← The Log
AI

DeepSeek R1 shakes up the AI cost math

In January 2025, the lab DeepSeek released R1, an open-weight reasoning model that matched or approached the performance of far pricier systems. Within a week its app topped the US app charts, and the shock to assumptions about AI spending helped wipe hundreds of billions off chipmaker valuations, including a record single-day drop for one of them.

The headline was cost. DeepSeek reported training its underlying model for a small fraction of what leading labs have spent, and it published the weights so anyone can run or fine-tune the model themselves rather than renting it through an API.

For a small or mid-size business, the lesson isn't to chase the newest model, but to notice the direction: capable AI is getting cheaper and more portable, fast. That means you can wait for prices to fall, keep your tooling vendor-flexible, and avoid locking a whole workflow to one provider's pricing. Build for swap-ability, not loyalty.

Further reading: original source ↗

Want this handled for you?

This is exactly the kind of thing our AI integration & automation work takes off your plate. Start with a free, no-pressure look at your site.

Come aboard
Get the Log by email

Plain-English notes on web, marketing, and getting found online. Occasional, no spam.