AI profitability is mathematically impossible

Post by jv22222 » Tue, Jun 09, 2026, 8:21 PM UTC

unknown_user_84 wrote: post #13

jv22222 wrote: post #6

Why is no one talking about open source models being burned direct to chip and running inference at 10k-15k a second? OS models close the gap (via distillation) with frontier models, then get burned to chip, then offer commoditized inference via data farms or local plugins. With thought loops this fast even if the models are less smart they can be self correcting to level them selves up.

I've seen Taalas come up on HN, but only once. I'm not a HN fanatic but I end up around page 3 before I kill-filp my browser into oblivion. Currently Taalas has Llama 3.1 8B burned onto a chip and offer chatbot and API access. That said they aren't selling the chips yet, on their website at least. I expect they are waiting for an openweight model they really feel is worth burning to a chip and/or training their own t…

Post by jv22222 » Tue, Jun 09, 2026, 8:25 PM UTC

saulpw wrote: post #7

jv22222 wrote: post #6

Why is no one talking about open source models being burned direct to chip and running inference at 10k-15k a second? OS models close the gap (via distillation) with frontier models, then get burned to chip, then offer commoditized inference via data farms or local plugins. With thought loops this fast even if the models are less smart they can be self correcting to level them selves up.

> With thought loops this fast even if the models are less smart they can be self correcting to level them selves up. You can't self-correct a model that's been burned to a chip. That seems like it'd be the main problem with ASIC AI, when everything's changing on a monthly basis, do you want to spend $x00 on a substandard model that'll be obsolete in 3 months, or wait 3 months?

Post by aesthesia » Tue, Jun 09, 2026, 9:08 PM UTC

anamax wrote: post #18

aesthesia wrote: post #11

There are some glaring local errors that make this analysis less than trustworthy. For instance, an assumption that corporate income tax applies directly to revenue, or a supposedly generous assumption that GPUs will fully depreciate after 3 years (6-year-old A100s are still in very high demand!). I would love to read a really well thought through investigation of inference costs and how they relate to token pricing,…

> GPUs will fully depreciate after 3 years (6-year-old A100s are still in very high demand!) Depreciation is a tax thing. While it is supposed to track useful life, it almost never does. For example, houses are depreciated on a 28-year schedule. I'm typing this from a house built in 1902.... Google has yet to decommission any of its Trilliums, and the V1s shipped in 2015. The prices to rent V2 (2017) and later are on…

Post by Tryk » Wed, Jun 10, 2026, 7:52 AM UTC

aatd86 wrote: post #26

u1hcw9nx wrote: post #5

Earlier quoted context omitted.

Classic story tellers vs people who can quantify. Story tellers: Full self driving was commonplace already in 2020.

No one claimed that. Besides negativity is a self-fulfilling prediction. And investing is not accounting...

Post by sameAsItEvrwas » Wed, Jun 10, 2026, 5:53 PM UTC

david_shi wrote: post #8

jv22222 wrote: post #6

Why is no one talking about open source models being burned direct to chip and running inference at 10k-15k a second? OS models close the gap (via distillation) with frontier models, then get burned to chip, then offer commoditized inference via data farms or local plugins. With thought loops this fast even if the models are less smart they can be self correcting to level them selves up.

If you burn a model to a chip what happens if there's a better model?