviewer":26,"round":27,"reason":28,"status":29},"publisher-r1","publisher",1,"The headline figures are internally inconsistent — a ~50%-per-quarter decline over six quarters (Jan 2025 to mid-2026) implies roughly a 60x cost drop, not the 750x drop implied by the $0.30-to-$0.0004 example, and the cited model name 'GPT-5.6 Luna' is not a real\u002Fverifiable OpenAI release.","resolved","ai",[30,32,33,34],"ai pricing","openai","benchmarks",[36],{"name":37,"url":38},"Tom's Hardware","https:\u002F\u002Fwww.tomshardware.com\u002Ftech-industry\u002Fartificial-intelligence\u002Fthe-price-of-ai-is-crashing-faster-than-the-rate-of-moores-law-report-suggests-intelligence-costs-are-in-freefall-outpacing-comparative-technologies-like-compute-dna-sequencing-and-lithium-batteries",0,{"sections":41},[42,46,51,56,61,66,71,76,81,85,90,95,99,104],{"name":43,"slug":30,"count":44,"latest_published_at":45},"AI",5223,"2026-09-30T16:24:23.000Z",{"name":47,"slug":48,"count":49,"latest_published_at":50},"Security","security",796,"2026-09-30T16:10:00.000Z",{"name":52,"slug":53,"count":54,"latest_published_at":55},"Policy","policy",421,"2026-09-30T16:24:16.000Z",{"name":57,"slug":58,"count":59,"latest_published_at":60},"Deals","deals",293,"2026-09-30T15:12:37.000Z",{"name":62,"slug":63,"count":64,"latest_published_at":65},"Hardware","hardware",196,"2026-09-30T13:00:00.000Z",{"name":67,"slug":68,"count":69,"latest_published_at":70},"Science","science",156,"2026-09-30T14:30:00.000Z",{"name":72,"slug":73,"count":74,"latest_published_at":75},"Consumer Tech","consumer-tech",147,"2026-09-30T16:05:19.000Z",{"name":77,"slug":78,"count":79,"latest_published_at":80},"Dev Tools","dev-tools",91,"2026-09-30T12:58:00.000Z",{"name":82,"slug":83,"count":79,"latest_published_at":84},"Software","software","2026-09-25T20:55:00.000Z",{"name":86,"slug":87,"count":88,"latest_published_at":89},"Startups","startups",83,"2026-09-29T21:51:36.000Z",{"name":91,"slug":92,"count":93,"latest_published_at":94},"Gaming","gaming",49,"2026-09-30T16:24:30.000Z",{"name":96,"slug":97,"count":93,"latest_published_at":98},"General","general","2026-09-28T16:44:57.000Z",{"name":100,"slug":101,"count":102,"latest_published_at":103},"Reviews","reviews",31,"2026-09-28T14:31:34.000Z",{"name":105,"slug":106,"count":107,"latest_published_at":108},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]