Swyx catches up with former guests and welcomes Eugene Cheah as a rare guest writer, who cofounded Featherless.AI, an inference platform with 2,000 open source models accessible via a single API for a flat rate. NVIDIA’s Blackwell series rollout to OpenAI has caused excitement, while cousin Lisa introduces MI3 25X and Cerebras files for IPO. Eugene shares insights on GPU economics and advises against buying H100s due to oversupply and better rental options. The rise of open weights models has shifted the market, with a decline in demand for foundation model training. Inference and fine-tuning are in demand, with alternatives to H100 available.
https://www.latent.space/p/gpu-bubble