More than three in four IT leaders across Asia Pacific and Japan say their infrastructure isn’t fully ready to support on-premises AI workloads — and in Singapore, 77% aren’t ready.

5 minutes ago

Photo by Isaac Maffeis on Unsplash

In the fastest car, every vehicle part and the team behind it work together to the last lap. The same applies to AI.

As Formula One (F1) returns to Singapore, enterprise leaders across Asia Pacific and Japan (APJ) are watching a very different high-stakes race take shape. Yet across the region, more than three in four IT leaders say their infrastructure is not fully ready to support on-premises artificial intelligence (AI) workloads.

If that's where the region stands, the race hasn't even begun.

The bottleneck isn't AI. Deploying advanced models on legacy infrastructure is like dropping an F1 engine into a commuter car. The horsepower is there, but the chassis simply can't handle the torque.

Stalled before the starting line: AI’s “valley of death”
Across my conversations in the region, one pattern stands out: AI is not creating entirely new infrastructure problems. Instead, it's exposing complexity that organisations have managed to work around for years, but can no longer ignore.

Even in Singapore, which reports the lowest level of infrastructure unpreparedness amongst the markets surveyed, 77% of IT leaders say they aren't fully ready, according to a cloud Index report. Across Japan, India, and Australia, the figures are still higher. The gap is real, and it runs across the most prepared markets in APJ.

The clearest sign of this is what happens when organisations try to move from pilots to production — the point where many AI initiatives quietly stall, or what's often called the “valley of death”. A pilot is forgiving. It runs on contained data, controlled conditions, and a team willing to work around the gaps. Production demands consistency, governance at scale, and an infrastructure that wasn't designed to be asked those questions.

That's the moment organisations discover that the problem was never AI. Legacy infrastructure wasn't built for AI being deployed across environments, at speed, under regulatory scrutiny. The model may be ready. The environment around it often is not.

Governance: the pit crew AI can't race without
Governance is where that operational gap is showing up most visibly, and for most leadership teams, arriving faster than expected.

That pressure is harder to defer as AI moves from experimentation into production, and infrastructure gaps become real governance risks for organisations across the region.

For organisations in Singapore, that shift is already intersecting with a harder question: who is accountable when an AI system makes a consequential decision? The Monetary Authority of Singapore (MAS) is strengthening its expectations around AI risk governance, and rightly so.

In almost every market I visit, governance has become a board-level issue long before the underlying infrastructure was equipped to support it.

Increasingly, that accountability question doesn't stop at who owns the decision — it extends to where the data behind it lives and moves. As cross-border data flows come under closer scrutiny across the region, data residency, and sovereignty are also becoming infrastructure decisions.

The regulatory specifics differ by market, but the operational pressure is consistent. From microbanking branches across the Philippines and Vietnam to mining sites in Indonesia with little or no local IT support, governance can't be managed from the centre and applied later. It has to be built into how the infrastructure runs from the start.

Running old and new engines on the same track
Most organisations already have years of infrastructure investment behind them. Core business systems running on virtual machines aren't going away, and nor should they. The real challenge is making those environments work alongside AI workloads under one operating model.

One reason this is so difficult is that AI strategy and infrastructure strategy have often evolved independently of each other.

Containers have become the practical path forward. In Singapore, the same cloud index report found that 88% of IT leaders expect application containerisation to increase over the next three years. But containers shift complexity rather than remove it.

What I consistently see in the organisations getting this right is that they stopped treating containers and legacy systems as separate problems. Everything runs under one operating model because those decisions are no longer separate. Infrastructure, governance, and operations now shape one another.

Winning the race takes more than horsepower
In F1, the fastest car isn't simply the one with the most powerful engine. It's the one where every part of the vehicle, and the team behind it, works together from the first lap to the last.

The same logic applies to AI. The organisations that will scale successfully aren't necessarily those with the most advanced models. They're the ones putting the operational foundation in place before the pressure to do so becomes unavoidable.

That window is narrowing across the region. In APJ, and especially in Singapore, where regulatory expectations are rising fastest, the organisations pulling ahead aren't waiting for a better engine. They're building a machine that can actually finish the race.