- Elon Musk predicts that neocloud companies will face challenges due to limited financing options.
- Kevin Williams, CEO of AscendAI, highlights risks in AI data center business models.
- Power supply and infrastructure development pose significant challenges for Utah's AI growth.
SALT LAKE CITY — Elon Musk offered a gloomy outlook this past weekend for neocloud companies as they try to build out data center infrastructure for artificial intelligence.
As the SpaceX CEO announced Sunday on X that the company had closed $40 billion in debt financing to purchase graphics processing units for AI systems, he painted a challenging picture for neocloud firms.
"As of the past few weeks, debt markets have almost entirely shut off access to money for any company that doesn't have substantial revenue independent of AI data centers," Musk stated on X. "SpaceX closed $40B of debt for GPUs, but we're investment grade and have Starlink & launch revenue. That means the 200+ neoclouds are going to see guppy-level death and consolidation. Some will make it through, but not many."
Kevin Williams, the CEO of AI consulting firm AscendAI, said Monday that demand for AI usage will likely only grow, but he said that doesn't necessarily mean business models centered around AI data centers are sound.
"It can be true that demand is going through the roof and data centers are still a very risky proposition," Williams said during an interview with KSL. "It still could be entirely true that a lot of these business models are flawed."
Williams acknowledged that tightened lending has the potential to slow down data center development.
"It could easily slow things down," Williams said. "If capital gets scarce and capital is hard to come by and you need capital to buy the infrastructure required to build these things, all that's left are the providers that have their own capacity that they can lever."
The tech CEO said he believed power supply could prove to be an even greater challenge than financing in Utah.
"That's going to mean that we're going to need to build more power infrastructure as well, which is enormously complicated and enormously expensive," Williams said. "Eventually you're just going to run out of power, and the timeline to build a data center is shorter than the timeline to build a power plant."
As Musk pointed to what he believes will turn into a consolidation of many neocloud companies, Williams said he is less confident that growth will automatically translate into physical infrastructure, noting that AI efficiency can still dramatically improve and that there is also the potential for new approaches to models.
"The data centers that get built will be those that have the cash and the power and the contracts that they can all plug in that make it a solid, independent business," Musk said. "Those are highly speculative — it is speculative. It will be quite risky."







