We are in a pretty amazing situation. If you're willing to go down 10% in benchmark scores, you easily 25% your costs. Now with Deepseek 3.2 another shot across the bow.

But if the ML, if SOTA intelligence becomes basically a price war, won't that mean that Google (and OpenAI and Microsoft and any other big model) lose big? Especially Google, as the margin even Google cloud (famously a lot lower than Google's other businesses) requires to survive has got to be sizeable.

Google trains its own AI with TPU's, which are designed in house. Google doesn't have to pay retail rates for Nvidia GPUs, like other hyperscalers in the AI rat race. Therefore, Google trains its AI for cheaper than everyone else. I think everyone else "loses big" other than Google.