> It's fantastic news, because OpenAI and AMD will now work together to develop decent software libraries for AI on AMD chips.
Too fast to jump to conclusion. I'd say they will work together to develop software specifically designed for OpenAI products. It is a giant question mark whether we'll get libraries for general purpose computing out of this.
OpenAI and AMD both have strong incentives to chip away at Nvidia's dominant position, and the best way to do that is by releasing software tools to the public.
Ask yourself: What would first-class PyTorch/Triton and Jax support for AMD hardware do to Nvidia's dominance and pricing power?
That argument could apply to every company who build their business on AI (except Nvidia) but we haven't seen any notable initiative or any significant movement in the past 3 years.
I would not put my bets on private companies' goodwill. I'd rather believe they'll do whatever is most important for their business priority, understand things at their face value, and then hope for the best.
I disagree. My argument could not apply to other companies, because AMD is the only other maker of GPU hardware that could conceivably compete with Nvidia at scale in the AI market. Intel's GPUs have never been in the same league. New alternatives like Cerebras, etc. are too small, too different, and too unproven at scale.
I'm not betting on anyone's goodwill, I'm betting on OpenAI's and AMD's self-interest. Please don't attack a straw-man.
All Tech companies including OpenAI have initiatives for their own chips. Nobody talks about Intel.
Tech companies have the following options:
1. Buy Nvdia high performance stack with stable and fast support and deploy your SW developers to quickly get started
2. Buy AMD whatever stack and deploy your SW developers to make a lot of ground work
3. Develop and deploy your own chips and use your SW developers to make your SW from scratch but exactly as you need it
The only reason they might go with No. 2 is if AMD gives them the HW more or less for free but maybe even then will mix 2 + 3 if they don't want 1. And AMD deal is showing exactly this, AMD has to give free quity to get a customer for their HW.
What people don't get, who in their right mind would switch from one vendor lock-in to another with the difference on investing SW development into the 2nd???? The resources spend on AMD will bind customers to AMD. It doesn't matter if RoCm is open source as long as it runs only on AMD. If RoCm would run on any AI chip (including Nvidia) then we would have a case of an interesting switch but then the question comes up, why buy AMD if RoCm doesn't require it?
Too fast to jump to conclusion. I'd say they will work together to develop software specifically designed for OpenAI products. It is a giant question mark whether we'll get libraries for general purpose computing out of this.