UMS Integration Limited (SGX: 558) - Deep Dive Research Report

Prepared 18 June 2026. Most recent reporting period: Q1 FY2026, quarter ended 31 March 2026, results released 11 May 2026.

Section 1: What the Company Does

UMS Integration makes the metal guts of the machines that make computer chips. It does not make chips, and it does not make the chip-making machines either. It sits one layer below: it takes blocks of aluminium and other metals, machines them to tolerances measured in microns, cleans and coats them to semiconductor-grade purity, and then bolts hundreds of these parts together into large modules that get shipped to the company that builds the actual wafer-fabrication equipment. That customer, overwhelmingly, is Applied Materials, the largest semiconductor equipment maker in the world.

Think of it this way. When TSMC or Samsung builds a new fab, they buy deposition and etch machines from Applied Materials. Each of those machines is a refrigerator-sized assembly of precision-machined chambers, gas-delivery manifolds, frames, and fluid systems. Applied Materials designs the machine and owns the customer relationship, but it outsources a large share of the physical manufacturing and sub-assembly of those modules. UMS is one of the firms that does that manufacturing. It is a contract manufacturer and integrated component supplier to the front end of the semiconductor capital-equipment supply chain.

The business was built by Andy Luong, a Vietnamese-American machinist who founded UMS in Singapore in 2001 and still runs it as founder and CEO with more than four decades in the trade and roughly a 15% personal stake. The pivotal decision in the company's history was vertical integration. Rather than just machine parts, Luong built the capability to do the full sequence in-house: precision machining, surface treatment (anodising, electroplating, cleaning), and final integrated-system assembly and testing. That lets UMS deliver a complete, ready-to-install module rather than a box of components, which is exactly what a customer like Applied Materials wants when it is trying to compress its own lead times. In 2009 Luong expanded into Penang, Malaysia, building a 400,000 sq ft factory that has become the engine of the company's current growth phase. In 2018 UMS bought into SGX Catalist-listed JEP Holdings, adding an aerospace precision-engineering arm.

The value proposition is reliability under tight tolerance at scale. A semiconductor equipment maker cannot afford a contaminated chamber or an out-of-spec manifold; a single defective part can scrap a customer's wafer lot worth a fortune. UMS's edge is that it has been qualified, audited, and embedded in Applied Materials' supply chain for two decades, doing the full machine-clean-coat-assemble loop under one roof in low-cost Asian locations. That qualification, and the vertical integration behind it, is the moat.

Section 2: Business Segments

UMS reports three segments: Semiconductor, Aerospace, and Others. The first dominates; the other two are diversification.

Semiconductor (roughly 85% of revenue)

This is the company. In Q1 FY2026 it generated S$58.9 million of the group's S$69.4 million revenue. The segment has two internal pieces that management discloses separately and that behave differently:

• Integrated Systems: full module assembly. UMS takes machined parts (its own and bought-in) and assembles complete sub-systems for the customer's deposition and etch tools. This is the higher-revenue, more cyclical piece, tightly coupled to Applied Materials' own build schedule. In FY2024 it fell hard (down 33% to S$94.4 million) when AMAT destocked, then recovered (integrated-system sales leapt 36% year-on-year in Q2 FY2025).
• Components: precision-machined parts sold individually. This piece has been the steadier grower (up roughly 10-15% through 2025), increasingly driven by the new second customer rather than Applied Materials.

The core capability here is the qualified, contamination-controlled, vertically integrated production line. To machine a part, anodise it, clean it to particle-count spec, and assemble it into a module that passes a semiconductor OEM's incoming inspection takes years of audited process control. That is what would be hard to replicate. Management talks about this segment as both the cash cow and the growth bet - the Applied Materials base throws off cash, while the new customer ramp in Penang is the growth story.

Aerospace (roughly 10-11% of revenue, via JEP Holdings)

UMS holds about 80% of SGX Catalist-listed JEP Holdings, which makes precision-machined aerospace components - engine parts, structural components - for the commercial-aviation supply chain. In Q1 FY2026 aerospace revenue was S$7.3 million, up 18%. This segment exists because it lets UMS redeploy its core competence (precision machining of difficult metals to tight tolerance under certification) into a completely different, long-cycle, certification-gated end market that does not move with the semiconductor cycle. Aerospace runs on AS9100 and customer-specific qualifications, multi-year programmes, and the post-pandemic build-rate recovery at Boeing and Airbus. Management frames it as a diversifier and a counter-cyclical ballast against semiconductor volatility, and has described aerospace conditions as resilient.

Others (roughly 4-5% of revenue)

A small bucket: electroplating, anodising, distribution, and other engineering services (including the Starke surface-treatment business, of which UMS bought the remaining 30% for S$8.22 million in early 2026). In Q1 FY2026 it was S$3.2 million. Strategically it is a margin and capability add-on - owning the surface-treatment step keeps more of the value chain inside the group and protects the integrated-manufacturing pitch.

Section 3: Products and Business Detail

The product is not a catalogue item; it is a qualified manufacturing capability sold by the part and by the module. The meaningful "products" are:

• Precision-machined components: chambers, lids, flanges, manifolds, frames, and fixtures machined from aluminium and other alloys for deposition and etch tools. Tolerances run to microns. The hard part is not cutting metal; it is doing it repeatably at semiconductor-grade surface finish and dimensional stability.
• Surface-treated parts: the same components after anodising, electroplating, or precision cleaning, performed in-house (Starke). Surface chemistry matters because residue or particles inside a process chamber contaminate wafers.
• Integrated systems / modules: complete sub-assemblies of a wafer-fab tool, assembled and tested, shipped ready to install. This is the high-value end and the reason UMS is a "one-stop" supplier rather than a job-shop machinist.
• Aerospace components (JEP): certified precision parts for aircraft engines and structures.

What makes it hard to make: the combination of certifications (semiconductor OEM supplier qualification; AS9100 for aerospace), the contamination control (cleanroom assembly, particle-count specs), and the process knowledge to hold micron tolerances across high volume. A new entrant cannot simply buy machines and underbid; it has to be qualified part-by-part by the customer, a process that can take a year or more per part.

Manufacturing and geography. Production sits in three places. Singapore is the historical headquarters and integrated-systems hub. Penang, Malaysia is the growth engine - the original 400,000 sq ft plant plus a newer "Plant 2" dedicated largely to the new second customer, with utilisation estimated below 50% (substantial headroom). California (USA) gives a local footprint near US customers. The Penang expansion is the single most important operational story: management spent heavily (around S$35 million of capex in the run-up) building and equipping capacity for the new customer's volume ramp, and free cash flow fell sharply in FY2025 (to S$2.1 million from S$22.9 million) precisely because of that inventory and capex build.

Milestones that changed the business: founding in Singapore (2001); Penang expansion (2009); JEP/aerospace entry (2018); onboarding of the second major semiconductor customer and Penang Plant 2 volume start (2024); secondary listing on Bursa Malaysia (2025); bonus issues including the 1-for-4 issue allotting 177.6 million shares in January 2026, lifting the share base to about 888 million.

Section 4: Customers

UMS sells to a very small number of very large, very demanding customers. This is the defining feature of the business and its biggest single risk.

Who buys. The dominant customer is Applied Materials, which has historically accounted for roughly 80% of group revenue. The buying decision inside a customer like AMAT is made by supply-chain and commodity-management teams together with manufacturing engineering and quality - not by a single purchasing manager. Their criteria are qualification status, on-time delivery, yield/quality (particle counts, dimensional conformance), capacity to scale, and cost. Sales cycles for new parts are long because every part must be qualified; once qualified, the part tends to stay with the supplier for the life of the tool platform.

The second customer ("customer L"). UMS has onboarded a second major semiconductor customer, widely understood to be Lam Research, and ramping it through Penang Plant 2. Management has described this customer's order flow as robust and noted the customer is diverting US supply sources to Asia. Over the longer term UMS has pointed to an aspiration of roughly S$300-500 million of revenue from this customer over the next few years - a number that, if realised, would transform the revenue base and meaningfully dilute the Applied Materials dependence.

Why they buy from UMS. Vertical integration (machining + surface treatment + assembly under one roof), an established qualification record stretching back two decades with AMAT, and low-cost Asian capacity with proven ability to scale. When a customer wants to move supply out of the US into Asia, an already-qualified, already-scaled vendor in Singapore and Penang is the path of least resistance.

Switching costs. High, in both directions. For UMS, losing a qualified part means losing it for years. For the customer, requalifying a part at a new supplier is slow and risky - which is what protects UMS's installed base. The flip side: that same lock-in concentrates UMS's fortunes on a handful of relationships.

Concentration. Extreme. With AMAT historically near 80% of revenue, UMS's revenue is effectively a derivative of Applied Materials' build rate, which is itself a derivative of global fab capex. The new customer is the deliberate antidote: every dollar it adds both grows the company and reduces the concentration ratio. Contract structure is largely purchase-order / forecast-driven build-to-order rather than fixed long-term volume commitments, so revenue visibility runs a few quarters out on customer forecasts, not years.

Section 5: Competitive Landscape

UMS competes in the market for outsourced precision components and sub-assembly to semiconductor equipment OEMs. This is a real, structured industry with both Singapore peers and global players. It is not a commodity machine shop business, but it is also not a wide-moat monopoly - it is a qualified-vendor oligopoly where each OEM dual- or multi-sources.

The closest Singapore-listed peers are AEM, Frencken, Grand Venture Technology, and Venture Corp, though each has a different centre of gravity. AEM is really a semiconductor test-handler company (closer to the test/burn-in side and the most direct AI-test name), not a metal-parts machinist, so the overlap with UMS is partial. Frencken is the most diversified - semis plus medical, analytical, automotive, industrial - which makes it steadier but a less pure semi-equipment supply play. Grand Venture Technology is the most directly comparable in that it also machines precision components and sub-assemblies for semiconductor and life-science equipment makers. Globally, the more direct functional competitors are US-listed Ichor Holdings (gas and fluid delivery sub-systems for the same OEMs) and Ultra Clean Holdings (sub-system assembly and components), both of which serve Applied Materials and Lam Research and compete for the same outsourced module work.

UMS wins on vertical integration and low-cost Asian capacity close to where fabs and OEM Asian supply chains are concentrating; it is the natural beneficiary when a US customer wants to shift sourcing to Asia. It is exposed where it is most concentrated - against a diversified competitor like Frencken or a scaled US sub-systems house like UCT or Ichor, UMS's reliance on one anchor customer is a structural weakness, not a strength. The barrier to entry is the per-part customer qualification and the capital and process knowledge to hold semiconductor-grade tolerances; it is high enough to keep out casual entrants but low enough that each OEM maintains several qualified suppliers, which caps pricing power.

Market caps are approximate peer-size references as of June 2026 and move with the market.

Section 6: Industry

UMS sits two steps removed from chip demand: chip demand drives fab capex, fab capex drives equipment orders at Applied Materials and Lam Research, and those orders drive UMS's build. So the industry that matters is the wafer-fabrication equipment (WFE) cycle.

Demand drivers today are dominated by AI. Build-out of AI compute requires leading-edge logic and large volumes of high-bandwidth and advanced memory, which require deposition and etch capacity - exactly the tool types UMS's customers build. Advanced packaging (the assembly of chiplets and HBM stacks) is an additional, newer driver that adds deposition and etch steps. Management and its customers have pointed to expectations of over 20% growth in semiconductor equipment in 2026, with the AI-driven capex cycle expected to run through roughly CY26-28.

The global WFE market is large (tens of billions of US dollars annually) and concentrated among a handful of OEMs - Applied Materials, Lam Research, Tokyo Electron, ASML, KLA. UMS sits in the outsourced-manufacturing tier beneath them. A structural tailwind specific to UMS is the relocation of supply chains into Asia: as US OEMs diversify sourcing away from the US for cost and resilience reasons, qualified Asian vendors with spare capacity (UMS's Penang Plant 2 at sub-50% utilisation) are positioned to capture share.

The industry is cyclical and historically sharp. The WFE cycle swings with memory pricing, fab utilisation, and customer inventory. UMS lived through this in FY2024, when revenue fell 19% and net profit fell 32% as Applied Materials worked down inventory. Regulation matters at the edges - export controls on advanced equipment to China can reshape where and how much equipment ships - but UMS's direct exposure is mediated through its OEM customers rather than borne directly. The certification regime (OEM supplier qualification, AS9100 for aerospace) is itself a barrier that shapes who can play.

Section 7: Growth Triggers

Drawn from the six most recent reporting periods (FY2024 results Feb 2025; Q1 FY2025 May 2025; Q2/1H FY2025 Aug 2025; Q3/9M FY2025 Nov 2025; FY2025 Feb 2026; Q1 FY2026 May 2026).

• New second customer ("customer L") volume ramp in Penang Plant 2. Repeated across all six periods. Volume production commenced in FY2024; management has consistently described order flow as robust, driven by the customer diverting US supply sourcing to Asia. (FY2024 results, Feb 2025; reaffirmed Q1 FY2026, May 2026)

  Order flow from the group's new key customer remains robust, as it seeks to divert its US supply source to Asia. (Q1 FY2026 commentary, May 2026)

• Penang Plant 2 capacity headroom. Utilisation estimated below 50%, with the plant predominantly dedicated to the new customer - implying substantial room to grow revenue without major new building capex. (Q1 FY2026, May 2026)

• Long-term revenue ambition of roughly S$300-500 million from the new customer over the next few years. A scale target that would materially rebalance the customer mix away from Applied Materials. (Q1 FY2026, May 2026)

• AI-driven semiconductor equipment demand through CY26-28. Management cited customers achieving record Q1 results and projecting over 20% growth in semiconductor equipment for 2026, driven by AI and advanced packaging. (Q1 FY2026, May 2026)

  Management notes key customers achieved record Q1 results and project over 20% growth in semiconductor equipment for 2026. (Q1 FY2026 commentary, May 2026)

• Advanced-packaging participation as a medium-term driver. Participation in customers' next-generation advanced-packaging programmes flagged as an additional growth lever beyond the core deposition/etch work. (FY2025 results, Feb 2026)

• Aerospace recovery via JEP. Aerospace described as resilient, with revenue up 18% in Q1 FY2026; the commercial-aviation build-rate recovery underpins the segment. (Q1 FY2026, May 2026)

• Starke full ownership. Acquisition of the remaining 30% of Starke (S$8.22 million) consolidates the surface-treatment value chain in-house. (Q1 FY2026, May 2026)

Section 8: Key Risks

Customer concentration on Applied Materials. The mechanism is direct: with AMAT historically near 80% of revenue, any cut in AMAT's build rate - whether from a WFE downturn, an AMAT inventory correction, or AMAT in-sourcing or reallocating work to another vendor - flows almost one-for-one into UMS's top line. This is not hypothetical. FY2024 revenue fell 19% and net profit fell 32% on exactly this dynamic as AMAT destocked. It is a high-probability moderate-to-severe drag whenever the cycle turns.

Execution risk on the new-customer ramp. The entire growth thesis rests on Penang Plant 2 filling up with the second customer's volume. The mechanism that hurts: if the customer's orders disappoint, qualification slips, or the diversion of US sourcing to Asia stalls, UMS is left with an under-utilised, capital-heavy plant and the depressed free cash flow already visible in FY2025 (S$2.1 million, down from S$22.9 million) without the revenue payoff. Management has set a high bar with the S$300-500 million ambition, which makes any shortfall conspicuous.

WFE cyclicality. The semiconductor equipment cycle is sharp and not fully in management's control. An AI-capex air pocket, a memory-price collapse, or a fab-utilisation downturn would hit both customers at once. The 2024 experience shows the amplitude.

Geopolitics and export controls. UMS does not sell restricted equipment directly, but its customers do. Tighter US controls on advanced-equipment shipments to China, or tariff actions, could reduce the order pool or reshape where customers want parts built. The "divert US supply to Asia" tailwind is itself a product of this geopolitical reshuffling and could reverse if policy pushes sourcing back onshore.

Margin and working-capital pressure during ramp. Heavy inventory build and capex to support the ramp have already compressed free cash flow and net cash. If the revenue lags the spend, returns metrics deteriorate before they recover.

Share-count creep from bonus issues. Successive bonus issues (the 1-for-4 in January 2026 added 177.6 million shares, taking the base to about 888 million) steadily expand the share count. They are economically neutral per shareholder, but they mean per-share metrics are measured against a moving, growing base, and they are not the same as returning capital.

Section 9: Walk the Talk

The six reporting periods used: FY2024 (Feb 2025), Q1 FY2025 (May 2025), Q2/1H FY2025 (Aug 2025), Q3/9M FY2025 (Nov 2025), FY2025 (Feb 2026), and Q1 FY2026 (May 2026). The most recent is within ~90 days of today.

Start at the bottom of the cycle. At the FY2024 results in February 2025, with revenue down 19% and profit down 32%, management's central promise was that the new Penang facilities were "largely completed and operational," that volume production for the new customer had "commenced," and that FY2025 prospects were "bright" as the ramp accelerated. This was a concrete, checkable claim about a plant and a customer.

Through 2025, the claim held up directionally but unevenly. Q1 FY2025 (May 2025) delivered the first proof point: semiconductor component sales rose 19% specifically on the new customer, even as integrated-system sales dipped on resolved supply issues. Q2/1H FY2025 (August 2025) was the strongest validation - quarterly revenue up 20%, the semiconductor segment up 27%, integrated-system sales leaping 36%, and Malaysia revenue up around 270% as the new-customer orders landed. Management's "bright FY2025" framing looked vindicated mid-year.

Then came the honest wobble. Q3/9M FY2025 (November 2025) showed Q3 revenue down 9% year-on-year, with the semiconductor segment down 8% and aerospace down 16% in the quarter, even as nine-month figures stayed positive and gross margin actually improved to 58.2%. Management did not pretend the quarter away; it kept investing (S$35.4 million of capex in Penang) and proposed a 1-for-4 bonus issue, signalling confidence through the soft patch rather than retreating. The full-year FY2025 result (February 2026) landed modestly positive - revenue up 3.7% to S$251.1 million, net attributable profit up 2% to S$41.6 million - and the board held the final dividend at 2 cents, a tangible "we meant it" gesture after a year where free cash flow had collapsed to fund the ramp.

The payoff arrived in Q1 FY2026 (May 2026): revenue up 20% to S$69.4 million and net profit up 43% to S$14.0 million, with the new-customer order flow described as robust and customers projecting 20%+ equipment growth. The promise made fifteen months earlier - that the Penang ramp would convert into growth - was delivered, with a lag and a mid-cycle dip in between.

The verdict: this is management that does roughly what it says, on a semiconductor cyclical's timeline rather than a straight line. Luong did not over-promise a smooth ramp; he promised a ramp and a brighter year, kept investing through the Q3 soft patch, held the dividend when cash was tight, and the result came through. The one place to watch is the new S$300-500 million new-customer ambition - that is a bigger, longer-dated promise than anything delivered so far, and it is the commitment future calls should be measured against.

Section 10: Shareholder Friendliness Index

Dividends. UMS is a consistent dividend payer. Full-year dividends were around 5.0 Singapore cents per share in FY2025 and about 5.2 cents in FY2024, on the back of a stronger FY2023 (the high-water year before the FY2024 downturn). The pattern is a quarterly interim of 1 cent plus a 2-cent final, held flat into FY2026 (a 1-cent first interim declared with Q1 FY2026, payable 24 July 2026). The dividend has been broadly held rather than grown, which is notable given the FY2024-FY2025 earnings dip and the heavy Penang capex - management chose to maintain the payout even as FY2025 free cash flow fell to S$2.1 million, pushing the payout ratio high (around 91% of earnings and well above 100% of free cash flow in 2024). That is a deliberate signal of commitment to shareholders, but it also means the dividend is currently being sustained partly from the balance sheet during the investment phase, not fully covered by free cash.

Buybacks and dilution. MoatMap recorded zero buybacks in the trailing ~90-day window (since 20 March 2026). Looking back further, UMS carries the customary annual SGX share-purchase mandate renewed at its AGM, but it has not run a large, sustained repurchase programme over the last three years - capital has gone to dividends and Penang capex, not buybacks, and no material multi-year buyback execution is documented in the filings or financial news searched. The more important share-count fact runs the other way: the count has been growing, not shrinking, because of bonus issues - the 1-for-4 bonus in January 2026 added 177.6 million shares, lifting the base to about 888 million. Bonus issues are economically neutral (every holder's slice is unchanged), but they are the opposite of retiring stock, and per-share figures are measured against a steadily larger base.

Verdict: Returns Capital (via dividends) / Neutral overall - a reliable, maintained dividend is the friendly signal; the absence of buybacks and the recurring bonus issues mean shares are being created, not retired, so the capital return runs through the dividend line alone.

Section 11: Insider Activities

Singapore (SGX) insider disclosures are filed via SGXNet, which is gated to automated search. Per the canonical data source for this venue, MoatMap's nightly scrape recorded zero insider transactions for 558.SI in the trailing 12-month window (data current 18 June 2026).

That empty window is itself informative when set against the founder's well-documented prior buying. Andy Luong, founder and CEO and the largest shareholder, has a multi-year track record of open-market purchases that lifted his stake to roughly 109.2 million shares (about 15.4% at the time, pre the latest bonus issue). The documented buys - including a 600,000-share purchase for about S$589,500, a 199,800-share purchase, and a ~100,000-share purchase at around S$1.15 - cluster in 2024 and around May 2025, i.e. at or just before the start of the trailing 12-month window. In other words, the founder was an active accumulator into the cyclical trough and the early new-customer ramp, but there is no recorded open-market insider transaction inside the last 12 months in the canonical feed.

Net assessment. On the canonical 12-month window the signal is neutral - no recorded insider buying or selling. The broader, longer-dated picture is mildly constructive: insiders collectively own roughly a quarter of the company and the founder has historically bought rather than sold, with his most recent documented purchases falling just outside this window during the period when the stock was depressed and the Penang thesis was unproven. There is no evidence of insider selling, which for a founder-led company through a sharp earnings recovery is itself a quiet positive. Readers wanting confirmation of any newer activity should consult SGXNet "Change in Interest of Director/CEO" and "Substantial Shareholder" notices directly, as that portal is not machine-readable here.

Section 12: Scenarios

Bull case. The AI capex cycle runs hot through 2026-2028 just as management's customers projected, and Applied Materials and the new customer both keep building. Penang Plant 2, today under half full, fills steadily as the second customer executes its plan to shift US sourcing into Asia. The new-customer revenue marches toward the S$300-500 million ambition, and in doing so quietly solves UMS's oldest problem: customer concentration drops from ~80% on Applied Materials toward something far more balanced, and the market re-rates the business from a single-customer AMAT proxy to a diversified, AI-levered Asian sub-systems champion. Advanced-packaging programmes add a second leg of deposition and etch content. Aerospace, riding the airframe build-rate recovery through JEP, throws off steady counter-cyclical cash. Operating leverage on the now-utilised Penang capex sends margins and free cash flow back up, and the maintained dividend, once stretched, is comfortably covered again. The company that spent 2024-2025 investing through a trough emerges as a structurally larger, less concentrated business.

Base case. The most likely path is the lumpy continuation of what the six calls already show. The AI-driven equipment cycle stays broadly supportive, so revenue grows but in the saw-toothed pattern of a WFE supplier - strong quarters like Q2 FY2025 and Q1 FY2026 punctuated by soft ones like Q3 FY2025 as customers adjust inventory. The new customer keeps ramping and gradually dilutes the Applied Materials weighting, but reaches the S$300-500 million ambition slowly and only partly within the visible horizon. Penang utilisation climbs off sub-50% without snapping to full. The dividend is held, capex normalises as the plant matures, and free cash flow recovers from its FY2025 lows. Management continues to do roughly what it says on a cyclical's timeline. Nothing breaks; nothing dramatically surprises to the upside.

Bear case. The AI capex enthusiasm cools or an inventory correction hits, and because UMS is still ~80% tied to Applied Materials, the downturn lands almost undiluted - a repeat of FY2024's 19% revenue and 32% profit falls, but now with a half-empty, capital-heavy Penang plant dragging on returns. The new customer disappoints: orders that were "robust" slow, qualification of additional parts stalls, or the geopolitical logic that was pushing US sourcing into Asia reverses under onshoring pressure and tariffs, leaving Plant 2 stranded. Free cash flow, already thin, stays compressed by the inventory and capex overhang, and the maintained dividend starts to look like it is being funded from the balance sheet rather than earnings, forcing a hard choice between holding the payout and protecting the cash buffer. The concentration that is a strength in good times becomes the whole problem in bad ones.