Southeast Asia’s AI narrative is usually told as a startup story. The reality is more steel, concrete, power contracts, and undersea cables than pitch decks.
A study titled “AI in Southeast Asia: An era of opportunity” by McKinsey and the Singapore Economic Development Board argues the region is becoming “the world’s AI arena”. The evidence it puts forward is blunt: over US$50 billion has already been poured into AI-ready data centres and cloud infrastructure by hyperscalers such as AWS, Google, and Microsoft — before you even count the second-order spending on connectivity, construction, and energy.
Also Read: Embracing AI in Southeast Asia: The strategy for avoiding cost overruns
This is the new great game in Southeast Asia: East and West stacks competing side by side, often within the same conglomerates. It is less ideology than latency.
The region is being rewired for compute
Singapore is still the anchor point. The report notes the city-state hosts more than 60 AI centres of excellence (CoEs), including those of Alibaba Cloud, IBM, NVIDIA, and Oracle. That density matters: it pulls in talent, vendors, and enterprise workloads, and it turns “AI adoption” into something companies can buy rather than build from scratch.
But the infrastructure story has shifted south. Malaysia is no longer just the “cheaper neighbour”; it is being positioned as a compute destination. The report highlights:
- AWS committing an additional US$9 billion investment in Singapore by 2028 and US$6 billion in Malaysia until 2038
- Google announcing a US$2 billion data centre and Google Cloud region in Malaysia (2024)
- Microsoft investing US$2.2 billion in cloud and AI services in Malaysia
- Alibaba Cloud opening its third data centre in Malaysia (July 2025)
- Tencent Cloud operating a data centre in Jakarta since 2021
That list is not just bragging rights. It’s a signal that enterprises in Jakarta, Bangkok, Manila, and Ho Chi Minh City can now choose between Chinese and US platforms without shipping data halfway across the planet.
Connectivity is being upgraded to match. The report flags the Southeast Asia-Japan Cable 2 (SJC2)going live in mid-2025: a 10,500-kilometre subsea cable designed to boost redundancy and low-latency links for AI and cloud traffic. The point is simple: compute without connectivity is just expensive heat.
“East meets West” is not a slogan — it’s procurement
The report describes a pragmatic regional approach: companies mix providers, sometimes within the same corporate group, to find the best fit for each workload. It cites a telling example from Indonesia: Tokopedia using Google Cloud for live video and analytics at scale, while GoTo Financial migrated Tokopedia’s core infrastructure to Alibaba Cloud data centres in Jakarta.
That kind of split is not indecision. It’s what happens when the region becomes a battlefield where providers must compete on price, services, and sovereignty—and where enterprises want leverage.
In consumer commerce, the competition is even more visible. The report points to TikTok’s re-entry into Indonesia (via a Tokopedia partnership), YouTube and Shopee rolling out YouTube Shopping in Indonesia, and Temu expanding across markets. AI infrastructure is not being built for fun; it is being built to win commerce, payments, and advertising.
The dirty secret: data centres are a risk business
For all the investment headlines, the report is unusually candid about the downside. Data centres come with volatile returns, and the volatility is structural: AI demand may not ramp at the pace the market is pricing in; hardware cycles are accelerating; and GPU prices can fall, turning today’s premium infrastructure into tomorrow’s stranded asset.
Also Read: AI is eating the world and startups are riding the infrastructure wave
Then there is the stuff nobody loves to talk about at launch events:
- Energy demand: AI compute is power-hungry, and grid constraints can become the actual bottleneck to “AI transformation”.
- Water and cooling: many modern data centres require significant cooling capacity.
- Carbon footprint and materials: the report notes rare earth dependencies and emissions pressures.
It also flags a particularly sharp figure: in Malaysia, data centres are expected to account for around 30 per cent of power demand by 2030. That is not a marginal planning issue. That is a national infrastructure question—one that can drag regulators, utilities, and hyperscalers into the same room, whether they like it or not.
Southeast Asia’s startups aren’t the main beneficiaries yet
The infrastructure wave does not automatically translate into a thriving local AI startup ecosystem. The report argues venture funding remains uneven. In 2024, of roughly US$20 billion in venture investment across the entire Asia–Pacific region, Southeast Asia’s young AI firms received as little as US$1.7 billion. The deal count gap is even starker: 122 AI funding deals in Southeast Asia versus 1,845 across APAC.
So yes, the region is becoming an AI arena. But the early winners are not necessarily local builders; they are often the platforms selling compute and the enterprises with the budgets to consume it.
The talent push is becoming part of the cloud pitch
Even hyperscalers know that infrastructure without skills is dead capital. The report quotes AWS’s Vikram Rao: “AI is the biggest opportunity since cloud computing and possibly even since the internet. . . . Our customer base has grown by five times over 2024 to 2025 alone, and with use cases across every industry.”
Rao also says: “We’ve trained over 1.8 million people in the region since 2017. We have initiatives such as AWS Skill Builder, which offers 600 free digital courses available in local languages…”
Also Read: The real risk in ASEAN’s AI race is not falling behind. It is falling apart
Training is not charity. It’s customer acquisition.
What to watch next: power, policy, and pricing wars
Southeast Asia’s AI infrastructure build-out is entering its more challenging phase. The first phase was announcements and land grabs. The next phase is operational reality: power availability, regulatory compliance, and pricing competition across providers.
If the region wants to be more than a consumption market, it will need to pair the hyperscaler build-out with mechanisms that help local firms capture value: funding, procurement access, and cross-border scale. Otherwise, Southeast Asia risks becoming what the supply chain already knows it can be: a world-class production zone—this time for compute.
—
The image was created using AI.
The post Southeast Asia’s AI boom is built on steel, not startups appeared first on e27.