Ben Bajarin says AI’s biggest miss was underestimating the GPU supply-chain tsunami
BenBajarin · x · 2026-07-28
Ben Bajarin revisits 2023–2024 forecasts and argues the biggest miss was underestimating how fast the entire accelerator stack had to scale together.
- He calls the new reality the “GPU Tsunami,” where power semis, analog, CoWoS packaging, switching, rack power, and data-center capex all expanded far faster than earlier models assumed.
- His current model puts the semiconductor market at $1.79T in 2026 and $3.49T in 2030, with adjacent AI infrastructure taking the combined 2030 silicon-systems pool to $3.95T.
- The 2027 accelerator model is now $720B, versus an old $125B endpoint, and one quarter of NVIDIA data-center revenue already reached about 60% of that former full-year endpoint.
- Other revisions include CoWoS capacity moving from 25k–30k wafers/month to 80k+, frontier rack power landing 3.0–3.6x above a prior 2030 expectation four years early, and a seven-company capex model now 2.4x an old global data-center endpoint.
More from Infra
- Kimi K3 tokenizer optimization cuts first-token latency by about 325 ms — philipkiely · 2026-07-28
- Palantir says its U.S. government AI platform is built on open weights and NVIDIA GPUs — eliano · 2026-07-28
- Kimi K3 lands day-one on Dell PowerEdge XE9780 for on-prem deployment — _akhaliq · 2026-07-28
- Claude Opus 5 leads benchmarks, but early users say it stops short on real work — The AI Daily Brief · 2026-07-28
- engy.ai launches verified inference API with cryptographic proof for open models like GLM-5.2 — markjeffrey · 2026-07-28
- Kimi K3 is reportedly running on 80 RTX 5090s with no HBM — markjeffrey · 2026-07-28