Southeast Asia does not have a shortage of AI pilots. It has a shortage of AI pilots that ever leave the lab. A recent regional survey found that 57 per cent of SEA firms can’t scale their AI pilots, a stretch of corporate limbo now known in industry jargon as the “frozen middle”.
Red Hat, the IBM-owned open-source software company, thinks it can help thaw it. The firm on Thursday launched its APAC AI Centre of Excellence (CoE) in Singapore, backed by the Singapore Economic Development Board (EDB). Its first named collaborator is the Defence Science and Technology Agency (DSTA), the body that develops technology for the Singapore Armed Forces.
Also Read: Southeast Asian tech leaders are learning to trust AI agents, but not with production
The pitch is simple: help organisations move AI from experimentation to production using Red Hat’s open-source stack and engineering staff. The details are harder to pin down.
What the centre will actually do
Red Hat says the CoE’s engineers will work along two tracks. The first is building reusable software architectures, essentially templates for problems that keep recurring across industries, so that each client is not reinventing the plumbing. The second is contributing code “upstream”, meaning into the shared open-source projects that Red Hat’s commercial products are built on, rather than keeping fixes locked inside one customer’s deployment. The company names system deployment, runtime optimisation and platform management as priority areas.
This is familiar Red Hat territory. Over the past two years, the company has pushed hard on open-source inference tooling, betting that enterprises will want to run AI models across their own servers and multiple clouds rather than rent everything from a single hyperscaler.
“AI adoption is moving rapidly from experimentation towards execution, and organisations need open platforms and hands-on engineering expertise to make that transition responsibly,” said Daniel Aw, Vice President and General Manager of Red Hat Asia Pacific.
What Red Hat did not say is just as telling. The announcement carries no investment figure, no headcount, no timeline and no disclosure of what EDB’s “support” amounts to in grants or incentives. For a centre pitched as a regional hub, that leaves readers to take its scale on trust.
The defence angle
The DSTA tie-up is the most concrete part of the launch, and the most sensitive. The two organisations plan to jointly develop solutions in AI, cloud computing, application modernisation and edge computing, which refers to processing data close to where it is generated rather than in a distant data centre. They will also set up practices for DevSecOps, which builds security checks into software development from the start, and “Day 2 operations”, the unglamorous work of keeping systems running, patched and stable after launch.
The work will span on-premise and off-premise environments. In defence, that distinction matters: sensitive systems often cannot touch the public cloud at all, which is precisely where an open-source, run-it-anywhere vendor has an edge over hyperscalers.
“As threats become more complex, it is important that we continue to push the boundaries of technology and innovation to stay ahead of the curve,” said Gayle Chan, Deputy Chief Executive (Information) at DSTA.
Also Read: Singapore’s AI dividend will depend on what happens after the pilot phase
The language is still exploratory. DSTA is collaborating to “explore and co-develop” solutions it “can then apply” to defence uses. That is a long way from a deployed system. It also raises questions the announcement does not touch: what kinds of defence applications are in scope, and what oversight governs AI built for military use. A bank’s chatbot gets public scrutiny when it misbehaves. Defence AI rarely does, which is why clarity on guardrails matters more here, not less. The governance gap is not unique to defence; enterprises across the region still aren’t sure who referees their AI agents.
Selling talent as much as tech
Red Hat is also framing the CoE as a skills play, a theme that lands well with Singapore’s policymakers. The centre includes an innovation team of engineers, architects, project managers and forward-deployed engineers. The last category, popularised by Palantir, describes engineers who embed directly within client organisations rather than advising from afar, an admission that most AI projects fail on messy, company-specific problems rather than model quality.
The centre will also run training in software engineering, automation, cloud-native development and DevSecOps for customers and partners. That addresses a real bottleneck. Singapore firms are scrambling for staff who can build and manage autonomous systems, and agentic AI is now the city-state’s hottest job skill. Meanwhile, messy data keeps Singapore SMEs’ pilots stuck, a problem no platform alone can fix.
“The Centre bolsters Singapore’s position as a key AI hub creating opportunities for local talent to deepen their engineering capabilities,” said Yueh Chinn Seah, Vice President of Digital Industry at EDB.
A crowded rescue mission
Red Hat is far from alone in pitching itself as the cure for pilot fatigue. Earlier this month, Singtel’s RE:AI partnered the SIT-NVIDIA centre to tackle the enterprise AI deployment gap. Singapore-based OneByZero raised US$20 million in Series A funding for the same mission, while Thinking Machines joined the Temus group in a regional push for production-grade AI. Above them all sit AWS, Microsoft Azure and Google Cloud, each with deep pockets, regional data centres and its own enterprise AI programmes. Open-source rivals such as SUSE and Canonical compete for the same hybrid-cloud budgets.
Also Read: Everyone is talking about AI adoption. But are enterprises actually ready for it?
Red Hat’s differentiator is portability. Its argument is that enterprises and governments should not have to hand their AI future to a single cloud provider. That message resonates with the region’s growing anxiety that Southeast Asia cannot build sovereign AI on borrowed choices. It is worth noting, though, that “open” and “independent” are not the same thing: Red Hat is ultimately a US company owned by IBM.
The real test
Centres of excellence are easy to open and hard to judge. The proof will not be the ribbon-cutting but the case studies: how many regional clients move workloads into production, how much code the Singapore team ships upstream, and whether the DSTA work produces anything beyond press releases.
Until Red Hat puts numbers on the table, the CoE is a promising signal rather than a solved problem. For a region stuck in the frozen middle, even a signal counts, but enterprises would be wise to ask for receipts.
The post Red Hat opens Singapore AI centre to pull enterprises out of pilot purgatory appeared first on e27.