ChipsThe story, in brief
AI performance costs are falling faster than those of any previous technology
13x per year. That's how fast AI performance costs are collapsing—outpacing every technology that came before it.
The KeyNews take
Why it matters
As algorithmic efficiency gains compound with hardware improvements and competition, the cost-per-fixed-benchmark is plummeting. This reshapes cloud economics, on-prem ROI calculations, and which models teams can afford to deploy at scale.
The key facts
4 to knowEpoch AI: 13x annual price decline for fixed benchmark performance
MIT: 3x annual algorithmic progress after removing hardware and competition gains
Reasoning models cost more per task due to higher compute requirements
Price and performance no longer track together; quality, speed, error rate now drive model selection
Go to the source
The Decoderthe-decoder.com
Publisher excerpt: AI is hitting a fixed benchmark performance level at a rapidly falling cost. Epoch AI measures a price decline of about 13x per year. After stripping out hardware gains and competition, MIT puts annual algorithmic progress at about 3x. That doesn't mean today's best models are cheaper, though.…