Tokens per Watt Decides Your 2026 GPU and Cooling
Inference is now 80-90% of AI compute, and tokens per watt, not FLOPS, picks your GPU and cooling in 2026. Here is the math that survives finance.
Read more3 articles on AI. Practical notes on infrastructure, security, and engineering leadership by Indra Gusti Prasetya.
Inference is now 80-90% of AI compute, and tokens per watt, not FLOPS, picks your GPU and cooling in 2026. Here is the math that survives finance.
Read moreHyperscalers will spend $650B on AI infrastructure in 2026, but 7 GW of announced capacity sits stranded behind transformer shortages and grid queues.
Read moreChrome 149's origin trial ships WebMCP, the Google, Microsoft API that lets sites declare callable tools to AI agents instead of being screenshotted and guessed at.
Read moreDeep dives into infrastructure, security, and technical leadership. No noise, just engineering rigor. Subscribe and grab the 2026 AI-agent & infrastructure security checklist.