- Joe@discuss.tchncs.deEnglish24 days
There is also a commercial aspect…
Bigger models are more expensive to train and serve…
Inference is currently insanely profitable if you have the hardware and the automation in place to support and serve it. At that point, it’s a money printing machine, and you want to squeeze as much out of it as you can.
While training new models is extremely expensive, and serving them probably makes less profit (at least initially).
Having an external brake applied to the frontier labs is likely good for their bottom line, while increasing hype and directing customers’ annoyance away from them.
It’s likely only a temporary benefit, though. The dragon will catch up and apply more pressure, both on inference price and capabilities.
- Dran@lemmy.worldEnglish23 days
Can you cite your source on the claim that “inference is currently insanely profitable”? Everything I read suggests that openai and anthropic lose money on their plans.
- Joe@discuss.tchncs.deEnglish23 days
My caveats were clearly stated… After capital expenditure, it’s just operational costs, where electricity & cooling are the big ones.
At that point, it is insanely profitable to serve. The cheap API prices on open weights models hints at the profit margins involved in the US (the frontier labs and hyperscalers don’t open their books for us), unsurprisingly)
Therefore, the longer they can serve existing and lower cost models at the current rates, the better for their bottom line. It’s just common sense in business.
It doesn’t mean the company as a whole is profitable. I expect we’ll see turmoil in the coming months and years, and the prize will be compute capacity, with electricity & cooling options.
- Dran@lemmy.worldEnglish22 days
You can’t just write off capital expenditure though. The hardware, even for “effecient” MOE inference is still very expensive to buy, house, run, and cool. Even assuming open-weight model serving at $0 r&d for the models themselves, mixing high-prefill workloads doesn’t batch well with decode heavy concurrency (or other prefill-heavy jobs). The moment you do anything nontrivial you start running into very complicated architectural problems to efficiently solve at scale.
Hardware that is useful for 5-10 years at most, plus development and support for the inference workflows, doesn’t leave a lot of margin on the table.
My gut, along with basically everything I read, suggests that not most (even pure inference) shops are not profitable and are still floating on loans or vc money.
- Joe@discuss.tchncs.deEnglish22 days
If you assume they are unprofitable, the Q only becomes whether they are more or less unprofitable by serving the older models for longer.