Neoclouds driving demand for more chips

For Intel, there is a strong demand for its products from neoclouds.

Neoclouds are booming around the world as businesses are finding them to be a more affordable option in their AI journey. Unlike large hyperscalers, neoclouds basically cater to AI and high-performance workloads.

According to Gartner, by 2030, neocloud providers will capture 20% of the US$267 billion AI cloud market globally and are now being seen as potentially challenging the dominance of hyperscalers especially when it comes to providing high performance computing. In Southeast Asia, neoclouds have seen increased growth with more data centers in the region catering to neoclouds as well.

In a conversation with CRN Asia, Anil Nanduri, Vice President, Product Management, GTM & Customer Success believes that one of the reasons why neoclouds are gaining traction among customers is because cloud hyperscalers cater to lot more use cases.

“For AI, it’s a very simpler stack. And so, neoclouds are formed to just make that simpler stack easily available. It's primarily a cost play and they're much more cost-effective than getting compute from a hyperscaler. But I think this is going to evolve because as neoclouds are trying to figure out how do they go up the stack, hyperscalers are looking at how do they cater to the broader ecosystem. So, I think this is just a cost game at the end. It'll be a TCR cost performance game. And there's also an availability of the models, which also becomes a factor, because you're not going to get all the models accessed locally,” he explained.

For Intel, there is a strong demand from neoclouds for its products. While there is a strong demand across the board, Nanduri pointed out that it only proves that the relevance of the CPUs have become a lot more important. In a neocloud, the CPU handles data pipelines, routing, and workload orchestration while stepping aside for the GPU to do the heavy math.

Xeon6+ and neoclouds

The Intel Xeon6+ processors that is built on Intel 18A, is engineered for sustained performance under real-world power constraints. This not only addresses the orchestration, concurrency, and data movement demands of emerging agentic AI, but the chip also provides greater performance density, power efficiency, and operational scale for cloud-native, agentic AI, and network-intensive workloads.

Xeon 6+ can be configured for AI rackscale infrastructure purpose-built for hosting agents at maximum density. For example, a single liquid-cooled rack can deliver 36,864 cores using 32U of compute space, which provides the highest agent density available (at approximately 100-kilowatt rack power compute).

Optimized for environments where watts per rack, throughput per core, and latency predictability are critical, Xeon 6+ emphasizes scale-out performance, making room for new AI workloads without requiring disruptive data center redesign.

“No matter where the AI models run, there is a need for CPUs to execute, verify, orchestrate, and a need more CPUs to keep up with more GPU computing,” he added.

At the same time, this is where Nanduri feels that companies can no longer just through more compute, memory and story to support their AI as its going to reach point where they cannot be sustained.

“It's more about catering to what is needed in future than about what you could do it in the past. And so, computing over time is always getting better and better. So, you'll find out that instead of running on three, four-year-old CPU hardware, if you upgrade to a Xeon 6 or a Xeon 6+, you're getting better use of your power. You can use less number of CPU cores to do the same amount of work. And so, that allows you to save more power to do other things, or add more CPUs, or add more storage. So, there's a benefit of upgrading the hardware to modernize your infrastructure. And not everyone can go liquid gold. So, you're going to have capabilities where infrastructures can only take certain kinds of servers. And Xeon 6+, caters to that,” he said.

Driving sovereign requirements

Apart from providing more affordable access to compute for AI, neoclouds are also helping organizations deal with the increasing regulatory requirements in their AI journey.

As Gartner has highlighted, increasingly stringent data sovereignty requirements are compelling organizations to seek greater control over where AI data is stored, processed, and governed. Coupled with rising geopolitical concerns, this regulatory pressure is driving enterprises to systematically evaluate their architectures to guarantee localized digital resilience.

In Southeast Asia, neoclouds booming in southern Malaysia are supporting the increasing need for compute and data regulatory requirements in both Malaysia and Singapore. The latest investment by Firmus to build a data center in Batam, Indonesia, with its close proximity to Singapore is also testament of the demand of neoclouds in the region.