Hacker News new | ask | show | jobs
by whatevertrevor 168 days ago
> Groq built the region's largest inference cluster in eight days in December 2024. From that Dammam facility, GroqCloud serves "nearly four billion people regionally adjacent to the KSA." This isn't API access. This is critical AI infrastructure for a nation-state, funded by the Public Investment Fund, processing inference workloads at national scale.

Maybe I'm just completely out of touch, and hardware has never been my expertise, but does it take O(days) and not O(years) to build data centers these days? I know Grok DCs in Memphis were built under a year cutting many corners and using plenty loopholes, but even by those standards, bringing up a full data center in just over a week sounds impossible without some insane construction automation to me.
3 comments
tw04 168 days ago
Physically building a datacenter yes. Repurposing a crypto datacenter for a new purpose, no. But days is hyperbolic no matter how you look at it unless they’re reusing ALL the existing infrastructure (network/security/hvac/physical server racks/some amount of compute). And aren’t actually including the time to procure the inferencing chips.

You can call up Cisco tomorrow and offer them a billion dollars to get you an entire datacenter worth of switches tomorrow and the answer is going to be no because they just don’t keep that much inventory sitting around. That’s why covid was such a shitshow.

I should give a caveat of: they could in theory redirect existing orders to you but would likely be violating contractual obligations and risk a lawsuit from the fortune 500 Peter they robbed to pay you Mr. Paul.

link
whatevertrevor 168 days ago
"Smart" accounting by not counting design, planning, procurement timelines would explain it, that makes sense.

If that's the case, it's saddening to see details like this hyped up to borderline fraudulent levels.

link
burnerRhodov2 167 days ago
Well, you have different players now... working at a different scale.

They are purchasing crypto infrastructure for a huge premium, importing massive turbines from old generation facilities, making huge personal investments in nvidia stock for "preference", and acquiring companies that happen to have an order.

It went from we need to make a data center as cost efficient as possible to we need a data center at all costs.

link
lumost 168 days ago
they are likely referring to their hardware deployment. Which can be done in days. The datacenter is owned and operated by someone else and they just bring in the racks.
link