The Map No One Is Drawing:
The factors beyond cost and water that quietly shape ASEAN's chip future

Last time, we looked at the ceiling over ASEAN's chip ambitions — the people, power and water that capital alone can't lift. This edition stays on those constraints and adds the factor most cost models don't capture: how the way the US and China are reorganising their supply chains is reshaping which sites actually win the work. Not as a verdict — as one more risk factor to weigh alongside water, power and talent when deciding where to build.

Start with what everyone knows: water, power, talent

Anyone in the field can run the resource comparison in a minute. Indonesia's Batam is rain-constrained, with raw-water supply barely grown in 25 years. Malaysia's Penang has the deepest semiconductor ecosystem in the region — but a structural, long-run water supply question that tightens every year demand climbs, leaning on river water piped in from neighbouring Perak. Vietnam's Haiphong sits in a water-rich delta where the problem is contamination, not volume, and pre-treatment is a solved, increasingly imported skill — Japanese players like Kurita have been quietly moving capability in.

Talent is the second axis, and it's tight everywhere. Malaysia is estimated to need on the order of ten times the engineers its universities graduate each year, and it loses a slice of them to Singapore annually. Vietnam has the larger talent pool today and lower wages, but a deeper structural question sits underneath: its total fertility rate has fallen below replacement, with population set to peak in the early 2040s. That is the fastest demographic shift among major ASEAN manufacturing bases — a long-term variable that nobody serious is ignoring. On power, Malaysia's grid is mature but increasingly stretched as fab and packaging investment piles into the same corridors; Vietnam's is improving but still catching up.

None of this is wrong, and a buyer should price all of it. But here's the trap: on resources alone, the comparison ends near a draw. Real investment decisions don't end at a draw. Something else is in the room.

The frontier moved — and the constraints got heavier

For years the mental model was simple: front-end wafer fabrication is the hard, thirsty, power-hungry crown jewel; back-end assembly and test is the lighter work ASEAN can absorb. That model is now out of date. As transistor shrinks slow and grow ruinously expensive, performance gains increasingly come from advanced packaging — stitching separate chiplets together and stacking high-bandwidth memory right beside the logic die (TSMC's CoWoS, 2.5D/3D integration). This is what makes a modern AI accelerator fast. In the industry's own framing, it's no longer a back-end afterthought but the frontier itself.

And here is the part that matters for the ceiling. Advanced packaging is not the light back-end work of the wire-bonding era. Wafer bumping and wafer-level packaging are wafer-scale processes — they drink cleaner water, draw heavy, uninterrupted power, and demand process engineers as deep as a fab's. The AI accelerators they build then compete with AI data centres for the very same megawatts. So the move up the value chain doesn't relax the people-power-water ceiling. It pushes back-end work toward front-end levels of constraint. The bottleneck didn't loosen. It climbed.

But cost and resources don't tell the whole story

If you keep the comparison to resources, you can't explain where the actual money is going. TSMC is spending roughly $165 billion in Arizona and has upgraded its second Kumamoto fab to a 3nm line (about $17 billion). ASE, the world's largest assembly-and-test firm, is running a record $7 billion 2026 capex with a major Penang expansion as one engine. Intel's roughly RM12 billion Penang advanced-packaging complex is set to begin operations this year. Taiwan's Chipbond just opened a ~$200 million wafer-bumping and WLCSP plant in Batu Kawan. Samsung, Foxconn, Luxshare, Goertek and many others run large operations across Vietnam.

None of those choices is the lowest-cost answer on a clean spreadsheet. On pure cost and water, you don't put a leading-edge fab in Arizona, and you don't crowd the new packaging frontier onto a single island that is itself tight on water. Yet that's exactly where the capital is landing. The reason isn't on the resource ledger. It's the geopolitical layer sitting on top of it.

The geopolitical factor — and how to read it

The next few sections sketch what is actually happening between the US and China — not as a story of winners and losers, but as a set of moves that, taken together, reshape which addresses foreign capital is willing to commit to. The aim is not a verdict on who is right. The aim is to give a site decision the same kind of clarity on this layer that it already has on cost.

What Washington is doing

The US approach has not been to confront China head-on but to reduce, over time and across several fronts at once, the inputs China would need to compete at the very frontier of chipmaking. It is, in effect, the logic of removing the firewood from under the cauldron — an old strategic idea, attributed to The Thirty-Six Stratagems, of weakening an opponent by draining what sustains them rather than fighting the heat directly. Three fronts run in parallel.

Technology. Export controls on extreme-ultraviolet lithography (EUV) and on advanced fabrication equipment are designed to keep the most advanced logic nodes outside China's reach. Made-in-China leading-edge production has advanced anyway — at high cost in yield, capital and time — but the gap on the frontier itself remains.

Energy. Sanctions and enforcement on cheap-discount crude routes that bypass the dollar system — most visibly Iran (which has long sent the bulk of its exported oil to China) and Venezuela (where a US naval blockade in late 2025 was followed by Maduro's removal in January 2026 and explicit licensing conditions tied to severing ties with China) — have squeezed two of China's lower-cost energy lifelines in the same window. Russian crude is what is left at scale.

Supply-chain location. The capital steering chip manufacturing into Arizona, Kumamoto, and parts of ASEAN is not pure cost optimisation. It is part of the same plan: physically disperse the world's most critical capacity away from a single contested island and into a set of trusted addresses.

From The Thirty-Six Stratagems: don't fight the steam, take away the fuel that makes it.

What Beijing is doing

None of this is happening to a passive opponent. China runs the world's largest industrial base and remains the largest single market for capital goods, vehicles and electronics. On the inputs the US is targeting, the moves are explicit.

On the technology front, SMIC has pushed older deep-ultraviolet (DUV) tools further than most foreign analysts thought possible, producing 7nm-class chips with multi-patterning at significant cost in yield and energy. Huawei has built its own AI accelerator family (Ascend) around what it can make domestically. Peking University researchers announced in early 2025 a 2D-transistor approach that aims to sidestep silicon scaling limits entirely. Whether those efforts close the frontier gap or not, they are a serious, well-funded attempt to do so.

On energy and finance, China has spent a decade preparing for exactly the squeeze now underway — building strategic reserves, settling oil purchases in yuan, and constructing alternative shipping and insurance networks. The squeeze still hurts, but China is not unprotected.

And on supply chains, China holds its own choke points. It refines the dominant share of the world's rare earths — the inputs almost everything downstream depends on, including the equipment that makes chips. Companies like DJI, CATL and BYD dominate categories outside semiconductors that the rest of the world has not figured out how to replicate. This is not a one-sided strangulation. It is a long, mutual war of attrition, with each side gripping the other's choke points. That framing is far more useful for anyone placing a bet in ASEAN than either of the comfortable extremes.

And ASEAN — Malaysia in particular — is where both sides meet

Here is where the picture turns specific. Malaysia's own ministers describe the country's draw as a "neutral geopolitical position" — and the word neutral turns out to be unusually literal. The capital pouring into Penang is not coming from one side alone.

From the US, Japan and Taiwan side, Intel's roughly RM12 billion Penang advanced-packaging complex is set to begin operations this year; Infineon is committing roughly $5 billion over five years to the world's largest 200-mm silicon-carbide plant in Kulim; ASE is running its biggest-ever capex year with Penang as a central engine; Chipbond's ~$200 million WLCSP plant just opened in Batu Kawan. Penang already handles roughly 13% of global semiconductor assembly, packaging and test (MIDA estimate) and aims for 15% by 2030.

From the China side, the same neighbourhood is being used too. Unisem — one of Malaysia's flagship packaging firms — has been majority-owned by China's Huatian Technology since 2018, providing Huatian with a USD/EUR-billed manufacturing presence outside the mainland (international revenue via Unisem reached roughly 35% of Huatian's total in 2025). xFusion, a former Huawei server unit, partnered with Malaysia's NationGate in 2023 to build a Penang Global Supply Centre — about 254,491 sq ft, with annual capacity above 150,000 GPU servers, aimed at covering most of xFusion's shipments outside China. Other Chinese fabless and design firms have been routing assembly work through Malaysian packaging houses, openly enough to be reported on by Reuters in late 2023.

Veterans inside the country are clear-eyed about where the gap still is. At SEMICON Southeast Asia 2026, HP Labs and Seagate veteran Tan Eng Tong told the audience plainly:

"Malaysia has zero. We have nothing."

That is the working arrangement, stated plainly. Malaysia hosts world-class foreign operations on every side of the geopolitical line, and a domestic process base that is still in early formation. Washington is aware of the cross-flow and has been tightening the perimeter selectively — additions to the US Entity List in March and September 2025 named procurement conduits routing restricted chips through Malaysia, Singapore and the UAE. But the broader trade has been allowed to continue, because the alternative — forcing Malaysia to pick a side cleanly — would damage the same supply chains the US is trying to rebuild.

The map underneath ASEAN

Step back to the regional level and the same logic sorts the rest of the map. ASEAN is not one bloc; member states sit at very different distances from China's economic orbit, and that distance — more than cost — is shaping which kinds of capacity each one is winning.

So how should a site decision actually weigh all this?

Pull the threads together and the practical picture comes into focus. Water, power and talent set the baseline — none of these bases has a clean answer on resources alone, and the rise of advanced packaging has made those constraints more binding, not less. On top of that baseline, the way US and Chinese capital are reorganising adds a second layer that is now decisive at the margin: which addresses are trusted by which side, which ones are openly used by both, and where the perimeter is being tightened.

For a real site decision, that means treating geopolitics as one more line in the diligence — alongside cost, water, power and talent — rather than as a separate, scarier conversation. Two readings of the region are worth holding side by side.

Malaysia, today. The deepest ecosystem, the most foreign capital arriving from every side, and the most visible signs of geopolitics actively shaping the map. It is being used as a working contact point by both blocs at once, which is a real strength in the short run and a real source of uncertainty in the long run, as the US perimeter continues to tighten selectively.

Vietnam, on the rise. Lower cost, more water, a younger workforce, and a deliberate "bamboo diplomacy" — bending with the wind, holding working relationships with Washington, Beijing and Moscow alike rather than leaning on any one. Packaging and back-end investment is now arriving (Hana Micron, Amkor, Intel and others have all announced or expanded major Vietnam operations). The profile rhymes in many ways with Malaysia circa 2010 — a base on the way up. But the strengths that make Vietnam attractive also carry their own exposures, and the demographic clock is real.

The honest takeaway is the frame, not a single pick. There is no one site that wins every dimension — and treating either Malaysia or Vietnam as a clean answer misses the point of the moment. The constraints are real, geopolitics is a real input alongside them, and the work is to weigh both, deliberately, for the specific decision in front of you. That map isn't in the spreadsheet, but it is one worth learning to read.

Next time, we'll look more closely at Vietnam — what makes its position genuinely interesting, and where the exposures sit underneath.