Would you believe a stock on SGX is up roughly 5x this year, riding the same AI wave as Nvidia and Micron, and most people barely noticed?
That stock is AEM Holdings.
They’re not the obvious plays. You won’t find them designing chips or training models. Definitely not the sexy stuff. Instead, they sit quietly inside the value chain, making the parts, the testing gear, and the consumable tools that the entire industry runs on.
And AEM is just the start.
What’s interesting is that many of these companies aren’t new stories. They’ve been around for years, steadily building capabilities, expanding customer relationships, and strengthening their business models.
The gap, more often than not, has been in how they are perceived.
But that may be starting to change. As the AI narrative broadens beyond just the obvious winners, more attention is shifting towards the companies that enable the entire ecosystem to function. And for investors, that shift matters. Because this is where the hidden gems live: companies poised to ride the next wave, but still flying under the radar.
How Singapore Ended Up with a Semiconductor Cluster
This didn’t happen by accident.
Singapore’s electronics story goes back to the 1960s and 70s, when MNCs set up shop and quietly transferred technology to local suppliers. By the 80s and 90s, we became a global hard disk drive hub, building up deep capabilities in mechatronics and prevision engineering.
In the mid-90s, JTC launched dedicated Wafer Fab Parks. Today they host the likes of GlobalFoundries, Micron, SSMC and UMC. Then in 2016, the government rolled out the Precision Engineering Transformation Map to push local players further up the value chain.
The result was a dense precision-engineering ecosystem wrapped around the semiconductor cluster. The four companies below are products of that ecosystem.
They’ve just been quiet about it, which is precisely why many of them have remained under the radar, despite sitting on critical parts of the global semiconductor value chain.
Here’s the simplest way to think about where each one sits.
| Company | Ticker | What they do | Key customer |
| AEM | AWX | Semiconductor test handlers (burn-in, functional, system-level test) | Intel (plus new names) |
| UMS | 558 | Precision front-end equipment parts & assembly | Applied Materials |
| Frencken | E28 | Diversified precision contract manufacturing | ASML (likely largest) |
| Micro-Mechanics | 5DD | Consumable tools & wafer-fab parts | Broad, the assembly & test houses |
Notice the pattern. Each is a specialised supplier to a blue-chip name you’ve definitely heard of. They don’t compete with the giants. They enable them.
There’s a second thing they share, and it matters more than most people think. These local names actually pay you to wait. AEM, UMS, Frencken and Micro-Mechanics all pay dividends, and because Singapore runs a one-tier tax system, those payouts reach you completely tax-free. Compare that with many US AI names, which tend to pay little or nothing, and where a dividend a Singapore investor does receive is reduced by 30% withholding tax before it even lands in the account. Same AI theme, very different income profile.
These may not be the companies making headlines, but they are deeply embedded in how the industry functions, and that’s worth paying attention to.
Let’s go through them.
AEM (SGX:AWX)
AEM sits in semiconductor testing, designing and building the equipment that stress-tests chips under real operating conditions. Its acquisition of CEI has also strengthened its manufacturing capabilities and integration across the value chain.
Now here’s the AI connection most people don’t realise. AEM’s gear doesn’t just check if a chip works. It tests chips at the actual heat and power levels they’ll face in the field, especially in the punishing thermal environment inside an AI server rack. Its differentiating technology, PiXL™, is all about thermal precision at very high power, done at scale so testing stays cost-effective.
Why does that matter? AI chips run hot. As chips become more powerful and energy-intensive, the ability to test them under these conditions becomes increasingly critical.
AEM doesn’t just sell the machines. A portion of its revenue comes from recurring streams such as consumables and services, which improve earnings visibility, something that the market tend to reward.
Intel is AEM’s biggest customer, but that’s also been one of the key overhangs on the stock. For years, investors viewed AEM as heavily dependent on a single customer, which limited how it was valued.
That is now changing. AEM has been actively diversifying beyond Intel, landing new customers, including a new fabless AI/HPC customer that is ramping quickly and could become its largest revenue contributor.
At the same time, AEM also struck a strategic partnership with ASE, the world’s largest OSAT (the firms that assemble and test chips for everyone else), which gives AEM access to a much broader ecosystem. More importantly, it creates a pathway to hyperscalers building their own AI chips, opening up a customer base that was previously out of reach.
Recent results reflect this shift, with strong growth and upgraded guidance.
Put together, these developments are meaningful. AEM is moving away from being a single-customer, cyclical supplier and towards a more diversified AI infrastructure enabler with multiple growth drivers. As that transition continues, both the business profile and how the market looks at it may start to shift.
One point to note: competitor Advantest has an ongoing patent suit against AEM, which AEM denies. The outcome remains uncertain and is worth monitoring.
UMS Integration (SGX:558)
If AEM is the back-end testing play, UMS is the front-end equipment play.
UMS makes high-precision components and provides electromechanical assembly for semiconductor equipment makers. Its anchor customer is Applied Materials, one of the world’s largest chip-equipment supplier, for whom UMS builds precision parts used in key semiconductor manufacturing equipment. That single relationship has historically driven more than half of UMS’s revenue.
That used to be the whole story, and the obvious risk. Lean on one customer and you live or die by their orders.
The big change over the past year is that UMS now has a second key customer ramping hard, a US chip-equipment maker shifting more of its supply base into Asia. UMS is busy qualifying a string of new products for this customer, and it is already starting to move the needle.
UMS didn’t wait for the market to worry about customer concentration. It went and fixed it. Customer concentration is one of the main reasons investors assign lower valuations to otherwise solid businesses. By bringing in a second major customer and scaling that relationship, UMS is actively reducing this risk.
You can see the progress in the numbers. In the first quarter of 2026, revenue jumped 20% to S$69.4 million and net profit attributable to shareholders leapt 43% to S$14.0 million, with growth across all core segments. Even its aerospace arm, which rides the global aviation recovery, climbed 18%. UMS stayed in a net cash position and continued paying dividends, holding to its long track record of distributing earning through a cyclical industry.
Looking ahead, the demand backdrop remains supportive, with semiconductor equipment spending expected to remain elevated and supply chains continuing to shift toward Asia.
If this second growth engine continues to scale, UMS starts to look less like a single-client supplier and more like a diversified semiconductor partner. And in this space, that distinction tends to matter more than it first appears.
One point to note: first-quarter operating cash flow turned negative, mainly due to higher working capital and time-related factors. Management indicated that delayed payments have since been resolved,but it is something worth watching.
Frencken (SGX:E28)
Frencken is a precision-engineering company and contract manufacturer, with most revenue coming from its mechatronics division. The difference is diversification. Frencken’s end products span semiconductor tools, medical devices and analytical instruments, with a separate arm handling automotive and industrial work. So it isn’t living and dying purely on chips.
Its customer list reads like a who’s who: Philips Healthcare, GE Healthcare, ASML, Teradyne, Seagate, Agilent, Continental, Bosch, even NASA’s Jet Propulsion Laboratory. The biggest is most likely ASML, the EUV lithography monopoly that sits at the absolute heart of advanced chipmaking. That’s the AI hook. You can’t make leading-edge AI chips without ASML’s machines, and Frencken supplies into that ecosystem.
Now for the honest part. Frencken’s recent performance has been softer, with revenue and profits affected by weaker orders in parts of its business. So this is not a stock firing on all cylinders right now.
But look under the hood and the picture is more encouraging. The Asian side of Mechatronics business continued to grow well on the back of semiconductor demand, while its medical and automotive segments also held up.
This is where Frecken’s diversification starts to matter. By operating across multiple end-markets, the company reduces its dependence on any single cycle, even if that sometimes means near-term performance looks uneven. Over time, this kind of balance can lead to more resilient earnings, which is something longer-term investors tend to value.
The balance sheet also remains healthy, with a net cash position providing flexibility to invest through the cycle. Management expects both revenue and profit to improve over the full year, with semiconductor orders in Europe recovering from the second half of 2026.
The road ahead is where Frencken gets interesting. Three forward catalysts stand out.
First, its automotive radar antenna business is approaching an inflection point and is expected to ramp from the second half of 2026. Second, Frencken is moving into robotics, developing components such as gearboxes for service robots and humanoids. Third, it is expanding capacity in Malaysia and building a new Singapore facility as it pushes toward the next phase of growth. Frencken isn’t just riding the AI wave. It’s actively building the next leg of growth.
Taken together, these moves point to a company that is gradually building multiple growth pillars. It may not always show up quarter to quarter, but overtime, the balance and optionality can become more visible.
Micro-Mechanics (SGX:5DD)
While AEM, UMS and Frencken sell big-ticket equipment and parts, Micro-Mechanics sells the consumables. It designs and manufactures the miniature tools used in semiconductor assembly and testing (die-attach, wire-bonding, encapsulation), plus precision parts for wafer-fabrication equipment.
Consumable tools wear out. Every chip that gets assembled and tested gradually wears down tooling that has to be replaced. So instead of relying on a lumpy capex cycle, Micro-Mechanics is tied more directly to chip production volumes. The more chips the world makes, the more its tools get consumed. It’s the classic razor-and-razorblade model with recurring demand built in.
That kind of model tends to offer more visibility and resilience compared to equipment makers, which is something the market often rewards over time.
The latest results reflect this. Net profit rose in the latest quarter and for the nine-month period driven by continued growth in its consumables segment. Margins remain strong, return on equity is above 20%, and the company continues to operate with a net cash balance sheet while generating positive operating cash flow.
There’s also a quieter strength here, diversification. Micro-Mechanics serves more than 600 customers across multiple geographies, rather than relying on a single major account. Its largely local-for-local manufacturing footprint also provides some buffer against tariffs and geopolitical friction, a useful advantage in today’s environment.
The road ahead is about getting closer to where the action is.The company is positioning itself closer to where future demand is emerging, building capabilities in key regions such as China, Taiwan and the US. At the same time, it continues to improve its product offerings through new materials and process enhancements that increase efficiency for its customers.
Through all this, Micro-Mechanics has consistently returned cash to shareholders. Micro-Mechanics has quietly built a business that combines steady demand, strong margins and consistent capital returns. It’s not flashy, but in many ways that consistency is the point.
The Takeaway
Each one is a real, profitable company with real customers and improving fundamentals, riding the same wave from a different spot in the AI value chain. AEM tests the chips. UMS and Frencken build the equipment that makes them. Micro-Mechanics supplies the consumables that keep the lines running.
And the wave itself is big. The global semiconductor industry hit a record in 2025 and is on track to cross US$1 trillion in sales in 2026, with the growth led by AI and high-performance computing. Industry leaders describe it as a multi-year giga cycle, not a one-year spike, with memory and advanced packaging as the tightest bottlenecks. If that holds, the companies feeding that supply chain have a long runway.
It helps, too, that these names are getting more attention at home. Singapore’s market is in the middle of a broader push to spotlight on quality small and mid-caps under its Value Unlock programme, and this cluster of specialised, blue-chip-serving suppliers is exactly the kind of business that effort is surfacing. SGX is encouraging companies to sharpen their corporate strategy and improve disclosures and investor engagement. The goal is to generate a better appreciation of the value in these companies among investors, which in turn should lead to better market valuations.
Many of these companies are not just riding the cycle, but also evolving – diversifying customers, building more recurring revenue, and positioning themselves closer to where new demand is forming. That may not show up fully at first, but over time it can change how a business is understood and valued.And don’t forget the income angle. The US AI giants ask you to pay up and wait for capital gains alone, while these four hand you a dividend along the way, tax-free in Singapore. That is a meaningful edge for anyone who likes getting paid to hold.
None of this makes them sure things. The AI wave is lifting all boats and Singapore’s own AI stocks have benefitted. No one knows how long this wave will last, but the rising capital expenditure from the hyperscalers does suggest that the momentum remains strong, and that is translating into improving business fundamentals for these stocks.
But the next time someone mentions AI stocks, you might find yourself thinking of more than just Nvidia.
Disclosure. This article is sponsored as part of SGX’s market development initiative. However, all opinions expressed are solely those of the author and do not constitute financial advice.
