In a surprising move, xAI and Anthropic recently announced a significant partnership: the Claude-maker is acquiring all of the compute capacity at xAI’s Colossus 1 data center. This massive 300MW acquisition immediately boosts Anthropic’s usage limits, providing a major uplift to its AI operations. For xAI, this multi-billion dollar deal transforms the company into a formidable compute provider, instantly monetizing a key asset.
xAI’s Strategic Compute Shift
While some initially speculated this arrangement was a jab at OpenAI amidst ongoing legal battles, Elon Musk clarified a different reason on X. He stated that xAI had already moved its AI training operations to the newer Colossus 2 data center. Thus, the original Colossus 1 facility was simply no longer essential for xAI’s internal needs.
From a short-term financial perspective, the benefits of this partnership are clear. With Grok’s usage experiencing a dip following earlier image generation challenges, xAI found itself with surplus computing power. Monetizing this excess capacity adds substantial funds to the balance sheet, a particularly useful move as xAI, now combined with SpaceX, accelerates towards an IPO.
Beyond immediate gains, this deal also lends credibility to SpaceX’s ambitious plans for orbital data centers. Having a major client like Anthropic already committed makes the prospect of space-based compute infrastructure seem far more viable. It effectively positions xAI not just as an AI model developer, but as a critical infrastructure provider in the burgeoning AI ecosystem.
A Unique Approach to AI Infrastructure
The Anthropic partnership sends an unusual message about Elon Musk’s strategic priorities for xAI. It suggests the company’s core business might be shifting towards the construction and operation of massive data centers, rather than solely focusing on developing AI models. This stance is notably rare in the current competitive tech landscape.
Typically, major tech players like Google and Meta, deeply involved in training their own advanced AI models, are continuously building more data centers for internal use. They consistently prioritize reserving their compute resources for proprietary AI tool development over renting them out. xAI’s decision to offload such a significant portion of its infrastructure is therefore striking.
Consider Google, which chose internal AI development over renting GPUs despite being “capacity constrained” as admitted by Sundar Pichai. Meta similarly established Meta Compute to guarantee GPU power for Mark Zuckerberg’s AI ambitions. Zuckerberg called this infrastructure a “strategic advantage,” highlighting that computing power is crucial for building tomorrow’s most lucrative AI products.
Both Google and Meta recognize that running short on compute means missing out on vital future opportunities. This mindset, which sees compute as a strategic imperative for product innovation, contrasts sharply with xAI’s current monetization strategy of selling off its core infrastructure.
The Rise of the Neocloud Player
By emphasizing data center development, both earthbound and potentially orbital, xAI is positioning itself as a “neocloud” business. This model involves acquiring vast quantities of GPUs from suppliers like Nvidia and then renting them out to other AI model developers, such as Anthropic. It’s a challenging business model, distinct from pure AI development.
The neocloud space is notoriously difficult, caught between the pricing power of chip suppliers and the fluctuating cycles of demand from AI developers. This reality often reflects in company valuations; while xAI was valued at $230 billion in its January funding round, a pure neocloud player like CoreWeave, with comparable computing power, is worth less than a third of that amount.
Musk’s neocloud vision is, as expected, more ambitious, envisioning some data centers in space by 2035. xAI also plans to produce its own chips at the Terafab, aiming to mitigate some of Nvidia’s pricing power. However, these initiatives don’t fundamentally alter the basic economics and inherent challenges of the neocloud business model.
The Future of xAI’s Software Ambitions
As recently as February, xAI showcased significant software ambitions, including coding plans bolstered by the Cursor partnership. They also explored leveraging computer usage into full-scale digital twins under the “Macrohard” project. These long-horizon, innovative projects demand dedicated computing resources to succeed.
Yet, by selling off large quantities of its valuable compute capacity to competitors like Anthropic, xAI risks undermining these very aspirations. It raises a critical question: how much future can such new software endeavors truly have if the company consistently prioritizes monetizing infrastructure over reserving it for internal development?
Source: TechCrunch – AI
