Raja Koduri: Western AI compute costs $50-60B per gigawatt, China targets under $10B
RajaXg · x · 2026-09-30
- Former Intel GPU chief and Oxmiq Labs founder Raja Koduri joined the TechSurge podcast to break down the $24T capital math behind AI compute.
- Key figure: each gigawatt of AI infrastructure costs $50-60 billion in the West, while China targets under $10 billion.
- Other topics: inference as a full memory hierarchy problem, spec now being 90% of chip design, a Steve Jobs decision story, and "token mills".
More from Infra
- EasyPPO: just fix the critic — stable PPO for LLM post-training with zero training collapse — teortaxesTex · 2026-09-30
- Why data centers can't just pump seawater: corrosion, salt fouling, and biofouling explained — AryHHAry · 2026-09-30
- Graphsub pitches in-memory graph DB for agent data: don't trust one AI corp with it all — arthurcolle · 2026-09-30
- Even if OpenAI and Anthropic blow up at IPO, going long GPU hardware still pays — GabGarrett · 2026-09-30
- GPU dominance in deep learning is largely accidental: the case for custom chiplet inference hardware — ai · 2026-09-30
- DeepSeek adds Huawei Ascend support across its open-source libraries — yacineMTB · 2026-09-30