There was a time in 2025 when it seemed like a new neocloud or GPUaaS provider was popping up every week.

Things have slowed down a bit since then, but there are still plenty of new players emerging to take a swing at the lucrative AI infrastructure market.

For Kevin Cochrane, CMO of global cloud provider Vultr, the drive for this was pretty simple. “How did we get here? Well, with the debut of GPUs and the ability to generate new applications and economics,” he says. “Suddenly, every single data center operator wanted to get in on the neocloud business.”

Sadly, as Cochrane notes, while the motivator might be obvious, the execution is challenging. Those seeking to get a piece of the neocloud pie have varied from existing smaller data center providers, to cryptominers, and some notably peculiar cases.

A Taiwanese whiskey importer, a recruitment website, a Singaporean healthcare company, an Indian TV production firm, a Chinese car loan financing company, an aviation operator, and an owner of Hard Rock Cafe franchises in Malaysia have all made an unlikely pivot into data centers. More recently, shoe brand Allbirds announced it was giving up on footwear to focus on servers instead. These stories make for great headlines and certainly for a chuckle at the editorial desk at DCD.

But, for Cochrane, this “boom” will inevitably lead to a “bust,” and that bust will take the form of a consolidated market.

For now, there is money to be made, and there are likely to be a handful of big winners. Synergy Research Group, in its latest neocloud market forecast, lists CoreWeave, Crusoe, Lambda, Nebius, Nscale and - oddly, a data center provider and not a cloud provider - Core Scientific as the top dogs of the neocloud world.

But below that tip of the iceberg, there are hundreds in the icy depths trying - and not always succeeding - to keep up. This is because buyers “are recognizing that there are things that matter more than the land, shell, and power and ability to buy the GPUs,” argues Cochrane.

He breaks the problem down into three areas. A lack of operational know-how, a lack of capital, and the growing issue of sovereignty. “Enterprises don’t know what to do with a GPU cluster,” he says. “The only companies that do know what to do with a bunch of naked hardware are the big off takers; Microsoft, Meta, Google, AWS, etc, and that's because they bring their cloud operations team and software stack to the mix.”

Certainly, the neoclouds that are becoming household names (or at least, households deep in the tech world) have made improving their platform a key priority.

Speaking to DCD at the recent Vast Forward 2026 conference, CoreWeave’s Chen Goldberg noted that its operating system is a key differentiating factor for the company. Goldberg is a former GM and VP of engineering at Google Cloud. She left the hyperscaler for CoreWeave in 2024, but noted that she was hesitant when the neocloud first reached out, as she didn’t want to just “work with a GPU reseller.”

She recalls: “They said, ‘You have to talk to Peter’ [Salanki, CoreWeave’s CTO]. I met with Peter, and he showed me the first version of mission control, and it addressed things that I’d seen at Google, of how hard it is to get observability data at the right time and take actions on top of it."

Goldberg continues: “I saw it as a team and a company that was ready to challenge the first principles of how cloud infrastructure should be built.”

For Goldberg, it is CoreWeave's approach to building a cloud stack that is “very powerful,” which has enabled its success.

“The reality is that networking, storage, and compute can all impact your workloads directly, and observability matters, security matters,” she says.

CoreWeave, like many of the neoclouds at this point, has built its platform upon the Vast Data AI OS - an operating system that underpins how these clouds are designed and operated.

Another customer of Vast Data is 5C Group, but when asked if the company considers itself a neocloud, David Bitton, 5C’s VP of AI, chuckles and ultimately says “no.” “When we started our ‘adventure’, we considered the neoclouds as a competitor,” he explains, before noting that the company’s niche has now enabled it to have neoclouds as customers, instead. Rather than try to offer the best platform, 5C has instead focused on building the “data center, cloud infrastructure, and bare metal on demand and they [the neoclouds] can put whatever they want on it and operate as a white label. Those companies suddenly became potential customers and not competitors.”

5C Group itself was formed as the result of a merger and acquisition. Hypertec Group, founded in 1984, focused on HPC and the designing, building, and operating of data centers. In May 2025, however, the company partnered with Together AI on a 2,000-GPU AI cluster that was set up in one month. The project was a success, and Hypertec wanted to get its own slice of the AI market.

“We had two choices,” Bitton says. “Either we start getting huge financing and building many clusters and wait to start selling those, or we can be niche and build on demand and just stop at the bare metal.”

Deciding to do the latter, Hypertec then spun off its cloud business from the wider group, and acquired 5C Data Centers, forming 5C Group. Today, 5C operates both a cloud business and a colocation provider. For 5C, it is not a problem if there is a neocloud market consolidation, though Bitton concedes that it could be a reality.

“It takes a lot of money to grow that business, and that's a bottleneck for some,” he says. “I could see mergers and acquisitions as a way to grow without having to invest in projects and build from scratch. “But, for us, our side of the business, it doesn't matter whether it's converging or merging and all that, we're still going to provide the data center and the compute infrastructure if needed.”

The requirement for ongoing investment is the second factor that Vultr’s Cochrane suggests is going to hit the industry.

“If you want, say, 1,024 GPUs as a small start up, then in three months you will need 5,000, and then three months later you need 10,000,” he argues. “It doesn’t stay put, the demand is growing exponentially, and the moment you tap out of the capital markets, the customer will need to go elsewhere.”

Certainly, we have held witness to the seemingly never-ending capital raises - be it in the form of debt or equity - by neoclouds, and those at the top of the food chain are fundraising in the billions.

By the end of 2025, CoreWeave had amassed $21.6bn in debt, and this is likely to grow in 2026.

Nscale raised $2bn in March 2026, and that same month, Nebius was seeking around $4bn in convertible senior notes.

“A lot of neoclouds will just wash out because if you are a syndicate of lenders in a major city, you aren’t going to lend to 200 different entities, you are going to pick the two or three best options, the grade A investment vehicles,” argues Cochrane.

Simply put, without ongoing capital injection, there will be limited options, and the odds of getting snapped up by a larger player will only grow. Shortly after DCD spoke with Cochrane, Brookfield-owned Radiant merged with UK neocloud Ori Industries.

Ori’s CEO and founder, Mahdi Yahya, said of the acquisition: “Through Radiant, we can challenge the supply-demand imbalance that has defined AI since the release of advanced LLMs in 2023. With deep, structural advantages in capital costs, powered land, compute, and software, Radiant is building the infrastructure to enable a global age of abundance for AI."

While just one example of M&A activity, the timing was serendipitous, and Nscale’s CRO Tom Burke certainly entertained it as an omen, raising it as an example when talking to DCD in late February.

“There are a lot of neoclouds that exist, and I think we are already starting to see some of that consolidation. It remains to be seen how many will survive in the long term, but I would argue that we are still just scratching the surface,” Burke says.

Burke points to the growing sovereignty requirements for AI cloud workloads, noting that the enterprise is currently “moving from experimentation to production deployments.”

“I think it remains to be seen how many neoclouds are going to be needed to fulfil this market demand, but I think we are a couple of years away from knowing. The question is, how many neoclouds are set up to be able to weather that storm? If you look today, it's a little top-heavy. There is a clear separation of the top four or five, and then everyone else is trying to find their niche.”

Despite this, Burke remains confident that Nscale will be one of the survivors. “Everything we’ve done at Nscale is to put us in a position that, should that even arise, we are able to make the best decision for the company on behalf of our customers and partners to tap into capacity,” he says.

This sentiment falls in line with Cochrane’s overarching prediction, including his thesis that sovereignty will play a major role in the matter.

“This is why I think a neocloud consolidation is going to happen,” he says. “There will be a limited number of truly global players, and there will be a bunch of regional and national ‘champions,’ but it will be limited.”

As for how that consolidation will present itself, Cochrane argues that the assets held by neoclouds are undeniably valuable.

“At some point in time, they will need to fold into existing entities that can operationalize it at scale, can upgrade, maintain, extend, and build upon it,” he suggests, adding that they can become something of a “point of presence” for those more established providers.

As for the companies trying to enter the market now, Cochrane has little hope.

“The time for a groundbreaking market entrance is well past,” he says. “I don’t think the capital markets want it, and I don’t think the chip manufacturers want it either.”