Singapore's semiconductor stocks: Selloff opens an entry point as the upcycle holds

The more durable signal is electronics. Electronic NODX rose 107.6%, led by disk media products (+312.4%), personal computers (+149.5%) and integrated circuits (+74.7%), and has now more than doubled YoY for three straight months.

Growth was broad-based, with NODX to Hong Kong (+112.1%), Taiwan (+96.9%), South Korea (+96.0%) and the US (+73.2%) all rising strongly.

For the first nine months, NODX is up 27.6%, well above Enterprise Singapore's full-year forecast of 14.0% to 16.0%. The Ministry of Trade and Industry raised its 2026 GDP growth forecast to 4.5% to 5.5% in August, and advance estimates show 3Q26 GDP growing 5.4% YoY, with manufacturing up 7.5%.

These exports are mostly finished AI hardware, such as storage, computers and chips. The SGX-listed semiconductor names sit one step earlier in the chain: their main driver is chipmakers’ spending on the equipment used to make that hardware, which SEMI expects to rise 23.1% in 2026 and 21.8% in 2027 as AI demand drives capacity additions. 

Three companies offer direct exposure to that spending, each from a different angle: UMS Integration (SGX: 558) through precision components for front-end deposition and etch equipment, AEM Holdings (SGX: AWX) through test and handling solutions for AI and high-performance computing (HPC) chips, and Frencken Group (SGX: E28) through mechatronics manufacturing for semiconductor equipment makers and other high-tech customers.

The mid-October selloff hit all three. On 15 October, the third straight day of declines, AEM fell 9.9% to SGD 8.80, UMS fell 10.0% to SGD 2.33 and Frencken fell 6.4% to SGD 2.20. In October to date, the three are down 13.5%, 14.9% and 16.4% respectively. The triggers were largely macro: US consumer prices rose at their fastest monthly pace in five months, lifting bets on further Federal Reserve hikes, while AI funding concerns resurfaced. Fundamentals look intact.

Industry data, however, point to sustained demand. TSMC, the world’s largest contract chipmaker, reported record 3Q26 revenue of USD 46.7 billion, up 41.2% YoY, on strong demand for its leading-edge processes, and expects capacity for its newest 2nm-class chips to more than double in 2027. Singapore’s electronic NODX also continues to more than double YoY.

Related article: AEM Holdings: Riding the AI/HPC test wave

Related article: UMS Integration: Placement-funded capacity lifts earnings, but the price has run ahead

Related article: Singapore's semiconductor stocks: Riding the AI-driven upcycle

1H26 earnings: strong at AEM and UMS, softer at Frencken

• AEM Holdings delivered the sharpest earnings inflection.

1H26 revenue rose 29.9% YoY to SGD 247.2 million, driven by a 191.7% surge from its fabless AI/HPC customer, now its largest. Profit before tax (PBT) margin widened to 15.5% from 2.1%, and attributable net profit rose to SGD 31.0 million from SGD 3.1 million. Recurring configurables and collaterals made up 77.9% of Test Cell Solutions revenue in 2Q26, giving AEM a steadier revenue base than in past cycles. Management raised FY2026 revenue guidance a second time to SGD 630 to 680 million and declared its first interim dividend for a corresponding period of 2.4 cents.

• UMS Integration delivered high-margin growth, with a second major customer now at scale.

1H26 revenue rose 25.2% YoY to SGD 156.5 million and attributable net profit rose 66.3% to SGD 33.4 million, with PBT margin of 24.3%, the highest of the three. Excluding FX gains and one-offs, pre-tax profit still grew about 39.4%. Revenue from Malaysia and South Korea, the markets tied to its new major customer (which we understand to be Lam Research), rose 81.9% to SGD 35.9 million, or 22.9% of group revenue. A placement of up to 52.9 million new shares in Malaysia will raise about SGD 138.6 million for automation in Penang and a new factory in Vietnam.

• Frencken Group's results were more subdued, though 2Q26 showed signs of inflection.

Revenue was broadly flat at SGD 427.8 million (-0.8% YoY), as softer Analytical Life Sciences demand in Europe offset semiconductor growth in Asia, and PATMI fell 3.4% to SGD 19.3 million. However, 2Q26 PATMI rose 13.7% YoY and 40.5% QoQ, suggesting 1Q26 was the trough. Management still guides for FY2026 revenue and profit above FY2025 and targets SGD 1 billion in revenue by 2028.

Table 1: Comparison across the three stocks

Outlook: a longer runway ahead

• AEM has the strongest growth outlook, and its pipeline is diversifying.

The AMPS platform backlog of over SGD 400 million is now split across two customers, the fabless AI/HPC customer and the PC/Foundry customer, and the Memory business ships its first Xact handler to a leading IDM in 4Q26 ahead of a 2027 ramp. AMD projects the data centre CPU market to grow more than 50% a year through 2030, while larger packages and longer test times favour AEM's thermal control and parallel test capabilities. Customer concentration has yet to ease, however, and as a high-beta AI/HPC proxy, the stock is likely to stay volatile.

• UMS has a broader, better-capitalised business, with capacity to take on more demand.

With management describing near-term components capacity as tight, the placement lets UMS serve demand its current footprint cannot, adding about SGD 26 million of revenue by FY2028E in our forecasts. Aerospace (+37.3% YoY in 1H26) adds a small second growth engine. SEMI expects wafer fab equipment spending to grow 23.1% in 2026 and 21.8% in 2027 before slowing to 14.1% in 2028, our valuation year. New product qualifications and advanced packaging are upside risks, though advanced packaging should pay off mostly beyond FY2028.

• Frencken offers a more gradual recovery.

Asian semiconductor growth should continue, while weakness in European life sciences persists through 2H26. Management expects 2H26 revenue to exceed both 2H25 and 1H26. Medical provides steady support, automotive radar could benefit from ADAS adoption, and a new Singapore facility due in 1Q27 supports longer-term growth.

Valuation: the selloff has reset multiples

The key change since our last sector update is valuation. It starts with share prices: even after the recent selloff, AEM, UMS and Frencken have returned about 413.8%, 108.5% and 61.1% year-to-date (Table 1).

Earnings forecasts have risen sharply too, but not as fast as share prices. We expect AEM's EPS to grow 389.1% in FY2026E from a low base, then 34.8% in FY2027E and 24.1% in FY2028E, against 72.2%, 29.1% and 22.0% for UMS, while consensus expects 10.9%, 14.1% and 18.1% for Frencken (Figure 1).

Figure 1: AEM is expected to deliver the strongest EPS growth over FY2026E to FY2028E

As share prices have risen faster than earnings, forward multiples expanded. By 13 October, AEM, UMS and Frencken traded at 38.0x, 33.7x and 23.7x consensus forward earnings. AEM and UMS were both more than 2.5 standard deviations above their ten-year historical averages and Frencken was 1.7 standard deviations above its own.

AEM's stronger earnings outlook provides some fundamental support for its outperformance, while Frencken's more modest gains are consistent with its slower expected growth. However, the scale of the gains across all three suggests investors are also pricing in greater optimism around the broader AI theme. We view this as a broad re-rating of SGX-listed semiconductor names, rather than one fully explained by each company's individual earnings outlook.

The selloff has since unwound part of that re-rating. As of 15 October, the three traded at 33.3x, 29.9x and 21.8x (Figure 2). That still leaves both AEM and UMS at around two standard deviations above their ten-year historical averages while Frencken is at around 1.3 standard deviations above its ten-year historical average. That said, multiples have compressed meaningfully from their peaks

Figure 2: All three still trade at around 1.3 to two standard deviations above their respective ten-year historical averages

Applying our fair P/E multiples to FY2028E EPS, we see upside of 24.2% for AEM and 9.0% for UMS (Table 2). Our fair multiples are above their respective ten-year historical averages of 13.1x for AEM and 15.0x for UMS. We believe this premium is justified by structural improvements in earnings quality. However, both fair multiples remain below their current forward P/Es of 33.3x for AEM and 29.9x for UMS. This means our upside projections rely on earnings growth through FY2028E rather than sustained elevated valuations.

We assigned AEM a higher fair P/E of 25x, versus 20x for UMS, given the greater resilience of its recurring configurables and collaterals revenue during the last downturn. In contrast, UMS’s capacity expansion is increasing its fixed-cost base, leaving earnings more vulnerable if equipment demand weakens. Its experience between 2022 and 2024, when EPS more than halved, illustrates how quickly earnings deterioration can undermine a premium valuation.

Table 2: Earnings table for AEM Holdings

Table 3: Earnings table for UMS Integration

Figure 3: AEM share price vs EPS chart

Figure 4: UMS share price vs EPS chart

Conclusion: the selloff opens an entry point, with AEM our preferred pick

The NODX data and upgrades to Singapore's growth forecasts reinforce that this upcycle has firm structural foundations. All three companies also sit in the iEdge Singapore Next 50 Index, which, together with MAS's SGD 6.5 billion Equity Market Development Programme (EQDP), could support greater institutional interest over time.

Even after the selloff, AEM and UMS have returned about 413.8% and 108.5% year-to-date. However, the pullback has compressed multiples and we view it primarily as a reaction to higher interest rate expectations. However, while rates may explain the near-term share price weakness, the more important fundamental question is whether concerns over AI funding will translate into lower AI capital spending. This would affect demand of AEM’s key customers, Intel and AMD, and UMS’s key customers, Applied Materials and Lam Research.

Those concerns have centred on the most leveraged parts of the build-out. CoreWeave, an AI cloud provider carrying about USD 50 billion of debt, saw its bond yields rise above 15% in mid-October, while SoftBank’s credit risk hit a four-year high after it borrowed heavily to complete its USD 64.6 billion investment in OpenAI. In other words, markets are questioning how the build-out is being funded, rather than whether demand for AI computing exists.

That distinction matters for our names. Most AI capital spending still comes from cash-rich hyperscalers, which are expected to spend more than USD 700 billion this year, and from chipmakers funding capacity largely from their own cash flows. TSMC, a major customer of both Lam Research and Applied Materials, reaffirmed its 2026 capex budget of USD 60 to 64 billion on 15 October, in the middle of the selloff, and expects capacity for its 2nm-class chips to more than double in 2027. The key test will be the hyperscalers’ 2027 capex guidance in their upcoming results, and any cuts there would be the main risk to our view.

• AEM Holdings (SGX: AWX), BUY (upgraded from HOLD), target price SGD 10.93. AEM offers the strongest earnings growth of the three and a growing recurring revenue base. At SGD 8.80, it trades at about 20.1x our FY2028E EPS, below our 25x fair P/E, implying 24.2% upside.
• UMS Integration (SGX: 558), HOLD (upgraded from TRIM), target price SGD 2.54. UMS has a second major customer at scale, placement-funded capacity and net cash. At SGD 2.33, it trades at about 18.3x our FY2028E EPS, implying 9.0% upside. Our HOLD reflects limited upside to our target price rather than a weaker view of the business.
• Frencken Group (SGX: E28), not rated. Frencken is the most defensive and diversified of the three and trades at the lowest multiple, but with flat 1H26 revenue and slower expected growth, that lower multiple does not in itself make it the cheaper stock.

Overall, we would use the selloff to add exposure selectively, with AEM our preferred pick. Volatility is likely to persist while US rate expectations remain unsettled and we will watch for any sign that AI funding concerns translate into lower AI capital spending, which would weigh on demand for all three. Investors may prefer to build positions gradually.

Note

iFAST Research rating system: iFAST Research employs a five-tier rating system: Buy (material upside potential, favourable risk-return); Accumulate (moderate upside, selectively add on weakness); Hold (limited upside, maintain existing positions); Trim (upside insufficient to justify a full position, reduce exposure on strength); and Sell (material downside risk, exit position).

Declaration:

For specific disclosure, at the time of publication of this report, IFPL (via its connected and associated entities) and the analyst who produced this report hold a NIL position in the abovementioned securities.

This research report was prepared with the assistance of artificial intelligence (AI) tools. iFAST Financial Pte Ltd does not rely exclusively on AI for content generation; the content of this report, including all investment theses, ratings, price targets and conclusions, has been independently reviewed and verified by the research analyst(s) to ensure accuracy and professional integrity.