Re: Ballot Option: Allow public AI for translations

Hi,

On Thu, Jul 30, 2026 at 11:32:11AM +0200, Aigars Mahinovs wrote:
> Any PC with a video card like AMD XTX 7900 or similar could run quite a few
> local AI payloads with decent speed. Finding a few of those for Debian
> infra should not be a blocking problem. Additionally there are many hosting
> providers all over the world offering AI-capable hardware to be rented on
> hourly basis for as low as 0.37€/hour in a fully green datacenter
> https://www.hetzner.com/dedicated-rootserver/matrix-gpu/ (as example only,
> no endorsement intended). And Debian has also not asked for any hardware or
> time donations from any providers yet. Historically Debian has been able to
> acquire needed compute power with little or no cost, when needs were
> clearly articulated and expressed.

It has become harder more recently, given a push from universities to
centralize their infrastructure or to move it to the clouds. Even with
hosting providers, negotiations became harder because of the price
increases to power and hardware cost.

Ironically the easiest would probably the hyperscalers, as for them the
ask is usually very small - but that is presumably what we'd want to
avoid.

I'd also say that we'd be poorly equipped to maintain a GPU cluster.
That said, I could still see LLMs become critical to certain workflows
(e.g. some batch processing), so having two machines with a GPU might be
fine. Just don't expect this to scale to regular development use cases.
And we'd need to find a place with sufficient power - this was
historically the biggest constraint we had.

> Providers of fronteer models would also likely jump at the opportunity of
> providing complimentary pro access to Debian Developers for project work,
> if we ever actually asked for that.

I'd also imagine this to be true, but this would not solve most of the
problems expressed in this thread.

Kind regards
Philipp Kern