Tech Industry
Yesterday
785
PM is irrelevant role and will die in next 2-3 years.
Tech Industry
3d
32566
What happens when most of your team is Indian?
Tech Industry
Yesterday
861
Is meta really worth the wait?
World Conflicts
Yesterday
631
Peaceful Protest Hasn’t Worked and Has Been Met With Aggression.
Work Visa
Yesterday
871
Hypocrisy of Indians
https://www.wsj.com/tech/google-expands-in-house-chip-efforts-in-costly-ai-battle-3121c852?st=6010gt7v4bee9ve&reflink=desktopwebshare_permalink Google investing in new Axion Chips to help them in the battle to dominate the AI market -- what are your bets on who will win the AI Race? Amazon? Microsoft? Google? Someone else? How do you think these chips affect the race in AI?
Why amazob?
Aws, investment into Claude, the ability to deliver ai to the home via Alexa, Amazons growing ad business
Amazon’s AI is like 5 years behind and invested in Claude later than Google 😂
Meta
Meta has custom chips?
@google last time I was there. They were developing their own chips.
I don't know who wins, if MS and Amazon follow suit, nvidia would lose out. All with titanium balls, short nvda.
I know everyone is working on it. Apart from Google, no one has announced yet. Guessing it would come within a year from others.
Amazon has had their own chip for years. Actually they have 2: one for training and one for inference.
Probably Google but not because of the chips, more because of huge access to data at scale and lots of great researchers and the ability to deploy things at global scale (somewhat) reliably
It’s bleeding researchers as culture is all time low and VCs are paying insane bubble valuations for startup ideas But. Compute and Data are way more important. Especially in LLM race where there’s barely any algo edge Google simply has the most data and the most compute Even with bad leadership and fleeing talent , it’s going to just come down to the simplest principles : compute and data
Does anyone really understand the performance improvements this chip makes? Just seems like Google is doing the same thing that Amazon did with their own ARM chips. Doesn’t seem that it’ll make a difference for specific workloads like say Nvidia Hopper CPUs which help with bandwidth between GPUs in a rack.
Exactly, this is just a general purpose ARM chip. Good for hopefully cutting your compute costs for a given workload, but not much more than that.
No it’s different for now. Google is ahead of Amazon and Nvidia to launch their own custom CPU based on Arm v9 ISA for infra cloud to get rid of the dependency on Arm. Amazon and Nvidia remain to rely on OTS CPU from Arm.
I think this article is phrased in a misleading way. My guess is that axiom chips are not specifically built for AI workloads, but rather just as a more power/cost efficient CPU for GCP & internal infra usage. Probably wanna diversify and get away from their Intel dependency
Yup journalists are idiots
Bingo
Meta definitely! They are developing their own chips too!
Lol, you all spent years developing Horizon Worlds. The world doesn’t want it and it sucks.
Sick burn Cory!!
Two guys in a garage
Every big player is going to have something. And it will be good enuf for most of their used cases
Damn Microsoft winning finally since 98