Meta's long-term vision for AGI: 600,000 x NVIDIA H100-equivalent AI GPUs for future of AI

Meta is on a war path towards artificial general intelligence (AGI), where CEO Mark Zuckerberg teased in a new Instagram post that the social networking giant is betting big on AI and AI GPUs.

3

The company is currently training its next-gen model Llama 3, where Meta is building massive compute infrastructure to support their future roadmap, including 350,000 x NVIDIA H100 AI GPUs by the end of the year. Zuckerberg added that overall, almost 600,000 x H100-equivalents of compute will be ready, as they're also buying AMD's new Instinct MI300X AI accelerators, not just NVIDIA AI GPUs.

We should expect all this AI GPU power won't just be from NVIDIA's current-gen H100 AI GPU that's based on the Hopper GPU architecture, but NVIDIA's beefed-up H200 AI GPU and next-gen B100 AI GPU that's based on the next-gen Blackwell GPU architecture.

Dylan Patel, Chief Analyst from SemiAnalysis, explained that NVIDIA will be increasing the output of its Hopper-based AI GPUs, which is a big family now. NVIDIA's current-gen Hopper AI GPUs include the H100, H200, GH100, GH200, and H20. Patel notes that NVIDIA is boosting production from 773,000 in Q1 2024, to 811,000 in Q2 2024... close to 100,000 more AI GPUs made in three months.

• Read more: NVIDIA AI GPU shipments expected to surge 150% year-over-year in 2024

Zuckerberg said: "Some updates on our AI efforts. Our long term vision is to build general intelligence, open source it responsibly, and make it widely available so everyone can benefit. We're bringing our two major AI research efforts (FAIR and GenAI) closer together to support this".

3

He continued: "We're currently training our next-gen model Llama 3, and we're building massive compute infrastructure to support our future roadmap, including 350k H100s by the end of this year -- and overall almost 600k H100s equivalents of compute if you include other GPUs. Also really excited about our progress building new AI-centric computing devices like Ray Ban Meta smart glasses. Lots more to come soon".

Dylan Patel told Tom's Hardware: "While NVIDIA has capacity to continue to grow in the 2nd half, the demand modeling we have done indicates a fall off in Hopper demand as hyperscalers shift to looking towards Blackwell".