Musk Bets $1.25T Merger on Space Data Centers, Anthropic's Legal Plugin Wipes $285B, & OpenAI Retires GPT-4o

Musk pairs rockets with AI in history's largest merger, Anthropic crashes legal stocks by launching workflow tools, and OpenAI follows through with their August decision

Musk Merges SpaceX & xAI, Betting on Data Centers in Space

SpaceX officially acquired xAI on Monday in a deal that values the combined entity at $1.25 trillion. It is the largest merger in history.

Bank valuation documents reviewed by CNBC put SpaceX between $859 billion and $1.26 trillion, and xAI between $219 billion and $294 billion. For context, OpenAI was last valued at $500 billion and Anthropic signed a term sheet last month at $350 billion.

The combined company brings rockets, Starlink, the X social network, and Grok AI under one roof. Tesla is not included, despite investing $2 billion in xAI just last month.

Elon’s stated rationale: orbital data centers. He wrote in a memo that “within 2 to 3 years, the lowest cost way to generate AI compute will be in space.” This would be thanks to unlimited solar power, no land constraints, and no grid bottlenecks. SpaceX has asked the FCC for authorization to launch up to one million satellites as part of an “orbital data center system.”

If you recall, we wrote two weeks ago about how AI data centers are bypassing the grid with gas turbines because the power infrastructure cannot keep up. Boom Supersonic is building jet engine turbines for data centers. OpenAI ordered 29 gas turbines for Stargate.

And xAI is burning approximately $1 billion per month for this build out, according to Bloomberg. SpaceX generated roughly $8 billion in profit on $15 to $16 billion of revenue in 2025. SpaceX’s profitable rocket business is now subsidizing xAI’s race to catch up with OpenAI, Google, and Anthropic.

And catching up is the right framing. Electrek called the deal a “bailout,” comparing it to Tesla’s 2016 SolarCity acquisition. The conflict of interest is obvious, but this ultimately happened because Elon valued xAI higher than new investors.

Elon was wrong with SolarCity, but he has been right before when the consensus was against him. Short sellers lost $40 billion betting against Tesla in 2020 alone.

How could Elon be proven right here? If SpaceX can build data centers in orbit far cheaper than on the ground, the expected burn for xAI is not as bad as it looks. CNN reported that SpaceX’s pitch relies on near-constant solar power with minimal operating costs, though Deutsche Bank estimates cost parity with terrestrial facilities is still well into the 2030s.

These orbital data centers could also let xAI become a CoreWeave of outer space, giving them a cost advantage no terrestrial competitor could match. CoreWeave built a $5 billion business renting GPU-packed data centers to AI labs. Doing that from orbit, powered by free solar energy, is a bull case.

This is a tall order, though. Grok remains a distant competitor to ChatGPT, Gemini, and Claude. xAI’s Colossus cluster, the 100,000-GPU facility we covered in January, is impressive infrastructure, but infrastructure alone does not close the model gap.

The merged company plans to IPO later this year, potentially raising as much as $50 billion in what could be the largest public offering in history. The question is how much SpaceX can monetize with this vertically integrated AI-space-internet platform.

---

Anthropic’s Legal Plugin Triggers a $285 Billion Selloff

Anthropic launched a legal productivity plugin for Claude Cowork on Tuesday. Within hours, it triggered a $285 billion rout across European software, legal data, and financial services stocks.

The plugin handles contract reviews, NDA triage, compliance checks, and templated legal responses. Anthropic emphasized that it “will not provide legal advice” and that all outputs require review by qualified professionals. It is a workflow tool, not a lawyer.

The market did not care about the disclaimer.

RELX, the parent company of LexisNexis, fell as much as 14.6%.

Wolters Kluwer dropped nearly 9%. Thomson Reuters fell roughly 10%. Experian lost over 7%. The London Stock Exchange Group fell 8.5%. A UBS basket of European stocks deemed vulnerable to AI disruption fell 4.9% to record lows.

This is the first time an AI lab’s product launch has wiped this much market cap from an entire industry sector in a single day.

The strategic signal matters more than the plugin itself. Anthropic is moving from model supplier to workflow owner. Until now, the AI labs have mostly competed on model quality and API pricing. This is Anthropic saying: we are coming for the application layer too.

Think about Anthropic’s moves over the past two weeks. MCP became the universal agent standard. Claude Cowork launched as their agentic desktop application. Now the legal plugin shows they are building vertical workflow tools on top of that platform.

This is the playbook: protocol layer (MCP), agent platform (Claude Cowork), then vertical applications (legal, and presumably more to come). Each layer makes the next one more valuable.

The legal tech incumbents charge premium subscriptions for access to case law databases and research tools. Anthropic is demonstrating that a general-purpose AI model trained on publicly available legal material can replicate a meaningful slice of those workflows.

But can Anthropic stay focused enough to capture this market? They’re valued at $350 billion, and RELX PLC only has $9 billion in revenue.

Investor uncertainty is reasonable, but if Anthropic out-executes these incumbents who should be all-in on legal, then they have every right to win this market.

---

OpenAI Retires GPT-4o, o4-mini & GPT-5 on February 13

OpenAI announced that it will retire GPT-4o, GPT-4.1, GPT-4.1 mini, o4-mini, GPT-5 Instant, and GPT-5 Thinking from ChatGPT. Everyone will use GPT-5.2 going forward, and the API is not affected for now.

Their justification: “the vast majority of usage has shifted to GPT-5.2, with only 0.1% of users still choosing GPT-4o each day.”

Keeping the API access makes sense, as OpenAI’s enterprise market share has already dropped from 50% to 34% as Anthropic and Google gain ground. Forcing a migration timeline on API customers would just accelerate their churn. It’s likely 4o’s API adoption usage is much higher than 0.1%.

OpenAI tried deprecating GPT-4o once before and reversed course after a user revolt. We covered the backlash in August: Plus subscribers saw their weekly advanced reasoning usage drop from 2,900 queries to just 200 after the new model router launched, and users described losing access to GPT-4o like losing a friend. We felt they made a mistake catering to the vocal minority: “with a billion users, upsetting just 1% of them is 10 million unhappy users.”

But OpenAI has been maintaining at least six consumer-facing models simultaneously. Each requires infrastructure, monitoring, and creates user confusion. Cutting that down to GPT-5.2 (plus o3 for reasoning) simplifies the product and reduces costs.

We wrote in our year-end review about how GPT-5 disappointed relative to expectations, and how GPT-5.1 launched without benchmarks or an API. GPT-5.2 is the version OpenAI is willing to bet the product on, and consolidating models arguably makes for a better user experience.

There’s no way you can keep everybody happy in consumer tech, and OpenAI committed to a decision they should have made back in August.