The rapid adoption of generative AI is fuelling unprecedented demand for GPU-intensive computing, and with it, investment in localised, high-performance infrastructure. Gartner predicts that by 2030, neocloud providers — cloud providers purpose-built for AI and high-performance workloads — will capture 20% of the US$267 billion AI cloud market.
That forecast points to a shift in enterprise cloud strategy, with neoclouds positioned as an answer to scarce GPUs and rising data-sovereignty concerns. Yet for CIOs and CTOs in Southeast Asia and Hong Kong, the promise deserves scrutiny.
Adrian Wong
In an exclusive with FutureCIO, Gartner director analyst Adrian Wong cautions against assuming neoclouds will magically solve hardware shortages, noting that the scarcity of high-end compute is most acute in heavily congested Western hyperscaler regions such as AWS us-east-1 and Azure West Europe.
In APAC, Wong argues, the picture is generally less severe. That helps explain why the region’s neocloud market remains relatively immature: CoreWeave and Lambda are still heavily North America- and EMEA-centric, while their APAC presence is only beginning to emerge.
At the same time, Southeast Asia’s wider data-centre market continues to expand rapidly. JLL says APAC will require US$772 billion in capital investment to add 24 GW of capacity between 2025 and 2030, underscoring the scale of the build-out.
The question, however, is not whether infrastructure is being built. It is whether a neocloud can genuinely deliver better GPU access, within the right jurisdiction, for the right workload.
Sovereignty and the ‘kill switch’
The promise of sovereign cloud capabilities is one major reason organisations are exploring neoclouds. Enrique Castera, Sr director analyst at Gartner, says some specialised providers are differentiating themselves through AI-optimised infrastructure and sovereign cloud features. “Some also focus on sovereign cloud capabilities, ensuring data and operations remain within specific jurisdictions,” notes the analyst.
“Sovereign neoclouds provide contractual guarantees that some or all aspects of the cloud environment, such as data, operations, and governance, remain confined to national boundaries, protecting them from foreign legal claims and extraterritorial access.” Enrique Castera
Yet CIOs should not confuse contractual language with operational immunity. Wong warns that many global IT vendors, including sovereign cloud providers, still carry what he describes as a U.S. “kill switch” risk in their contracts, meaning that extraterritorial laws such as the CLOUD Act can force a provider to terminate access or updates.
He recommends that sourcing and legal teams scrutinise the full legal stack, including terms of service, acceptable use policies, and end-user licence agreements, rather than relying on the master agreement alone.
Wong says the two most important contractual protections are to:
Contractually require the vendor to allow data retrieval and provide termination assistance in the event of a “kill switch” event.
Agree to a notice-and-cure period instead of immediate termination. This provides time to resolve short-term trade restrictions or mount a legal challenge to the termination.
The operating cost of sovereignty
Building a sovereign stack alongside existing global cloud deployments can introduce fragmentation and operational overhead. Wong notes that the cost depends heavily on the architecture chosen: some distributed or isolated private-cloud models reduce fragmentation, but they also shift responsibilities such as capacity planning, facility readiness, and hardware lifecycle management back to the enterprise.
“The perceived availability advantage of high-end GPUs on neoclouds is less about superior procurement capabilities and more about geographic demand for this infrastructure.” Adrian Wong
The ‘minimum viable’ alternative
As a more pragmatic option, Gartner proposes a Minimum Viable Sovereign Stack, designed to keep a business operating through a severe geopolitical disruption or “kill switch” event. The model focuses on running containerised workloads on Kubernetes over bare-metal commodity hardware, avoiding cloud-tethered proprietary licensing, and deploying only “just enough” IT to sustain critical functions.
To avoid fragmentation, he suggests organisations deploy distributed or isolated private clouds, allowing them to deploy hyperscaler infrastructure within a location of their choosing.
“The level of fragmentation that exists between public cloud and distributed cloud environment varies from solution to solution (not just provider to provider).” Adrian Wong
As outlined by Gartner, the MVSS approach recommends:
Simplifying the infrastructure: Running containerised workloads on Kubernetes on bare-metal, commodity hardware.
Complete data centre ownership: Physically stockpiling spare hardware in advance.
Deploying “just enough” IT: Substituting legacy systems with lightweight, container-native applications custom-built to provide critical functionality.
That is not a shortcut. It is a resilience strategy.
Castera captures the broader market shift succinctly: “The AI cloud market is entering a new phase where sovereignty, performance, and infrastructure specialisation are becoming primary decision factors for enterprises.”
“As demand for GPU-intensive workloads accelerates and traditional cloud models struggle to keep pace, it is creating the conditions for a new class of providers purpose-built to deliver AI infrastructure at scale.” Enrique Castera
For Southeast Asian organisations, the practical test is straightforward: does a neocloud genuinely offer better GPU access, stronger sovereignty, and acceptable operational resilience in the specific jurisdiction where the workload will run? If not, the label matters less than the underlying architecture and legal reality.
