An Indonesian e-commerce platform signing memoranda of understanding for AI data centers in Thailand and Indonesia might look like a footnote — but it's the latest signal that AI datacenter expansion Southeast Asia 2026 is becoming a real alternative to stretched US and EU capacity.
View the data behind this chart
| Thailand MoU | Indonesia (full target) | Indonesia (phase 1) | |
|---|---|---|---|
| Planned capacity | MW100 | MW10 | MW5 |
What Polibeli actually announced
Polibeli, a Cayman Islands-incorporated B2B e-commerce platform wholly owned by China's Xingyun International and listed on Nasdaq since August 2025, has confirmed it is "evaluating opportunities" in AI data centers as part of a broader long-term growth strategy, according to DataCenterDynamics.
On 30 June, the company signed a non-binding MoU with Authaikam Company Ltd to assess building an AI data center in Thailand, with the site potentially offering up to 100MW of capacity. Two weeks later, a second non-binding MoU followed with PT Grosirone Prima Nusantara for a 10MW AI data center in Indonesia, with an initial phase of 2MW to 5MW.
Both agreements are explicitly preliminary — covering information exchange, site visits, land and power assessments, and due diligence rather than committed construction.
Why an online wholesale platform wants into data centers
Polibeli describes itself today as an online wholesale platform facilitating B2B product sales in Indonesia — a long way from power distribution units and liquid cooling loops. In its own words: "The company believes that AI represents a significant long-term trend across the global technology landscape. Leveraging its regional presence, customer network, and experience in technology-enabled markets, the company intends to actively evaluate opportunities that may support its future growth and enhance value for customers and shareholders."
It isn't alone. DataCenterDynamics notes a growing list of unlikely entrants chasing AI infrastructure, from a rebranded former shoe company to a Taiwanese whiskey importer, a recruitment website, a Singaporean healthcare firm, an Indian TV production company, a Chinese car loan financing business, an aviation operator, and a Hard Rock Cafe franchise owner in Malaysia. For UK buyers, that pattern matters as much as any single MoU: it signals speculative capital chasing AI real estate wherever land, power and permits align — not always where operational maturity does.
Thailand and Indonesia's growing pull for AI capacity
The specifics of these two MoUs point to genuinely useful capacity if either progresses: a 100MW Thai site would sit comfortably alongside mid-sized hyperscale campuses, while the Indonesian project's phased 2MW-to-5MW approach mirrors how many emerging-market projects de-risk build-out before committing to full scale.
Thailand and Indonesia datacenters are attracting this kind of interest because both markets offer land availability, developing power infrastructure, and government appetite for digital investment that's harder to find in supply-constrained hubs like London, Frankfurt or Northern Virginia. That combination is exactly what underpins the case for AI infrastructure geographic arbitrage.
The UK buyer angle: alternative regions, real trade-offs
For UK enterprises weighing where their next AI workloads run, Southeast Asia's emergence as an alternative AI cloud region is worth watching rather than acting on immediately. Power availability and grid connection queues in the UK and EU remain a genuine constraint on scaling GPU clusters, and that's pushing some buyers to look further afield for capacity and cost.
But sourcing compute from a market where the leading local project is still at MoU stage — not shovel-ready — carries obvious latency, data sovereignty, and supply-chain risk that needs weighing against any cost saving. Buyers should treat this as an early signal to monitor, not a procurement option to lock into today.
For those still building capacity closer to home, tools like an AI GPU calculator help model workload requirements before committing capital, while IT finance options can bridge the gap between wanting scale now and waiting for new regional capacity to mature.
Due diligence for anyone eyeing Southeast Asian capacity
Non-binding MoUs signed by a wholesale e-commerce platform with no prior data center track record are not the same as a hyperscaler breaking ground. UK buyers should distinguish between announced intent and delivered infrastructure, and factor in that power assessments, land due diligence, and regulatory sign-off in these markets are still in early stages.
- •Confirm whether power and grid connection studies have actually been completed, not just scoped
- •Check the operational track record of local partners named in any MoU or joint venture
- •Model total cost of ownership including cross-border connectivity and latency, not just headline site cost
- •Watch for phased capacity commitments (like the 2MW-to-5MW Indonesian first phase) as a sign of realistic risk management
Where this leaves procurement planning
Whether or not Polibeli's projects proceed beyond preliminary evaluation, the pattern reinforces that AI capacity sourcing is diversifying globally. UK buyers building resilient strategies should keep evaluating both domestic build-out — including building an on-premise AI cluster and AI server cooling solutions for on-site density — and emerging regional partnerships as complementary, not competing, paths to secure compute.
Reviewing IT infrastructure solutions and current GPU accelerators pricing alongside data from AI servers studies gives buyers a grounded baseline before any Southeast Asian capacity becomes commercially available.