The situation in AI right now basically looks like this. Alon Musk is suing OpenAI and signs Enthropic as a customer. XAI folds into SpaceX and nine out of 12 XAI co-founders leave. Anthropic rents compute from Amazon who invests in OpenAI but Microsoft also breaks the deal with OpenAI but still holds 27% in shares and at the same time profits from Anthropic through cloud revenue.

On April 27th, Microsoft and OpenAI broke off their partnership. They've been together for many years, holding each other's hand, and boom, it's over. Why did it happen? To answer this question, we have to go back to October 2025. OpenAI completes its forprofit conversion. OpenAI and Microsoft go back a long way. Microsoft has been investing into OpenAI's research, but in exchange, Microsoft Azure was supposed to be the primary cloud provider to OpenAI. On top of this exclusivity, there was revenue sharing, intellectual property licensing, and this is important, the right to block OpenAI's forprofit conversion. This means that OpenAI could not restructure or execute IPO without Microsoft's approval. And the cherry on top of all of it is the 27% stake in OpenAI exclusive API hosting rights and the AGI clause.

The clause stated that if the moment when OpenAI declares AGI comes, Microsoft loses access to all of OpenAI's future models developed after that date. Wait, what? If OpenAI declares the AGI... Microsoft loses access to all of its future models that would supposedly be an order of magnitude better. Yeah, that one's been a weird clause from the start. Anyway, when AGI is achieved, OpenAI can unilaterally withhold model access from Microsoft. But in October, Microsoft added a condition to that clause saying that an independent expert panel would need to verify any AGI declaration. This condition strips OpenAI of the unilateral trigger. In exchange, OpenAI commits to purchase $250 billion of Azure services over the next 7 years. 6 months later, April 2026, the whole thing flips upside down. They break off the contract, but the 27% stake is still there. The AGI clause disappears. Microsoft stops paying revenue share to OpenAI. OpenAI's payments to Microsoft continue through 2030. And OpenAI is free to go to AWS or Google Cloud for cloud services.

In those 6 months, OpenAI went from hyperrowth rocket to we don't think we can pay the bills. And that crack happened inside the company. They missed their internal goal of 1 billion weekly active users for ChatGPT by end of 2025. They missed multiple revenue targets early this year, especially as Anthropic started eating its lunch in coding and enterprise. And they projected this year's losses of around 14 billion on 13 billion in revenue. This report by the Wall Street Journal dropped like a bomb and their own CFO Sarah Frier told the board that OpenAI has locked in $600 billion in future data center commitments. At the same time, Anthropic's Claude Code is blowing through the roof in coding and is actively beating OpenAI.

In February 2026, they sign a $38 billion deal with Amazon. The reason they signed with AWS is because they needed capacity to train new models and that doesn't conflict with the API clause. In March this year, Amazon invested $50 billion into OpenAI to be a cloud provider for OpenAI's enterprise agent management platform. Agent management is a specific category that doesn't trigger Microsoft exclusivity clause. Amazon's investment breaks into 15 billion immediately and $35 billion contingent on conditions, including an IPO or the achievement of AGI. This is a sign of OpenAI's desperation. The 15 billion upfront and 35 billion contingent split... is how you structure a deal when you know that the company needs the money badly. Amazon structured their investment so that they cannot lose. It pays $15 billion upfront and even if the remaining chunk never triggers, they already sold their chips to OpenAI and that is one of the largest chip adoptions they can get in today's market.

Now, while these three are having a party, Google is watching its biggest cloud rival lock in OpenAI. And so on April 24th, just 3 days before Microsoft OpenAI breakup, Google invests $40 billion in OpenAI's grave enemy, Anthropic. Why anthropic? Because by April 2026, OpenAI's cap table was essentially exhausted. Microsoft 27%, Amazon 8%, employees around 25%, and the nonprofit business about 26%. There was no room. Also, don't forget Google is already fighting the largest antitrust case in tech history. The Department of Justice proposed a remedy that included banning Google from AI investments, which means that investing in both anthropic and OpenAI... would be the clearest possible evidence of an attempt to control the entire AI model market.

Anthropic's explosive growth... Claude's demand is growing faster than any single infrastructure provider can physically keep up with. Google's TPU network and Amazon's Tranium clusters have ceilings and Anthropic is already pushing against them. So what happens when Anthropic needs more compute? It turns to Elon Musk. The same Elon Musk who is suing OpenAI in federal court who dissolved his own XAI company on May 6th and on the same day rented Colossus 1 to Anthropic.

XAI was founded in July 2023... raised $20 billion in January 2026 at $230 billion valuation. But under the surface, XAI burned $1 billion every single month... and 11 out of 12 original co-founders had left by March 2026. So, why merge SpaceX with an unprofitable business? Through this merger, SpaceX got Colossus 1, the Grock model ecosystem, X's 600 million user social media platform. Because SpaceX wants to go IPO at $1.5 trillion valuation. And this combo justifies the valuation because without XAI, SpaceX is a rocket company... But a rocket company plus AI plus satellite plus social media conglomerate might.

The losses of XAI got absorbed into SpaceX and cleaned the balance sheet for the IPO. SpaceX announced a deal giving Anthropic access to all of Colossus 1 capacity. Colossus 1 was built for XAI... to build an LLM that would hopefully beat Anthropic. So the irony that the data center built for Grock is being used to run Anthropic's model is quite sharp. An empty supercomputer on the balance sheet is a liability but a supercomputer contracted to the world's fastest growing AI company is a revenue generating asset. Anthropic went from 1 billion to 30 billion annual recurring revenue in 15 months. Salesforce took 15 years to hit 10 billion ARR. Anthropic did triple of that in 15 months. From Anthropic side... Colossus 1 gives them immediate compute capacity and now they're partnered with a rocket company, the only rocket company that shows a plan on how to build the first AI data center in space.

What is happening right now is that we're entering the race to IPO for Frontier AI companies. And they all know that the window is open for roughly 18 months. SpaceX is targeting June 2026 at $1.5 trillion valuation. Anthropic's IPO target is October 2026 at 900 billion in valuation. And OpenAI target Q4 2026, slipping toward 2027.

OpenAI situation is more complicated because they are on a harsher timeline. Amazon's 35 billion expires the last day of 2028. If they don't IPO by then, they won't get the money and therefore they're not going to have the money to cover the insane compute obligations. If OpenAI's IPO slips to 2027 and Anthropic's IPO happens in October this year at $900 billion valuation, the market narrative will flip gloriously. Anthropic will become the star, the undeniable star that will be very difficult to overthrow. This is why OpenAI is doing all these moves to get to that IPO as soon as they possibly can because they're losing the market to Anthroic.