The 2017 whitepaper predicted AI demand would outrun GPU supply
"Greg Osuri | Founder of Akash Network: Liberty, Community, and the Future of Decentralized AI" (Beacon Podcast)
“We actually talked about machine learning in the White Paper. There was a whole page on it. We talked about how machine learning is only getting expensive… it was very obvious to us, like, that GPUs being the core for AI will be, and the demand for AI as it’s going up, GPUs will be in short supply.” — 00:16:48
Context: Retrospective claim (spans into the 00:17:30 block) that the GPU-scarcity thesis dates to 2017; he attributes the bottleneck to TSMC/ASML supply chains, noting fabs take 18 months to 2 years per new process (00:22:56).