Table of Contents
JLL noted that the relationship between hyperscalers and neocloud providers is evolving from competition to partnership
In sum – what to know:
AI creates a new data center asset class – Training workloads need up to 10x higher power density, enabling AI-focused facilities to achieve lease rates as much as 60% above traditional data centers.
Inference drives distributed growth – By around 2027, inference is expected to surpass training, accelerating demand for edge and geographically distributed AI infrastructure.
Hyperscalers turn to neoclouds – As AI capacity needs grow, hyperscalers are increasingly partnering with GPU-focused neocloud providers to gain speed, scale and reach.
The rapid expansion of AI infrastructure is creating a major new market opportunity for neocloud providers and reshaping the data center industry. According to a recent report by JLL, AI represents the most significant shift in the sector since the first wave of cloud adoption more than a decade ago.
As AI workloads scale, annual revenue in the data center semiconductor market could reach nearly $500 billion by 2030. AI training workloads require roughly ten times the power density of traditional enterprise computing, driving demand for a new class of AI-optimized facilities. These assets can command lease rates up to 60% higher than conventional data centers, according to JLL.
Demand for AI infrastructure is expected to accelerate sharply over the next five years. JLL points to a critical inflection point around 2027, when inference workloads are projected to overtake training. This transition will push compute closer to end users and support more distributed, edge-based deployments.
This shift aligns closely with the core strengths of neocloud providers. Neoclouds are specialized cloud companies that offer GPUs as a service, focusing on flexibility, rapid deployment and geographic reach. These attributes position them well to support distributed AI workloads at scale.
“Early movers like CoreWeave have demonstrated this opportunity, scaling from a startup to a $19 billion valuation in just four years by building AI infrastructure that outcompetes cloud offerings on performance, flexibility and price,” JLL said.
“The relationship between hyperscalers and neocloud providers is evolving from competition to partnership. While hyperscalers dominate cloud services through scale, they’re increasingly partnering with neocloud providers for AI workloads that require larger capacity requirements, broader geographic reach and faster deployment times,” the company added.
In a previous interview with RCR Wireless News, Reece Hayden, principal analyst at ABI Research, explained how neoclouds position themselves relative to traditional hyperscalers, and why their architecture and business models set them apart.
“Neoclouds at their core are really just AI infrastructure provision. And it’s really about providing GPUs in the most effective way in purpose-built architecture to enterprises, AI ISVs, AI vendors,” the analyst said. Rather than focusing on services or software ecosystems, Neoclouds lead with hardware availability and GPU-centric architectures designed specifically for AI training and inference, according to Hayden.
Their consumption models vary, ranging from bare-metal GPU rental to colocation arrangements where the neocloud supplies the servers, racks, and cooling. Hayden notes that “neocloud is providing the infrastructure, the servers, the racks to enable the cooling, all of the different factors you need to deploy and implement GPUs into data centers.”
The contrast with traditional hyperscalers is sharp. Hyperscalers remain broad service platforms rooted in CPU-based general-purpose compute. Hayden points out that many in the industry describe hyperscalers as “CPU clouds,” whereas Neoclouds design infrastructure around GPUs from the ground up. Scale is another defining difference. Hyperscalers are global; neoclouds tend to be smaller and more regional, with strong traction in sovereign data center projects, particularly in Europe, according to the analyst.
