Elon Musk folded his model-building ambition into a new business: becoming AI’s hardware landlord. After failing to match frontier rivals with xAI’s Grok, Musk spent a week with the Anthropic team and concluded he’d rather power them than compete. He leased his 220,000-GPU Colossus 1 cluster to Anthropic, ending the GPU famine that had capped Claude Pro usage throughout 2024. According to Nathaniel Whittemore, the move immediately doubled user rate limits and signaled Musk’s pivot from chasing OpenAI to competing directly with AWS and Azure for compute dominance.
On The AI Daily Brief, Whittemore explained this resolves a mutual crisis. Anthropic had best-in-class coding models but no scale; Musk had unprecedented scale but a mediocre model. The alliance lets Anthropic focus on software while Musk’s SpaceX provides the physical backbone. Musk reportedly justified the deal by saying no one at Anthropic set off his “evil detector.” He’s now converting electrons to tokens faster than anyone, generating billions in revenue to fund his own model training on the newer Colossus 2 cluster.
“Elon Musk is moving from a ‘model builder’ phase into a ‘compute kingmaker’ role similar to Nvidia’s Jensen Huang.”
- Nathaniel Whittemore, The AI Daily Brief
Anthropic’s explosive growth made the compute crunch acute. CEO Dario Amodei revealed the company saw an 80x annualized growth rate in revenue and usage this year - far exceeding its planned 10x annual target. This velocity, combined with Musk’s hardware, creates a formidable force. On All-In, David Sacks cited Anthropic’s revenue tripling from $10B ARR in January to $30B ARR by March, then hitting $44B ARR in April. He argues that trajectory could make Anthropic the most powerful monopoly in history, using safety rhetoric as a potential tool for regulatory capture.
Anthropic is channeling that scale into building agentic infrastructure, not just bigger models. At its Code with Claude event, the company skipped a flagship model release to showcase “harnesses” like Dreaming - a scheduled memory review that lets agents learn from past sessions - and Outcomes, which uses a separate grading agent to score work against user-defined rubrics. Internal tests showed Outcomes improved file generation quality by 8.4% for Word documents and 10.1% for PowerPoint slides. Boris Cherny, creator of Claude Code, stated there is no manually written code left at Anthropic; Claude agents coordinate via Slack, code in loops, and resolve issues autonomously.
The deal reshapes the competitive landscape. Musk is betting his engineering speed and access to power gives him a structural advantage over cloud providers. As Chamath Palihapitiya predicted on The AI Daily Brief, power and data center constraints are forcing these “trade-back” deals where software leaders rent hardware from specialists. For Musk, it’s a pivot to what he does best: solving hard physical problems at scale. For Anthropic, it’s the runway to chase “infinite” context windows and multi-agent coordination - shifting from a chatbot that answers questions to a system that functions like a reliable, evolving employee.
“Anthropic uses safety rhetoric for regulatory capture. This strategy could lead to a $1 trillion revenue run rate by 2027.”
- David Sacks, All-In
The collaboration underscores a broader industry shift where access to compute is becoming a more decisive moat than algorithmic breakthroughs. While OpenAI and Google chase the next model release, Musk and Anthropic are building the infrastructure to turn those models into persistent, scalable workforces. The question is whether this hardware-for-software trade makes Musk the industry’s indispensable kingmaker - or just a temporary landlord until someone builds a bigger warehouse.
