Neocloud Rising

Written By Scott Robohn

Scott Robohn

There’s a new buzzword working its way into the AI vocabulary: Neocloud.

What Is Neocloud?

Here's the concept:

  • AI clusters are expensive to build and operate.

  • Unless you're in the business of training AI models, building your own AI training infrastructure - the hosts, the GPUs, the network, the power and cooling, the physical infrastructure - probably doesn't make sense.

  • Access to energy might be the most significant constraint for housing large numbers of GPUs.

  • We have a model for this - the Cloud - and we can apply it to AI model training, effectively providing platforms for Training-as-a-Service and other GPU-intensive use cases.

We should point out that it's not just rent-a-cluster. Neocloud providers are developing their own orchestration layers and other management tools. Some certainly started out as GPU hosting, and some have a heritage in crypto mining, so, evolving from GPUaaS to TaaS is a natural step for many of these providers.

Some example Neocloud providers are:

You can catch nuances in their outward-facing stories that they don’t all target the same use cases. Other Neocloud providers exist, and more are emerging; we'll do more comparative analysis on these in future posts.

What are Use Cases for Neocloud?

  • Crypto Mining

  • TaaS (technology-as-a-service)

  • AI Inference

  • VFX and Rendering

  • Managed Services

  • High Performabce Computing

But it’s important to note that we're not starting from scratch. With "cloud" in the name, you see right away that we're building on consumption models spearheaded by the cloud providers. This has implications and expectations for Neocloud providers, especially in service definitions, operational models, pricing, security, and more.

In a May 2025 analysis of AI Trends, Mary Meeker et al note the rapid acceleration of adoption for key technologies over the past 30+ years. They write,  

"Meeker's data shows AI is being adopted faster than the early internet. ChatGPT, the report's standout example, reached 100 million users in just two months, far eclipsing the growth rates of TikTok, Instagram, or Netflix. By April 2025, ChatGPT had 800 million weekly users and was handling over 365 billion searches annually. These numbers make ChatGPT the fastest-growing consumer technology product in history, achieving in two years what took Google more than a decade."

This rapid growth in consumer adoption is the hunger that generates demand for new models and new AI applications.

There are two additional key accelerators provided by Neocloud companies: 

  • Focus on application development: By taking away the complexity, infrastructure cost, and time to deploy GPU clusters and power them, these providers allow AI developers to focus on their data and application development.

  • Remain at the forefront of power and cooling: With the focus on key highly-specialized technologies in this arena, Neocloud companies can maintain critical staff to absorb new GPU technology, experiment and deploy new power distribution and energy solutions, and related infrastructure issues.

These items further accelerate the ability to develop new models and new applications, eliminating the time and resources that would be needed for an Enterprise to invest in and expand these capabilities.

Everything Builds on the Previous Generation 

This acceleration in adoption is not surprising at all. The buildout of a stable global internet provided the connectivity platform needed for everything else after it. It gave reach to all different types of data from distributed sources and disparate formats. That same platform displaced software delivery via CDs in the mail, allowing "software to eat the world.” That same connectivity platform (with continual buildout and evolution) also provided the infrastructure that facilitated the emergence of cloud platforms. And cloud platforms changed our views regarding different compute consumption models and shared security models. 

Neocloud benefits from each of these tech leaps forward. And somewhat ironically, Neocloud is a perfect example of Making Hardware Cool Again. (Would you like a side of networking with your GPU order?)

We're clearly in the early days of Neocloud. People are arguing about the nuances of the term. That's ok - we don't expect there to be a single uniformly agreed-upon definition this early in the game. But this market will definitely grow. New service models will likely emerge, especially depending on the economics of access to energy, GPU supplier diversity, the evolution of cooperative training across multiple physical data centers, and multi-domain orchestration systems that can get all the compute, storage, and network pieces to stay in sync. Stay tuned.

Scott Robohn
Next

Redefining App Building with AI at the 2025 AWS Summit NYC