Neoclouds. Neoscalers. Whatever you call them, a new breed of infrastructure player has captured the industry’s attention. Everyone wants to sell to them. Or be seen working with them. More agile than their hyperscale counterparts. More financially flush than Mr. Monopoly. And more willing to double down on AI. These are the providers of tomorrow, today.

But what exactly constitutes a so-called neocloud? Neo comes from the ancient Greek word néos, meaning new. Does that mean Meta would constitute a neocloud, given it’s believed to be entering the cloud game? Or how about CoreWeave or Crusoe, vendors that started life as cryptocurrency mining operations before later pivoting to AI? Are they only neoclouds now that they’ve taken on a largely AI focus?

And, more importantly, how do such players want to be seen?

Definitions and denominations

Analyst giant Gartner defines neoclouds as “cloud providers built specifically for AI and high-performance workloads” – with such vendors projected to capture 20% of the $267 billion AI cloud market by 2030.

An explainer from Cisco goes one further, stating said players leverage high-performance hardware compute accelerators – “primarily graphics processing units (GPUs)” – to support AI application needs of enterprises, model builders, and even hyperscalers.

Meanwhile, a similar Equinix outline points to neoclouds offering more than just GPUs and compute infrastructure, referencing an array of services including AI-optimized object and file storage; model training and fine-tuning; and low-latency, high-bandwidth networking and connectivity.

The consensus then centers around one key factor: AI. These players put AI at the forefront of everything they do and offer, differentiating themselves from general-purpose players by riding the wave of AI intensity, while also ensuring agility given their size when it comes to buildout that their monolithic competitors simply cannot compete with.

Another key differentiating factor as to what exactly constitutes a neocloud is how they use their networks. While traditional internet traffic has a “uniqueness” to it, with a larger number of data centers and a mix of workloads resulting in an array of outputs, Backblaze’s quarterly “Network Stats” analysis reveals neoclouds make use of fewer, more sizable datasets moving in short, sustained bursts across persistent endpoints and pipelines.

The vendor’s “Q2 2026 report,” for example, found that neocloud traffic shows a more moderate, synchronized volatility, with ingress and egress scaling together in sustained, business-hour patterns. That’s in comparison to hyperscalers which enjoyed sharper volatility, with egress swinging between 50 and 150 gigabits per second (Gb/s) against a stable, tightly controlled ingress baseline.

Challenging the neocloud narrative

Gleaning an accurate perspective on how these emerging players envision their own position requires a view direct from the source.

At the vanguard of the neocloud boom is CoreWeave, perhaps the darling of specialized infrastructure. Boasting a market cap hovering around the $50 billion mark at the time of writing, the New Jersey-based outfit has solidified its status through massive, multi-year capacity agreements with the likes of Meta and OpenAI.

Its meteoric rise was epitomized in its placement on the first-ever Gartner Magic Quadrant for cloud AI infrastructure, with CoreWeave coming out as the top-ranking “visionary” player – a segment that included old hacks like IBM.

Despite forcing its way into the emerging player limelight, CoreWeave isn’t too much of a fan of the neocloud name.

“Think about categories in tech that actually stuck: Internet of Things, mobile apps, cloud computing, search engine, database, container orchestration, machine learning platform, etc. Each one tells you what it does. ‘Neocloud’ tells you when it showed up. There’s no purpose in the name. Just newness, which is the one thing guaranteed not to last,” Chen Goldberg, CoreWeave’s EVP and former Googler, wrote in an impassioned LinkedIn article.

Goldberg even took issue with the pattern of industry framing toward such features like GPU access and flexible pricing, arguing the vendor’s place in the world was more “a story about connecting systems in unconventional ways.”

“Connecting the technology stack end-to-end. Every layer is built with intent, for workloads that look nothing like what the cloud was originally designed to run,” Goldberg wrote. “Our customers don’t come to us to rent GPUs. They partner with us to run production AI on our purpose-built stack.”

Goldberg’s framing of where CoreWeave stands leaned on something said at this year’s GTC event by none other than Nvidia CEO Jensen Huang, who instead likened CoreWeave to a “new generation of AI-native clouds.”

Corey Sanders, CoreWeave’s SVP for product management, outlined the framing further to SDxCentral, contending it was less about being just a newcomer and more of a cloud built specifically for AI workloads.

While hyperscale rivals are all building comparable AI-native stacks, Sanders, a Microsoft alum, argued it was difficult for those players to copy such offerings without redesign or redeployment of existing cloud infrastructure or cloud solutions, adding: “It’s a little bit trickier to deliver a very specific AI solution focused on storage, caching, and throughput if you need to support a wide range of different services solutions.”

“From my perspective, we’re in a place where we’re still new, we’re still innovating,
but it feels very temporal a way to describe us. And AI cloud for us at least fits better,” Sanders added.

Building from the network inward

One such player who would qualify the neocloud name is Megaport. The Michael Reid-led firm started as a network-as-a-service (NaaS) provider, but in a little less than a year has supplemented its core remit with a bare metal compute stack from Latitude.sh.

Megaport offers compute services following its acquisition of Latitude – given that its provisioning, monitoring, and routing operations are handled autonomously – that tick the baseline neocloud box. Gui Soubihe, founder of Latitude and now part of Megaport’s leadership, put it that the company “enabled customers to go to our platform, provision one physical server in Japan, and have it spun up in five seconds, and then … reinstall this machine and allocate [it] back to the cloud pool, so another customer could use it.”

But unlike compute-focused CoreWeave and company, Megaport started with the network and built outward. Megaport CEO Reid said this approach allowed the company to take its existing global network and add compute capacity quickly, “without negotiating every agreement or building the supporting infrastructure from scratch.”

“Companies starting from the opposite end first need to secure the facilities, power, and space to house their compute clusters, and then solve the network on top of that. That is an unavoidable, time- and capital-intensive undertaking,” Reid said.

Soubihe went deeper, reminding that scaling high‑end AI clusters requires an entire extra level of networking capability: “The deployment of GPUs requires a lot of networking, not to the outside world but inside the data centers. You have to interconnect these all together as a cluster.”

The Latitude founder described Megaport’s NaaS heritage as a differentiator rather than an add‑on. And in fact, the acquired brand “didn’t have this expertise” in building those GPU fabrics at scale, so Soubihe admitted the combined company “pulled talent from Megaport to help us implement and deliver this east and west connectivity that these GPUs required inside [the] data center.”

“Power is being a constraint in the U.S. and globally, and Latitude would not be able to do the kind of deployments that we are doing without Megaport,” Soubihe added.

That combination of pre‑existing global network reach and bare‑metal automation is now being turned toward large‑scale AI, with Soubihe suggesting Latitude’s business has grown six‑fold following the acquisition, adding Nvidia GPUs and a powerful, Vast Data-supported storage layer on top of its historic CPU footprint.

While CoreWeave’s reluctance stems from what it sees as a vague, temporary tag that doesn’t reflect its long-term positioning, Reid too tempered his view on Megaport’s potential moniker in that while it fits the bill for what a neocloud is, such a definition is “simply too narrow.”

“We prefer to describe Megaport as an automated infrastructure platform. The common thread across our compute and network businesses is the automation of global physical infrastructure – making what would traditionally take months to deploy available on demand in seconds,” Reid said.

The ultimate debate over neocloud naming conventions is secondary to the actual structural shifts these emerging vendors are forcing upon the market. Whether they define themselves as AI-native stacks or automated infrastructure platforms, the common denominator is a shared intent to break the monolithic, general-purpose grip of the big boy hyperscalers. As the market matures, terminologies will likely fade, but more agile, purpose-built blueprints for AI compute will remain.