The cloud was built around a relatively simple idea: make computing available as a service. Instead of buying and operating every server, a company could rent capacity from a large provider and scale as its needs changed.
That model changed the economics of software. It also created a highly concentrated infrastructure market. A small number of large providers now offer the broadest mix of computing, storage, networking and software services. Their scale is difficult to match, but scale is not the only advantage that matters in the next phase of computing.
Artificial intelligence is creating demand for a different kind of cloud service. Training and running advanced models require specialized processors, high-speed networks, large amounts of power and carefully designed data-center environments. The customer is not simply looking for a virtual machine. The customer is looking for access to a coordinated computing system.
This is where neocloud providers are emerging.
The term describes companies that focus on a narrower part of the cloud market, often accelerated computing for artificial intelligence and other demanding workloads. Their proposition is not that they can immediately replace the largest cloud providers across every category. It is that specialization can make them more responsive to a customer with a specific technical requirement.
Why specialization matters
A general-purpose cloud platform must support a wide range of applications and customers. A specialized provider can design its infrastructure around a smaller set of workloads. It can make decisions about processor configuration, networking, cooling, storage and software integration with a particular use case in mind.
The opportunity is therefore less about building another broad technology supermarket. It is about turning scarce and complex infrastructure into a dependable service.
The operating challenge
Specialized computing is capital intensive. Hardware is expensive, power availability can constrain expansion, and equipment can become outdated quickly. A provider must keep utilization high enough to cover those costs while still giving customers predictable access to capacity.
Software and operations are just as important. Customers need scheduling, monitoring, security, data movement and developer tools. They also need confidence that a workload will perform consistently. A collection of powerful chips is not, by itself, a cloud platform.
The largest providers still have significant advantages. They have established customer relationships, global infrastructure and the ability to bundle computing with many other services. Neocloud companies must therefore compete where their specialization creates a meaningful difference: access, responsiveness, performance, transparency or support for a particular class of workload.
A more diverse cloud market
The market may ultimately contain several layers. Large providers can continue to offer broad infrastructure, while specialized providers focus on particular forms of accelerated computing, geographic regions or industry requirements. Customers may use more than one model depending on the workload.
The rise of neocloud is a sign that cloud computing is becoming more diverse, not necessarily that the existing model is disappearing. As computing becomes more specialized, the value may move toward providers that understand the complete system behind a workload. The winners will be the companies that combine scarce infrastructure with reliable operations and a clear reason for customers to choose them.
Don Basile writes about computing infrastructure, advanced technology and the business questions shaping the next generation of digital systems.