Prediction: Models Will Grow Larger, Chips Won't Get Cheaper
teortaxesTex · x · 2026-08-20
A debate on the trajectory of model size and hardware costs:
- The author predicts models will get larger while chips won't get cheaper, making future inference (e.g., Qwen 6 20B) extremely expensive, and suggests undervolting/downclocking.
- A quoted view counters that models will get smaller and chips cheaper.
More from Infra
- GitHub Commits Surge to 1.4B/Month as Cursor Launches Origin Code Hosting — krishnan · 2026-08-20
- Colibri: Pure C engine runs 2.8T parameter MoE models on consumer hardware — tom_doerr · 2026-08-20
- Apple Silicon achieves ANE+GPU dual acceleration, boosting prefill by 50% — bakawolf123 · 2026-08-20
- NVIDIA: Vera Rubin Platform Delivers 10x Tokens/Second per Megawatt — nvidia · 2026-08-20
- Opinion: SpaceXAI emerges as the new No.3 frontier lab with 10GW compute ambitions — thedealdirector · 2026-08-20
- NVIDIA cuOpt becomes fastest open-source solver on Hans Mittelmann benchmarks — NVIDIAAI · 2026-08-20