SpaceX is raising funds through an IPO to build a "space data center," yet it is renting out computing power at high prices to Google and Anthropic! Netizens are hotly debating: Why is this?

```

SpaceX is turning its massive computing infrastructure into a fast-moving cash machine, but contradictions behind this business logic are drawing increasing market scrutiny.

According to Reuters on the 6th, SpaceX has successively signed huge computing rental agreements with Anthropic and Google—Anthropic pays $1.25 billion per month, and Google will pay $920 million per month starting in October this year. These two contracts represent annualized revenue of around $26 billion, with a combined total contract value of over $70 billion.

This series of deals provides a strong narrative backing for SpaceX's sprint toward the largest IPO in US stock history, with a targeted financing scale of up to $75 billion.

However, multiple market observers and analysts on social media are raising a more fundamental question: Why is SpaceX renting computing power to competitors instead of using it to train models in its own AI lab xAI? The answer to this question may point to an expensive design error inside xAI, which is triggering a chain reaction of skepticism toward the entire transaction logic and IPO narrative.

Questionable pricing for compute rental; critics call it "circular finance"

Sharp doubts about the rationality of the pricing for these two deals have emerged on social media.

Former assistant research professor Roger broke down Google’s deal with detailed mathematical calculations on X. He pointed out:

CoreWeave ($CRWV) offers GB200 computing at about $10.50 per hour, equivalent to $5.25 per hour per Blackwell GPU chip. Based on this, the annual cost for 110,000 chips is about $5.06 billion.

In other words, Google could obtain the same compute capacity from CoreWeave at a lower price, even paying a 100% premium to CoreWeave would still save about $1 billion—yet Google chooses to rent from SpaceX at a higher price.

Based on this, Roger concludes, that the prices Google and Anthropic are paying SpaceX exceed the cost of building private data centers, and categorizes such transactions as "circular finance"—giant tech companies using pension funds to mutually boost valuations and create artificial revenue growth.

AI researcher Gary Marcus approaches it from another angle, arguing that focusing on how much SpaceX charges is "the wrong question." He believes the real issue is why SpaceX is making these deals—in his view, the answer is xAI has realized it cannot win the frontier model race, so it needs to maintain good financial performance before the IPO through such deals.

IPO narrative contains contradictions, "space data center" vision questioned

SpaceX’s IPO documents list “orbital data centers” as a core growth narrative, with plans to deploy them around 2028. However, X user Chief Agenteer directly pointed out the logic gap between this narrative and reality.

He wrote: "SpaceX is renting out idle computing power, which itself shows that space data centers are fundamentally unnecessary and will never be economically viable. If they can't use all the computing power built on the ground, why build more in space?"

This skepticism connects two originally separate pieces of information: on one hand, SpaceX is forced to rent out ground computing power due to xAI training setbacks; on the other, its IPO documents paint an ambitious blueprint for space-based computing expansion. The tension between the two makes some investors question the overall strategic coherence of SpaceX's AI business.

xAI training setbacks; Colossus 1’s chaotic architecture as the spark

According to tech media WccfTech, SpaceX’s xAI lab faced serious technical obstacles at its Colossus 1 data center in Memphis, Tennessee. The center deployed three Nvidia GPU architectures—H100, H200, and GB200—in a mixed setup, which prevented xAI from effectively training its Grok models there, eventually forcing the migration of training tasks to the Colossus 2 data center.

This technical predicament directly explains SpaceX's motive in renting out compute. WccfTech's analysis points out, SpaceX is essentially trying to monetize xAI’s poor design decisions—by signing cloud service agreements with external clients like Google and Anthropic, converting originally idle or inefficiently used Colossus 1 compute capacity into considerable rental income.

This background further increases doubt about xAI's independent R&D capabilities. SpaceX's AI lab is currently facing both core talent drain and R&D setbacks, and the Colossus 1 architectural issues make matters worse.

Currently, SpaceX plans to go public next week. While high-visibility income from its compute rental business provides strong support for the IPO, controversies over transaction pricing logic, xAI's R&D abilities, and the commercial viability of space data centers may be variables investors need to weigh carefully in the valuation process.

Risk Notice and DisclaimerThe market involves risk, investment needs caution. This article does not constitute personal investment advice, nor does it take into account individual users’ special investment goals, financial situations, or needs. Users should consider whether any opinions, views, or conclusions in this article fit their specific circumstances. Invest accordingly, at your own risk. ```