Thailand does not design AI chips. It has no equivalent of Nvidia, no homegrown cloud giant like Microsoft or Oracle, and no major cybersecurity firm on the scale of CrowdStrike. And yet, according to a new analysis from Kiatnakin Phatra Securities (KKP) — one of Thailand’s largest investment banks and brokerages, part of the KKP financial group — the country may be about to catch a fresh wave of foreign investment simply because of what it already has: the unglamorous, unavoidable physical infrastructure that every AI data center on Earth needs to stay powered and cool.
Suppapong Iemkongaek, a securities analyst in KKP’s research division, argues that Thailand doesn’t need to build the brains of artificial intelligence to profit from it. It just needs to keep supplying the muscle.
The Electricity Problem Nobody Priced In
The AI boom is usually described in terms of chatbots, chips, and cloud subscriptions. Behind the scenes, it is really an electricity problem. Every large language model runs on racks of servers packed into data centers, and those racks are becoming staggeringly power-hungry. Bank of America Global Research estimates that power draw per server rack — essentially, one column of stacked computing equipment — could jump more than twelvefold in just four years: from 121 kilowatts in 2024, to 177 kilowatts in 2026, to 646 kilowatts in 2027, and past 1,500 kilowatts from 2028 onward.
That surge in power also means a surge in heat, and the traditional copper wiring used to move electricity around a data center simply cannot keep up — it loses efficiency as it heats up, and it isn’t stable enough for these new loads. The fix, according to KKP, involves two shifts: 800VDC systems (a higher-voltage, direct-current power architecture that moves electricity more efficiently than conventional AC wiring) and optical connectivity (using fiber-optic light signals instead of copper cables to carry data with far less energy loss). Neither of these exists on a large scale yet. Building them out is where Thailand sees its opening.
Four Unglamorous Businesses Investors Should Know
KKP breaks the opportunity into four categories that don’t require Thailand to own a single chip patent:
- Server board power — the small circuit systems that manage and convert electricity right next to the chip itself
- PSU and power racks — the power supply units and rack-level hardware that deliver and convert electricity to servers
- Cooling systems — the equipment that keeps increasingly hot, power-dense server rooms from overheating
- Facility infrastructure — the physical build-out of data centers themselves, upgraded to handle high-voltage systems like 800VDC
Bank of America values this specific slice of the global market at roughly $7.9 billion in 2025, growing to $12.3 billion in 2026, and then to $21.8 billion within one to three years — meaning the addressable opportunity could roughly double in the next year or two alone.
There are already signs of momentum on the ground. Applications for data center investment promotion filed with Thailand’s Board of Investment (BOI) — the government agency that grants tax breaks and other incentives to attract foreign manufacturers — exceeded 700 billion Thai baht (roughly $21 billion) in the first quarter of 2026 alone, before slowing in the second quarter as some investors waited for regulatory clarity. For context on how early-stage this still is, Thailand currently operates less than 1 gigawatt of live data center capacity, compared with roughly 7 gigawatts already running in neighboring Malaysia — a gap that shows both how far behind Thailand starts and how much room there is to grow.
For investors specifically looking for a way into this story, KKP points to Delta Electronics (Thailand), listed on the Stock Exchange of Thailand under the ticker DELTA — the Thai-listed arm of the Taiwan-based Delta Electronics group, a major maker of power supplies and cooling equipment — as a medium- to long-term pick for AI-related exposure. That said, the stock is not cheap: DELTA shares fell from around 360–370 baht to roughly 250 baht, partly on weaker-than-expected second-quarter 2026 profit margins, and its price-to-earnings ratio still sits near 100 times — versus roughly 20 to 30 times for Nvidia itself. In plain terms: the market is already pricing in a lot of future growth, so the easy money may already be gone.
The Catch: Thailand Is Still Downstream
Here’s the honest part of the pitch. In global supply chains, “downstream” activity generally means assembly, manufacturing, and lower-margin production, while “upstream” activity — chip design, core research, proprietary technology — captures far more economic value. Right now, Thailand sits almost entirely downstream. Foreign companies with manufacturing bases in the country, including Analog Devices, Lumentum, Seagate, and Western Digital, are driving real investment and activity, but most of what happens on Thai soil is production and assembly rather than design.
That also means the jobs boom hasn’t matched expectations. Modern factories lean heavily on automation, so even large new investments don’t translate into proportionally large hiring. As Suppapong put it, that’s still better than nothing — the activity keeps Thailand plugged into the global technology supply chain, even if the biggest profits are captured elsewhere.
How Thailand Stacks Up Against Its Neighbors
Thailand isn’t the only Southeast Asian country chasing this money, and the regional competition helps explain both the opportunity and the pressure Thailand is under.
Malaysia is currently the region’s biggest data center player by a wide margin. YTL Power International, in partnership with Nvidia, has completed a 600-megawatt AI data center at its Green Data Center Park in Kulai, Johor, running on Nvidia’s latest liquid-cooled GB200 chips, and the Malaysian government has earmarked roughly 5.9 billion ringgit (about $1.4 billion) in its 2026 budget to grow the country’s AI sector further. Malaysia’s GPU appetite has been extraordinary: the country reportedly imported around $6.45 billion worth of AI chips in just the first four months of 2025 alone, more than any other country in the region, with Microsoft, Oracle, and other hyperscalers (large cloud providers that build their own massive data centers) all committing billions more. The tradeoff is that this scale is now generating friction; Malaysia saw its first data center protest in February 2026, when residents near a Johor site rallied against dust, pollution, and water concerns — a reminder that speed has costs of its own.
Vietnam is playing a longer game, trying to move up the value chain rather than just build data centers. The government approved the country’s first wafer fab (a factory that manufactures chips from raw silicon wafers), a roughly 12.8 trillion Vietnamese dong (about $500 million) project aimed at producing specialized chips for defense and AI applications, while Nvidia-partner FPT is building a $200 million AI factory and established players like Intel and Amkor Technology run large chip assembly and testing plants in the country. It’s the same downstream-to-upstream ambition Thailand has, just pursued through chip fabrication rather than power infrastructure.
Indonesia is taking a third path, building data centers to serve its own enormous domestic market rather than to export capacity. Its data center growth is driven primarily by domestic demand, given one of the largest internet populations in the region, with investment concentrated around Greater Jakarta. Microsoft, Google, and Oracle have all launched or expanded cloud regions there in the past two years.
And looming over all of them is Singapore, which isn’t competing on scale at all. Microsoft committed $5.5 billion to Singapore in early 2026 as part of a broader $6.5 billion Southeast Asian package, while Amazon Web Services had already committed $9 billion to expand its Singapore infrastructure. Singapore has positioned itself as the region’s “command node” — the hub for latency-sensitive financial workloads and regional headquarters — rather than a manufacturing base.
Set against that field, Thailand’s pitch is narrower but arguably more defensible: it isn’t trying to out-build Malaysia, out-design Vietnam, or out-govern Singapore. It’s betting that the unglamorous physical layer — power conversion, cooling, facility construction — is a business almost every country’s AI boom will need, regardless of who wins the chip race.
แนะนำ
The Long Road Upstream: Not Now, But Maybe by 2035
Is Thailand likely to eventually make its own chips, rather than just the boxes and wiring around them? Suppapong thinks it’s plausible, but not soon. One meaningful signal: Analog Devices reportedly has plans to expand its Thai operations into more upstream production processes, which Suppapong expects could take about five years to materialize into real capability, with the broader shift toward Thailand becoming a genuine upstream player unfolding over five to ten years.
Thailand does have real structural advantages for that climb: an established education system, existing supply-chain relationships, relatively strong power infrastructure, and a location that has proven attractive to manufacturers for decades. It can also learn from neighbors like Malaysia and Vietnam, who are further along the same path. Suppapong also flagged two specific sub-sectors — AI chip manufacturing and analog chip production, the latter being chips that manage and regulate voltage across everything from electric vehicles to computer power supplies — as the most realistic entry points for U.S. companies already doing downstream business in Thailand to help pull the country upstream.
What This Actually Means If You’re Trying to Profit From It
If you’re an investor, entrepreneur, or business owner watching this from outside Thailand, the takeaway isn’t “buy Thai AI stocks” — it’s narrower and more useful than that. The real opportunity sits in the physical supply chain feeding every AI boom everywhere: power conversion hardware, industrial cooling systems, and facility construction for high-density computing. Those are global demand categories, not Thailand-specific ones, and Thailand is simply one accessible entry point into them, alongside Malaysia’s manufacturing scale, Vietnam’s chip ambitions, and Indonesia’s cloud build-out.
The caution is just as important as the opportunity. Valuations in this space have already run ahead of fundamentals in places — a 100-times earnings multiple on a power-equipment maker is a bet on years of uninterrupted growth, not a value play. And the biggest prize — upstream chip design and the margins that come with it — remains, by Thailand’s own analysts’ admission, a decade away at best. For now, the money to be made is in keeping the lights on and the servers cool, not in building the AI itself.
Key Takeaways
- AI’s real growth constraint is electricity: server rack power demand could jump more than 12x by 2028, forcing a shift to 800VDC power systems and fiber-based optical connectivity.
- Thailand can’t compete on chips or software, but is positioning itself as a supplier of power conversion, cooling, and facility infrastructure for data centers — a market growing from roughly $7.9 billion to $21.8 billion in a few years.
- Thailand remains a “downstream” assembly hub rather than an upstream design center, and automation means new investment isn’t translating into large job growth.
- Regional rivals are ahead in different ways: Malaysia leads on data center scale and GPU imports, Vietnam is chasing chip fabrication, Indonesia is building for its own domestic market, and Singapore commands the region’s highest-value cloud and financial workloads.
- Analysts expect Thailand’s move toward higher-value “upstream” chip production to take five to ten years, with Delta Electronics (SET: DELTA) cited as a medium-term stock pick — though its valuation already reflects significant optimism.
Frequently Asked Questions
Q: Can I actually make money from Thailand’s AI infrastructure boom right now?
A: Yes, primarily through publicly traded companies in the power, cooling, and industrial infrastructure supply chain rather than AI software itself. Analysts have flagged Delta Electronics (SET: DELTA) as one Thailand-listed example, though its valuation already reflects substantial expected growth.
Q: Why does Thailand matter for AI if it doesn’t make chips?
A: Every AI data center needs power conversion systems, cooling equipment, and physical facilities regardless of whose chips sit inside it. Thailand has existing industrial capacity in exactly those categories.
Q: What is the difference between “upstream” and “downstream” in this context?
A: Downstream refers to assembly and manufacturing of finished products — generally lower-margin work. Upstream refers to higher-value activities like chip design and core technology development, which capture more economic value.
Q: How does Thailand compare to Malaysia’s AI data center industry?
A: Malaysia is significantly further along, with roughly 7 gigawatts of operating data center capacity compared with Thailand’s under 1 gigawatt, plus billions of dollars in committed hyperscaler investment and a completed Nvidia-powered facility in Johor.
Q: Is Vietnam a bigger threat to Thailand’s AI ambitions than Malaysia?
A: They’re pursuing different strategies. Vietnam is investing in chip fabrication and design capability (its first wafer fab, Nvidia-linked AI research), which is a more upstream ambition than Thailand’s current power-and-cooling focus.
Q: What is 800VDC and why does it matter for data centers?
A: It’s a higher-voltage, direct-current electrical system that moves power more efficiently than standard AC wiring, which is needed as AI servers draw far more electricity and generate far more heat than older computing equipment.
Q: Will Thailand ever design its own AI chips?
A: Analysts believe it’s possible but likely five to ten years away, contingent on companies like Analog Devices expanding into more advanced production processes already based in the country.
Q: Is Delta Electronics a safe stock pick given its high valuation?
A: Not necessarily “safe” — its price-to-earnings ratio near 100 times is roughly three to four times higher than Nvidia’s, meaning the stock already prices in substantial future growth, so downside risk exists if that growth is delayed.
Q: Does the AI data center boom actually create jobs in Thailand?
A: Some, but fewer than headline investment figures might suggest, because modern data centers and manufacturing facilities rely heavily on automation rather than manual labor.
Q: How does Indonesia’s approach to AI data centers differ from Thailand’s?
A: Indonesia’s data center growth is driven mainly by serving its own huge domestic population and digital economy, while Thailand’s strategy is aimed more at supplying global data center infrastructure and hardware.
Q: What role does Singapore play in this regional competition?
A: Singapore isn’t competing on manufacturing scale at all — it has positioned itself as the region’s high-value “command node” for cloud infrastructure and financial-sector computing, attracting massive hyperscaler investment despite its small size.
Q: What should a foreign entrepreneur take away from Thailand’s AI infrastructure story?
A: The opportunity lies less in “AI” as a buzzword and more in the concrete supply chains — power hardware, cooling, industrial construction — that every country’s AI buildout depends on, wherever in Southeast Asia that buildout happens.
