Musk proved to the world with 220,000 GPUs: the battle over large language models has long moved beyond theoretical debates about algorithms and parameters.

Author and source: 0x9999in1, ME News

TL;DR

• Institutional realignment: In May 2026, xAI was officially absorbed internally by SpaceX and restructured into a new subsidiary, SpaceXAI. The large language model developer has formally transitioned into a “space cloud infrastructure provider.”
• Massive compute infusion: SpaceXAI is leasing its entire 300-megawatt compute capacity at Colossus 1, a supercomputing center in Memphis housing 220,000 NVIDIA GPUs (H100/H200/GB200), to former rival Anthropic.
• Enemies shake hands: After passing the "Evil Detector" test, Musk opens his computing resources to Anthropic to counter the monopoly alliance of OpenAI and Microsoft.
• The code nuclear explosion: Claude Code, now granted unlimited firepower, instantly lifts computational restrictions—Pro and Max accounts are freed from peak-time limits. AI agents will completely transform software engineering, marking a technological inflection point for the "one-person company" model.
• Interstellar Cloud Computing: Both parties have officially launched exploration of "Orbital AI Computing," aiming to use Starship to deliver gigawatt-scale data centers into space, thereby addressing the critical crises of power and cooling shortages on Earth.

The board is overturned: Silicon Valley no longer believes in tears, only in computing power.

In Memphis in May, the air was thick with tension.

It’s not the hot weather. It’s the colossal machine named Colossus 1, relentlessly consuming electricity at a terrifying power of 300 megawatts. Two hundred twenty thousand NVIDIA GPUs. What does that mean? It’s the most ferocious silicon-based matrix in human history—a computing behemoth capable of reshaping the Big Bang in the digital universe.

But now, this beast has a new owner. Or rather, it has gained a new "rider."

On May 6, 2026, Musk dropped a bombshell: xAI, once regarded as the elite unit countering OpenAI, was quietly dissolved as an independent company. No farewell, no mourning—xAI was seamlessly absorbed into SpaceX’s vast structure, becoming a new internal code name: SpaceXAI.

Immediately afterward, an even more fantastical plot unfolded.

SpaceXAI immediately signed a major agreement with Anthropic, handing over the entire Colossus 1 supercomputing center—built with immense effort and powered by mobile gas turbines and Tesla Megapack battery systems—to the creators of Claude.

Do you understand?

Elon Musk personally buried xAI’s independent operations and handed the sharpest sword to Sam Altman’s fiercest rival. This is not merely a business partnership—it’s a bloody power realignment at the heart of Silicon Valley.

The facts are unequivocal. No exaggeration, no fabrication. From the Financial Times to xAI’s official announcement, it is clearly stated in black and white. Musk casually remarked on the X platform: “I spoke with Anthropic’s leadership, and they did not trigger my ‘evil detector.’”

Behind the casual remark lies deadly danger.

Why? Why SpaceX? Why Anthropic? What kind of future can 220,000 GPUs really build?

Let’s peel back the layers of this century-long deal, like peeling an onion.

Unraveling the Inner Game: Why Must xAI “Die”?

To understand SpaceXAI, you must first account for xAI.

From 2024 to 2025, xAI was at the height of its glory. It built Colossus in just 122 days, then doubled its GPU count to over 200,000 in another 92 days—so fast it left Silicon Valley breathless. The Grok model also surged forward, even attempting to outpace ChatGPT in reasoning and rebellious spirit.

But reality is harsh.

The AI model business is incredibly saturated. Despite Grok being trained on X platform’s data, its market penetration among end consumers has consistently struggled to compete head-on with the first-mover ChatGPT or the safety-focused, gentle Claude. Rumors suggest that xAI, despite owning hundreds of thousands of GPUs, has at times seen alarmingly low utilization rates.

What kind of shrewd person is Musk? How could he possibly tolerate idle resources?

Since a direct frontal assault on the model layer is too costly, just flip the table and strike from a lower dimension. If you can't beat the water sellers, monopolize the water source instead. The "death" of xAI is essentially a highly pragmatic escape strategy.

Incorporating it into SpaceX is a real masterstroke.

Why? Because the Earth can't take it anymore.

When you have 220,000 GPUs, and even plan for a million in the future, your greatest adversary is no longer code—but the laws of physics. It’s electricity. It’s water. It’s land. Colossus 1 in Memphis has already been forced to deploy gas turbines to generate power. What does this indicate? It shows that the world’s traditional power grids can no longer support these energy-hungry AI giants.

Turning xAI into SpaceXAI would allow Musk to unlock his full potential.

He is no longer just a model developer—he has transformed into an “interstellar cloud computing infrastructure provider.” Building data centers on Earth too difficult? Then let’s build them in space! Using Starship’s massive payload capacity, deploy servers directly into low Earth orbit. What’s out there in space? Natural vacuum cooling at over minus 200 degrees Celsius, and unobstructed, 24/7 pure solar energy.

This is a cognitive downgrade attack. While OpenAI is still on Earth, struggling with governments over hundreds of megawatts of grid allocation, SpaceXAI has already set its sights on a gigawatt-scale orbital computing network.

So, xAI is not dead. It has simply shed its skin and grown lungs capable of breathing in a vacuum.

The handshake of rivals: Anthropic's hunger and Musk's calculations

The stage is set. SpaceXAI has even moved its training tasks to the more advanced "Colossus 2."

So, who will use the newly vacated Colossus 1?

The answer is Anthropic. This will undoubtedly be the most dramatic moment of 2026.

Recall how Anthropic came about: it was founded by idealists who left in protest over OpenAI’s shift toward commercialization and closed-source development. As for Musk, he even sued OpenAI and Sam Altman, accusing them of betraying their original mission to benefit humanity and turning into a profit-driven machine worth $85 billion.

The enemy of my enemy is my friend.

Anthropic is desperately hungry for compute. They need it badly. The Claude models have an excellent reputation, especially in logical reasoning and long-text processing, and are revered by countless professional developers. But what Anthropic lacks is the underlying raw power. To keep up with scale, they’ve previously had to juggle between Amazon AWS and Google Cloud, even committing to tens of billions of dollars in infrastructure investment.

But those traditional cloud providers cannot offer Anthropic what it needs most: instant, concentrated, pure computational power.

At that moment, Musk pushed open the door.

300 megawatts. 220,000 NVIDIA GPUs. Available for connection now, immediately, and without delay.

This is an irresistible temptation. For Anthropic, it’s not just about saving money—it’s about gaining time. In the life-or-death race toward AGI, securing a cluster of 200,000 GPUs a month earlier means digging a ten-meter-deeper moat around the next-generation model.

What does Musk stand to gain?

First, monetize your money and resources. If you don’t need them right away, why not rent them out to others to share the high costs of depreciation and electricity? Just a few days ago, SpaceX applied the same logic by leasing computing power to the AI programming startup Cursor.

Secondly, there is balance. Supporting Anthropic is like delivering a decisive strike to OpenAI’s flank. As long as Claude remains strong enough, OpenAI can never comfortably monopolize the market, and Microsoft’s Azure cloud business will remain under constant pressure.

Finally, and most concerning: ecosystem lock-in.

What does the agreement say? Anthropic not only leased supercomputers on Earth, but both parties explicitly expressed interest in jointly developing "orbital AI computing." Musk is using the existing 220,000 GPUs as bait to firmly tie Anthropic to SpaceX’s interstellar endeavor.

In the future, when Claude becomes the world’s best AI brain, its underlying infrastructure will float in low Earth orbit built by SpaceX. This is the kind of open strategy only Musk could devise.

Violent Aesthetics: Claude Code’s Nuclear Explosion and the Arrival of the “One-Person Company”

Let’s refocus on the impact of this transaction on the technology itself.

With 220,000 GPUs fully unleashed against Anthropic, what will happen? The first to be disrupted will be programmers' jobs.

On the same day the announcement was made, Anthropic held its first developer conference in San Francisco, where they directly declared: thanks to the massive computing power provided by SpaceX, the rate limits for Claude Code have been doubled! The peak usage caps for Pro and Max accounts have been completely removed! API request limits for the Opus model have been significantly relaxed!

This isn't squeezing toothpaste. This is blowing up the dam gates outright.

In the past, what was our biggest pain point when using AI-assisted programming? Context would break, AI would hallucinate with slightly more complex architectures, and every deep code refactoring would hit the API rate limit.

What about Claude Code now?

Imagine an tireless super-architect. When you hand it an ancient codebase with millions of lines of code, it doesn’t need to read it in batches—it swallows it instantly. With Colossus 1 powering it, its backend can spin up thousands of inference threads simultaneously.

You only need to provide the high-level business logic, and Claude can complete the underlying database schema design, middleware configuration, frontend component library setup, and even automatically write and run all unit tests within seconds.

This is no longer a "coding assistant." This is a "silicon-based outsourcing team."

Sharp observers should have realized that this marks a complete break from traditional software engineering paradigms. The much-discussed "One Person Company" (OPC) model has, at this moment, become physically feasible.

What is an OPC? It’s a super individual who leverages deep industry expertise to direct a team of AI agents in accomplishing tasks that once required a hundred-person team.

Before Colossus 1, OPC was merely a hopeful vision. Due to AI’s computational bottlenecks, individuals still had to spend significant time “fine-tuning” AI outputs. But now, quantity has transformed into quality. When computational power becomes sufficient to enable large-scale Chain-of-Thought searches and Monte Carlo tree simulations, Claude’s code generation accuracy will approach, and even surpass, that of senior human engineers.

Developers only need to handle the "vibe" — AI takes care of all the remaining coding.

From restructuring e-commerce platforms, to optimizing high-frequency trading systems for low latency, to the underlying architecture of complex 3D rendering engines—with infinite computing power, Claude will become the external brain for every superindividual. Behind it all, the 300-megawatt gas turbine in Memphis rumbles on.

Ultraman's Sleepless Night: Big Model Three Kingdoms Battle and Dimensional Reduction Strike

This merger directly reshaped the technology landscape in the second half of 2026.

If the previous AI battlefield was a contest among many powers, the situation is now unmistakably clear: a stable triangular dominance has emerged.

The first pole is the traditional powerhouse alliance of OpenAI and Microsoft. They possess the earliest technological advantages, the largest enterprise distribution channels (Windows/Office), and the deepest financial resources.

The second pole is Google, the village's last hope. With its self-developed TPU advantages, DeepMind's deep expertise, and dominance in the Android ecosystem, Google remains an indispensable giant at the table.

And now, the third pole has firmly taken root: the "Interstellar Rebel Alliance" between SpaceXAI and Anthropic.

The frightening aspect of this new alliance is that it plays by no conventional rules.

While Microsoft was still buying land and nuclear power plants across Earth in an attempt to secure energy for Azure, SpaceX is already preparing to load servers into Starship and launch them into space. This is a game-changing advantage.

This has not only left Sam Altman deeply uneasy but also sent shivers through the entire traditional cloud computing market.

If SpaceXAI truly leverages this extreme physical advantage to successfully implement a business model of "interstellar cloud services for AI training/inference," what challenges would AWS, Azure, and Google Cloud face?

Keep in mind that SpaceX’s current valuation is nearing $1.25 trillion, and the market is abuzz with rumors of its upcoming IPO in 2026. At this critical juncture, acquiring xAI, integrating the Colossus supercomputer, and partnering with Anthropic is no casual move—it’s crafting the most ambitious narrative yet for the largest tech behemoth in history to go public.

Musk is telling Wall Street: I’m not selling rockets or satellite broadband—I’m selling the interstellar digital infrastructure upon which future human civilization will depend.

In the face of this grand narrative, simply comparing whether ChatGPT or Claude writes more beautiful poetry seems utterly feeble.

Gaze at the stars, deploy your computing power: it’s only just beginning.

Winds begin at the tips of water plants; waves form in the slightest ripples.

When we look back at the turbulent month of May 2026, we will see that the founding of SpaceXAI and its alliance with Anthropic were far more than just a routine press release.

This is a violent collision between the physical and digital worlds.

Musk proved to the world with 220,000 GPUs: the battle over large language models has long moved beyond theoretical debates about algorithms and parameters—it has become a heavy metal showdown of energy, cooling, and who can more极致ly integrate heavy industry with microelectronics.

And what about Anthropic? They’ve received the firepower they’ve long desired. With this gift from the madman of Mars, what waves will Claude stir in the ocean of code? What extraordinary business miracles will the super individuals, who single-handedly deploy thousands of AI agents powered by Claude, create?

The answer is being wildly computed in Memphis's data center at a speed of trillions of floating-point operations per second.

Silicon Valley never believes in tears or moral self-righteousness. It believes only in power. And now, the purest, most violent computing power has shaken hands with the calmest, most rigorous logical mind in an epic moment.

Look up.

Perhaps in the near future, when we use AI to generate a line of code that changes the world, that data won’t come from a data center deep in the desert, but from a SpaceX interstellar server twinkling hundreds of kilometers above your head, in the silent vacuum of space.

Hash power is power. And the game of power has just left the atmosphere.

Source:

1. TradingKey, "Musk Further Expands Space Blueprint, SpaceX Fully Acquires xAI to Build Space AI Strategic Core," February 2, 2026.
2. Digital Today, “xAI Integrates into SpaceX AI Division SpaceXAI, Accelerating Space Data Center Plan,” May 7, 2026.
3. ANI News, “SpaceX Signs Deal Giving Anthropic Access to Its Massive AI Supercomputer,” May 7, 2026.
4. xAI Official Website, "New Compute Partnership with Anthropic," May 6, 2026.
5. The Indian Express, "Anthropic Turns to Elon Musk-Owned SpaceX for AI Compute as Claude Demand Surges," May 7, 2026.
6. Financial Times, "SpaceX to Rent Data Center Capacity to Anthropic," May 7, 2026.