Frontier models + infra/power: surging demands, DOE funding [developing]
Key Questions
What power demands are projected for AI data centers?
Goldman Sachs forecasts 220% growth in global AI/data center power demand by 2030. US surge hits 50%, with projects like Abilene Christian data center.
What DOE funding supports AI infrastructure?
DOE allocated $294M via Genesis Mission for AI-for-science projects. This aids surging demands alongside Meta's OSS models post-LLaMA4.
What partnerships fuel Anthropic's frontier models?
Anthropic expanded with Google/Broadcom for multi-GW TPU capacity. MS-NV-Anthropic $30B deal provides 1GW compute.
What open-source models is Meta planning?
Meta plans to open-source next-gen models like Gemma4 after LLaMA4, amid hybrid OSS push led by Alexandr Wang influences.
What investments back CoreWeave and G42?
CoreWeave raised $8.5B; G42 valued at $1.4T. These fund frontier infra amid Qwen and other model surges.
How is AI power need addressed in offices?
Offices could repurpose for AI power via AC and infrastructure tweaks, easing data center strains.
What UAE-US AI partnership developments?
U.S.-UAE AI partnership launched first meeting; OpenAI's $30B Abu Dhabi center faces Iran threats.
What funding records did OpenAI achieve?
OpenAI raised $122B, backed by Amazon, Nvidia, SoftBank for compute expansion. Microsoft-NVIDIA-Anthropic Ignite 2025 partnership unites giants.
Meta Alexandr Wang new OSS/hybrid models post-LLaMA4; Anthropic Google/Broadcom multi-GW TPU; Goldman 220% power 2030; DOE $294M; Meta OSS; 50% US surge/Abilene; MS-NV-Anthropic $30B/1GW; G42 $1.4T; Gemma4/Qwen; CoreWeave $8.5B.