Meta has released their AI model Llama 3.1 as open source which has a lot of implications for companies the board follows like Nvidia. The model is offered for free to be downloaded and it’s already a competitive model to GPT-4o from OpenAI. The Llama 3.1 405B model has increased to 128,000 tokens, or nearly 400 pages of text possible to input or output.
Some interesting bits from the press release related to companies,
Beyond releasing these models, we’re working with a range of companies to grow the broader ecosystem. Amazon, Databricks, and NVIDIA are launching full suites of services to support developers fine-tuning and distilling their own models. Innovators like Groq have built low-latency, low-cost inference serving for all the new models. The models will be available on all major clouds including AWS, Azure, Google, Oracle, and more. Companies like Scale.AI, Dell, Deloitte, and others are ready to help enterprises adopt Llama and train custom models with their own data. As the community grows and more companies develop new services, we can collectively make Llama the industry standard and bring the benefits of AI to everyone.
Wanted to update that I see this move to open source AI from Meta as potentially beneficial to Supermicro. Meta is Supermicro’s largest customer and two quarters ago they accounted for 26% of revenue.
I recently learned from another podcast that Meta has very strict guidelines about what type of hardware is allowed in their data centers. One of the main specifications is that absolutely no proprietary software or hardware components are allowed, because they don’t want vendor lock-in. Supermicro has passed this requirement to be open and configurable, and that includes their liquid cooling solution.
Meta with this move to open source AI has also committed to investing heavily in the creation of Llama 4 and Llama 5 in a multi-year build out. This shows that Meta’s commitment to AI is likely to lead to multi-years worth of purchases from Supermicro.
The CEO Charles Liang had been talking about a customer in a CNBC interview where the customer was concerned if Supermicro could deliver at scale. Liang confirmed they indeed could meet the demand. I believe this is referring to a conversation between Mark Zuckerberg and Charles Liang, although Zuckerberg was never mentioned by name.
Additionally, Supermicro is expanding their San Jose manufacturing facility, which is right next to Meta HQ so they are likely get the fastest time to market of any customer as well.
When I was building personal computers Supermicro was one of the big vendors that I used. I think they will also be a beneficiary of the PC upgrade also.
