On June 5, the first Malaysia-built XPeng G6 rolled off a line in Melaka. Not in Guangzhou — in Pegoh, inside a plant owned by EP Manufacturing Berhad, a company that spent forty years making automotive parts, not assembling cars. Ten weeks earlier, the same EPMB orbit had started turning out the MG MGS5 a few bays over. We didn't read either as a car launch. We read them as a stress test, landing on the exact kind of plant we built Foreman for.
We've written before about why we started in Malaysia — the plants are close enough to walk into, the MES is already paid for, and the engineers are too honest to let 'almost works' slide. What we underpriced eighteen months ago was how fast the demand side would move. It's moving now, and it's moving on the floor, not in a keynote.
What actually happened, in ninety days
- June 5, 2026 — the first Malaysia-assembled XPeng G6 came off EP Manufacturing Berhad's CKD line at HICOM Pegoh in Melaka: component kits shipped from China, completed locally. The X9 MPV followed in late May. It is XPeng's second assembly hub in Southeast Asia after Indonesia, and the Melaka plant's annual capacity has been expanded to 30,000 units.
- March 2026 — the MG MGS5 EV (five-star Euro NCAP and ASEAN NCAP) started rolling off the PEPS-JV facility in Melaka, in partnership with the same EPMB. Officially launched April 24, with 1,500–2,000 units planned for the year.
- Up in Kulim, Inokom — a Sime Motors plant — is already CKD-assembling the Chery Omoda E5 and the BMW i5. Three marques, one assembly footprint.
- BYD's own plant in Tanjung Malim — 600,000 square metres, announced last August, production targeted for the second half of 2026 — is now hanging in the air. BYD balked at MITI's export-ratio and floor-price conditions, and its leadership has been seen walking Inokom's Kulim line instead. As of June, the path isn't settled.
Read together, that is a wave of Chinese-marque EV programs being localised in Malaysia inside a single year. And the companies absorbing the wave are not the OEMs. They are the contract assemblers and the parts makers underneath them — the Tier-1 and Tier-2 plants that have spent decades shipping components and are suddenly being asked to run multi-brand vehicle programs on compressed timelines.
What this does to a Tier-2 plant
A Tier-1 like EPMB doesn't get to specialise anymore. In the same quarter it is running a Chinese SUV program, a Chinese MPV program, and a European-rated EV — each with its own OEM, its own quality gate, its own scorecard, its own ramp curve. Now push that down a tier. The stamping shop, the injection-moulder, and the wire-harness maker feeding those lines used to serve two stable programs. They are being pulled into four volatile ones.
- More programs, more scorecards. Every new marque arrives with its own PPM target, its own OTIF definition, its own 8D template. A plant that tracked one customer's quality now reconciles four — by hand, in four spreadsheets, every week.
- Shorter ramps. A CKD program that localises inside a year doesn't hand a supplier two quiet quarters to stabilise. Yield problems that used to surface over months now surface over shifts.
- Thinner slack. Volume up, mix up, changeovers up — and the same Monday OEE meeting trying to cover all of it. The meeting was already the bottleneck. Now it's underwater.
- Policy volatility runs downhill. BYD's stalled plant is the tell: the program a supplier tooled up for can move three hundred kilometres north, or slip two quarters, on a condition negotiated in Putrajaya. The OEM absorbs that as a press cycle. The supplier absorbs it as stranded capacity.
Why this is an argument for a decision layer, not another dashboard
A plant running one stable program can survive on a wall of dashboards and a Monday meeting. A plant running four volatile programs cannot. There are not enough engineering hours in the week to export four customers' data into four spreadsheets, reconcile four scorecards, chase four ramp curves, and still walk the floor. The work the Monday meeting used to absorb now overflows it — and no amount of better charts fixes a capacity problem.
This is the gap Foreman was built for, and the localisation wave is widening it faster than our own slide deck predicted. An agent that reads the MES the plant already runs, watches every program at once, drafts each OEM's scorecard narrative before the QBR, flags the OTIF risk on a shipment before it slips, and ranks the root cause on a yield drift before the next shift repeats it — that is not a luxury when the program count triples. It is how the same headcount keeps up.
The solutions we built for inside the plant — quality intelligence, the OEM scorecard agent, OTIF risk, in-shift OEE recovery — were drawn from plants running two or three programs for customers they had served for years. The plants we walk into now are running more, faster, for OEMs they had never heard from twelve months ago. The shape of the problem didn't change. The volume did.
Reading this from the floor
We bet the company on a thesis: that the most honest place to build a manufacturing decision layer is on top of the plants that already have the data and are about to be asked to do far more with it. The Malaysian EV-assembly wave is that thesis arriving ahead of schedule. Malaysia first was never 'start small.' It was 'start where the conditions are most honest, and let the rest follow.' The conditions just got harder, which is exactly the point.
None of this needs a greenfield factory or a rebuilt data stack. The plants turning out XPeng and MG bodies this quarter are the plants that were stamping parts last year, on the MES they already paid for. The job isn't to replace any of that. It is to put an agent on top of it that can hold four programs straight when the engineers can only keep two in their heads at once.