@theo Lmao delete this
The United States is no longer the best place to build an AI lab.
@theo Lmao delete this
The United States is no longer the best place to build an AI lab.
Positive users defend the US as still the best place for AI labs citing its capital and talent edge, while negative users cite tech layoffs and hiring collapses to mock the claim.
“America is no longer the best place to build an AI lab”
Lmfao good luck with that

“Amazon, Google, Meta, Microsoft, and Oracle collectively hold an estimated 71% of the world’s cumulative AI compute as of Q4 2025, measured in H100-equivalents of computing power. This is up from 63% in Q1 2024.”
+ only American labs have frontier models right now
+ we’re in early takeoff
Not hubris just calling the trajectory as I see it
Where is community notes when you need them
“America is no longer the best place to build an AI lab”
Lmfao good luck with that

@bayeslord This hubris is beyond asinine.

@bayeslord @grok how much computing power is used to train the latest big models and how many NPUs does it take

@bayeslord @HarmonyHacker seems like something that can change relatively quickly

@bayeslord listen, while you're correct right now, these things can change. belief that they can't is just wrong. And stuff like THIS is what can cause it to change.

@bayeslord Racks on train cars man.
Datacenters on rails bro.
Steam engine power dude!
Don’t gotta worry about all the local ordinances and community meetings!
Total Win for Ai Dude! Let’s get investors and IPO challenge that Space nerd!
Ai Datacenters on choooo chooooo

@billionsmustliv @bayeslord @HarmonyHacker Virtually every GPU is owned by an American company and are already under export laws that don't allow them to even use them overseas. What exactly do you think can change quickly?

@bayeslord It will be the only place

@bayeslord Wish people would just quote things instead of fucking vagueposting

@bayeslord I think America is still the best place and apparently with all human and financial capital they will keep their edge
But the problem in my mind is different

Frontier models (Grok-3, GPT-4.5, Claude 4 variants, Gemini 2.5+) use ~10^25 to 5×10^26 FLOPs.
This takes clusters of 50k–200k+ NVIDIA GPUs (H100/H200/Blackwell). xAI’s Colossus (100k–200k GPUs) trained Grok-3 at record scale.
“NPUs” are mostly for efficient inference on phones/edge devices. Large-scale training needs data-center GPUs with fast NVLink interconnects, not consumer NPUs.
US labs still dominate building these clusters.

@bayeslord seems like he forgot about the semiconductor export controls too

@bayeslord Microsoft: 15k layoffs. Meta: 8k cuts. Entry-level hiring collapsing. Laughing doesn't change the numbers.

@bayeslord Real