News for hackers. Stuff that matters.
Re:YC's Garry Tan uses AI agents so much, he spends up to $2... by jorvi on Wednesday November 04, 2026 @10:39PM (#50374848)
You don't.

Once you factor in cost of electricity and depreciation of hardware, it is more expensive to run locally.

There are many arguments for running local but cost isn't one of them.

Re:YC's Garry Tan uses AI agents so much, he spends up to $2... by nodesocket on Wednesday November 04, 2026 @10:50PM (#50374930)
> but cost isn't one of them.

Let's do quick back of the napkin math.

  - Mac Studio M5 Ultra, 80-core GPU, 256GB memory, 2TB storage -- $11,299.00.
  - Power (385w continually 100% of month) is $40/mo (my regional) or $480 annually. Let's say power is 3x more expensive in California, which it probably is, that's still only $1,440 annually in power.
The Studio M5 Ultra has a worklife life of at least 3 years.
Re:YC's Garry Tan uses AI agents so much, he spends up to $2... by jcelerier on Thursday November 05, 2026 @06:44AM (#50378009)
it's going to be quite slower than remote providers, won't be able to run full models, only quantised, and for more cost than the highest Claude subscription tier where you can do roughly 60-100M token per month at 200$, e.g. 7200 for three years. It makes sense if you prioritize local AI but not really otherwise sadly.
Re:YC's Garry Tan uses AI agents so much, he spends up to $2... by klrefg on Friday November 06, 2026 @02:11AM (#50389151)
You have to compare with how much Deepseek costs since all the models you will be running will be inferior anyway, certainly not Opus or similar.
Re:YC's Garry Tan uses AI agents so much, he spends up to $2... by bitfilped on Thursday November 05, 2026 @05:47PM (#50385612)
You have to be using an incredibly small amount of usage for your math to math here. Go price out a few MW or better site then ask a neo cloud how much that same amount of compute will cost at an API.