AI infrastructure arms race: Google-SpaceX compute deal, Nvidia revenue-share, Microsoft Frontier, Meta Compute, Veritone-Oracle partnership
Key Questions
What is the Google-SpaceX compute deal?
Google signed a $920M per month agreement with SpaceX for 110,000 NVIDIA GPUs, totaling $30B, to support its Gemini Enterprise offerings. This deal intensifies competition in AI infrastructure among major tech players.
How is Meta expanding into the AI compute market?
Meta launched its Meta Compute cloud business to sell excess AI compute capacity, including through the Muse Spark API and a partnership with CoreWeave. This positions Meta as a direct competitor to AWS, Azure, and GCP in the cloud services space.
What is the Veritone-Oracle partnership about?
Veritone partnered with Oracle to achieve over 20% cost savings while processing 22.2 trillion tokens in its data refinery. The collaboration supports AI training data demands and emphasizes cost optimization in infrastructure.
Google signed $920M/month compute deal with SpaceX for 110,000 NVIDIA GPUs ($30B total) for Gemini Enterprise, intensifying infrastructure competition. Nvidia offers compute-for-revenue-share to startups and steps back from equity checks, funding emerging cloud providers with GPU buyback. Microsoft launched Frontier Company ($2.5B, 6,000 engineers) for enterprise AI deployment, competing with AWS's $1B FDE venture. Meta launched Meta Compute cloud business (stock surge, Muse Spark API, CoreWeave deal) to sell excess AI compute, directly competing with AWS, Azure, and GCP. New: Veritone partnered with Oracle for 20%+ cost savings, processing 22.2T tokens in its data refinery, signaling continued demand for AI training data and cost optimization. These moves escalate the AI infrastructure arms race and reshape cloud market dynamics.