expectedwrong hindsight

Two Cents Per Million

Google is doing several things at once, none of them accidental.

1 min read 215 words #google #gemini #ai-pricing #llm
hindsight — nailed it

Google's infrastructure pricing play continued exactly as described. Token costs kept dropping. Gmail AI search shipped. The strategy of making the underlying infrastructure too cheap to compete with was the right read.

Three things happened today, and they are probably the same thing.

First: Gmail now does conversational search. You can ask it things like a person and it answers. This sounds like a minor productivity feature until you use it and realize your inbox is no longer a place where information goes to die.

Second: Gemini Flash token caching is priced at $0.02 per million tokens. Not $0.02 per thousand. Per million. That number is not a typo and it is not a promotional rate — it is what Google decided this costs.

Third: Tim Brooks posted something from a Google account.

The war framing is corny but it's also correct. What Google is doing is not competing on benchmarks or shipping demo videos that look like they were produced by a company that has seen the sun. They are making the underlying infrastructure so cheap that building on anything else requires a justification. And they are stuffing AI into products that a billion people open every morning before they're fully awake.

At $0.02 per million cached tokens, the pitch isn't "our model is smarter." The pitch is: at this price, the cost of not using us is the interesting number.

Nobody planned this. Or, someone did, and they've been quiet about it for a while.