Debate: replace GPUs with chips physically hardwired to run one specific model?
giffmana · x · 2026-11-02
A discussion about AI chip routes: one participant envisions replacing general-purpose GPUs entirely with speciality chips — physically integrated at the circuit level to run one specific model, e.g. a DeepSeek 7 13B chip that can only run that model. @giffmana's reply mocks the narrow assumption that GPUs exist only for LLM training. The exchange highlights the core trade-off in the GPU vs ASIC inference debate: peak inference efficiency versus flexibility and fast model iteration.
Related event: Debate flares over purpose-built inference chips replacing GPUs(3 posts)→
More from Infra
- Rumor: Musk to renegotiate TeraFab terms with TSMC, dropping self-build insistence — SumitGup · 2026-11-02
- Crusoe CEO says neocloud debt fears overblown amid 'tidal wave' of compute demand — nathanbenaich · 2026-11-02
- Kapitsa hits 346 tok/s on a 35B MoE and runs a 180B model on 48GB Macs via SSD offloading — DotaMate · 2026-11-02
- Texas Mega-Project Pipeline Tops $500B, Led by $100B Stargate AI Hub — MatthewChang · 2026-11-02
- SemiAnalysis: 30.3GW of planned 2028 US data center capacity at risk of pause after midterms — Eliv_nurotic · 2026-11-02
- B300 GPU rental prices plunge from $7.9/hr to $4.6/hr in a month as providers beg for customers — SumitGup · 2026-11-02