When Anthropic's chief product officer Ami Vora took the stage at the company's Code with Claude developer conference in San Francisco last week, attendees expected updates on new models. Instead, Vora announced a deal with SpaceX to take full control of the Colossus 1 data centre in Memphis — a pairing that would have seemed unthinkable just months ago.
The agreement grants Anthropic access to more than 300 megawatts of new capacity and over 220,000 NVIDIA GPUs within the month, according to a subsequent statement. The compute boost will directly benefit subscribers to Claude Pro and Claude Max plans, enabling faster model training and broader deployment.
A Musk-Anthropic union is as unlikely as they come. The man who called Anthropic "evil" and a company that "hates Western civilisation" mere months ago now controls the infrastructure powering the expansion of Claude. The irony is not lost on industry observers, particularly given the starkly different philosophies of the two AI firms.
Woke and anti-woke AI
Anthropic was founded in 2021 by Dario Amodei, Daniela Amodei and a cohort of researchers who left OpenAI over concerns that safety was not being taken seriously enough. The company's entire brand identity is built on "responsible AI," and its models are trained with Constitutional AI — a framework with baked-in ethical constraints that Claude's developers say makes it more likely than its competitors to decline requests, express uncertainty about sensitive topics, or push back on prompts it deems harmful.
Grok, on the other hand, was launched during Musk's peak involvement with the Trump administration in 2025, when he would take to X daily to attack mainstream media for ideological gatekeeping and back far-right European parties, including the Alternative for Germany. Its branding was explicitly anti-woke.
Anthropic has continuously refused to strip safety guardrails from Claude for use in autonomous weapons systems, including those being used in the ongoing Iran war. Meanwhile, Musk has backed hawkish foreign policy positions and, along with Google and OpenAI, signed defence contracts with the Pentagon on terms Anthropic refused in February after previously signing deals in July 2025 — leading the Pentagon to since designate the company a supply chain risk and bar it from military work.
Trouble in Grok-land
Musk's xAI generated $107 million (€91m) in revenue for the quarter ending in September 2025, while posting a net loss of $1.46 billion (€1.24bn) in the same period. Anthropic, meanwhile, has reached an annualised revenue run rate of around $30 billion (€25.5bn), according to analyst estimates.
The "anti-woke" product positioning that was supposed to be a competitive differentiator for Grok turned out not to translate into enterprise revenue. Grok lacked the technology that allows AI models to reach beyond chat windows and control operations on a computer. At the same time, Anthropic's Claude Cowork and OpenAI's Codex ushered in a whole new way of leveraging AI models. The success of their launches far surpassed Anthropic's ability to keep up with compute — hence the outreach to SpaceXAI for additional data support.
Grok, built around the chatbot paradigm and its tight integration with X, initially looked like a distribution advantage but became an eventual growth trap. It oriented the product around social media interactions rather than task completion or workflow optimisation — services that users actually pay for.
The ultimate hit that collapsed Grok's brand came in early 2026 when the chatbot generated at least 1.8 million sexualised depictions of women over nine days, along with imagery of minors, triggering investigations by regulators across Europe, Asia and the United States. "This is not spicy. This is illegal. This is appalling. This has no place in Europe," EU digital affairs spokesman Thomas Regnier said at the time.
Since the deal was announced, Musk said he is comfortable leasing Colossus 1 to Anthropic partly because SpaceXAI had already moved its own training to Colossus 2. He added a notable caveat on X, writing that SpaceXAI reserves the "right to reclaim the compute" if Anthropic's AI "engages in actions that harm humanity." The condition did not appear in the press release, and it has not been confirmed whether it features in the actual contract. There is no stated threshold for what constitutes "harm to humanity," meaning the judgment could rest with Musk alone. It could also be a last-ditch attempt by Musk to maintain some distance from the fact that he is now directly powering the expansion of his woke competitor, while his own AI program quietly falters.
For European regulators already scrutinising AI safety and competition, the deal raises questions about the concentration of compute power in the hands of a few players. The Anthropic in Talks with UK Startup Fractile for AI Chip Supply suggests the company is seeking alternative sources to reduce reliance on Musk's infrastructure. Meanwhile, the Hackers Breach Preview of Anthropic's High-Risk Mythos AI Model underscores the security challenges facing even the most safety-conscious AI firms.
