Just months after announcing an ambitious 2-gigawatt data center complex in Texas, AI startup Poolside finds itself scrambling for new partners after its deal with CoreWeave collapsed and a $2 billion funding round anchored by Nvidia fell apart. The setback reveals the complex challenges facing companies trying to capitalize on the AI infrastructure boom, even as demand for computing power continues to outpace supply.
According to sources familiar with the matter, Poolside’s Horizon project – which would have filled the Texas facility with tens of thousands of Nvidia’s latest Blackwell AI chips – began to falter late last year when the startup couldn’t bring the first cluster of chips online in time to meet CoreWeave’s deadline. This prompted the pair to end their contract, leaving Poolside to hold talks with Google and other cloud providers to revive the project.
The Funding Crunch and Investor Skepticism
Poolside’s troubles extend beyond infrastructure partnerships. The startup’s talks to raise $2 billion from investors, which would have quadrupled its valuation to $12 billion, have also stalled. This included discussions for Nvidia to invest as much as $1 billion in the round. Sources say Poolside struggled to convince potential investors that it could train AI models as sophisticated as those from OpenAI, Anthropic, and Google.
This funding challenge stands in stark contrast to OpenAI’s recent success. The AI giant just raised a record $122 billion in funding, including $3 billion from retail investors for the first time, valuing the company at $852 billion. OpenAI is generating $2 billion in monthly revenue and has over 900 million weekly active users, demonstrating the massive scale and investor confidence that established players command.
The Energy Infrastructure Dilemma
Poolside’s Texas project sits on a half-million-acre site in the Permian Basin, owned by the billionaire Mitchell family who pioneered hydraulic fracturing. The location offers an abundance of natural gas and claims to have a maximum capacity of more than 10 GW of power. Poolside also plans to develop 2.5 GW of solar and wind power there, reflecting the energy-intensive nature of AI infrastructure.
But this rush to build energy infrastructure for AI data centers carries significant risks. David Crane, CEO of Generate Capital, warns that the industry risks overbuilding power plants, potentially leaving power companies with excess capacity costs. “As much as the data centre people tell you their demand for electricity is infinite, it feels to me like there will be a time when they’ll be overbuilt,” Crane told the Financial Times. He advocates for ‘take-or-pay’ contracts where data centers cover infrastructure costs regardless of usage.
The numbers support Crane’s concern. US data center power demand is projected to surge from 34.7GW in 2024 to 106GW by 2035, according to BloombergNEF. Companies like NextEra Energy plan to build at least 15GW of new plants for data centers over the next nine years, while utility capital spending plans have risen 19% for 2026 to 2030.
Strategic Restructuring and Future Prospects
In response to its challenges, Poolside has split itself into two companies – one focused on infrastructure called “PIC” and the other on model building – in an attempt to better appeal to distinct sets of investors. The startup has yet to release a model but hopes to release an application programming interface (API) by the end of April that will show enough promise to rekindle funding talks.
Poolside’s founders, Jason Warner (former CTO of GitHub) and AI entrepreneur Eiso Kant, are betting that the AI market will be large enough for second-tier models to gain a profitable foothold. The startup also sees an opening to sell its products to government and defense contractors after Anthropic was blacklisted by the Pentagon in a spat over the military use of its AI.
Despite the setbacks, Poolside could still benefit from the ongoing scramble for electricity and land between tech giants like Amazon, Google, Meta, Microsoft, and Oracle. Hyperscalers are spending hundreds of billions on data centers and power infrastructure as demand for AI continues to outpace supply. “There are a lot of people in the market for capacity,” said a person familiar with Poolside’s strategy.
The Broader Industry Context
Poolside’s experience reflects a broader trend in the AI infrastructure space. While established players like OpenAI secure massive funding rounds and European startups like Mistral AI invest $830 million in data centers to strengthen regional autonomy, smaller players face significant hurdles in scaling their operations.
The energy demands of AI infrastructure are creating both opportunities and challenges. As Ben Hertz-Shargel, Global Head of Grid Edge at Wood Mackenzie, notes: “The AI ship has sailed, but the energy cost of serving it is very much in question.” This tension between explosive growth and sustainable infrastructure development will likely define the next phase of AI expansion.
For now, Poolside remains confident that Google or another “stronger and better partner” will step in to help them purchase AI chips and finance the required power. But if the startup can’t swiftly find a new counterparty, it could reignite concerns about the riskiness of so-called “neocloud” startups that have been snapping up land, power, and chips to capitalize on the AI boom.