SG Semicon Plunged 30%: Why I’m NOT Selling AEM, UMS & Micro-Mechanics

After a remarkable run, local semiconductor heavyweights AEM Holdings (SGX: AWX), UMS Integration (SGX: 558), and Micro-Mechanics (SGX: 5DD) have taken a sharp beating, falling 25% to 30% from their mid-year peaks.

Beyond a natural cooling off from months of semiconductor euphoria, investor sentiment was further rattled this week by breaking reports of China’s breakthrough in domestic DUV lithography alongside rapid DRAM capacity expansion led by CXMT (SSE: 688825)

These developments have stoked fears of future market oversupply, triggering a sharp sell-off across global semiconductor leaders that is now echoing down local supply chains.

It is natural to feel alarmed by such a steep decline, and some investors may panic-sell to lock in whatever gains remain. But while the volatility certainly caught my attention, I haven’t sold a single share.

Here is why I am comfortably holding on to my positions in AEM, UMS, and Micro-Mechanics.

Fundamentals Remain Intact: The Signal vs. Noise Disconnect

The primary reason for staying the course is straightforward: the operational realities of these three businesses have not deteriorated.

Markets trade on generic “semiconductor oversupply” headlines, but a closer look at each company reveals a clear insulation from commodity memory risks.

AEM Holdings: System-Level Testing (SLT) Solutions for Advanced Packaging

AEM’s primary growth engine is driven by its SLT solutions for High-Performance Computing (HPC) and AI logic processors.

Fulfilling long-dated, non-cancellable orders for Intel (INTC) and the aggressive volume ramp for its second major AI/HPC anchor customer will drive significant top- and bottom-line expansion over the coming years.

Furthermore, AEM’s strategic partnership with global packaging leader ASE Technology (TWSE: 3711) provides a direct gateway to integrate its test architectures deeper into OSAT advanced packaging environments worldwide.

What about the memory oversupply fears?

Memory is a fresh growth runway for AEM, but their focus is strictly on high-end, multi-chip setups. When High Bandwidth Memory (HBM) is stacked with AI processors on a single package, managing heat and signal flow becomes a huge challenge.

AEM’s advanced SLT platforms ensure these complex, integrated packages perform seamlessly under real-world server stress. CXMT flooding the market with low-margin, legacy DDR4/DDR5 capacity will not displace this need.

Most importantly, because memory test is an entirely new growth vertical starting from a low base, it can only expand AEM’s addressable market rather than expose its core.

UMS Integration: High-Precision Chambers for the HBM Pivot

UMS sits right at the front end of semiconductor manufacturing, supplying the high-precision chambers and sub-assemblies required by equipment leaders like Applied Materials (AMAT) and Lam Research (LRCX) to build their etching and deposition machines.

Any lingering fear of an industry oversupply is effectively crushed by Lam’s latest blowout Q4 2026 results — reporting record revenue of US$6.72 billion (up 30% YoY) and guiding for an astonishing US$8.1 billion in the upcoming quarter.

You can clearly infer from this that global demand for advanced wafer fab equipment remains surging. This is fuelled heavily by the HBM buildout, which isn’t being derailed by CXMT’s capacity ramp.

I am not discounting China’s tech capabilities. In fact, based on what we’ve seen, you should expect domestic Chinese firms to achieve technical breakthroughs faster than conventional consensus expects.

However, even when those breakthroughs happen, they won’t soak up global demand overnight.

Think of it like AI LLMs: as impressive as Chinese models have become, their rapid progress hasn’t slowed the explosive adoption of ChatGPT, Gemini, or Claude.

UMS remains strategically positioned to capture high-margin order flow over the next few years.

Micro-Mechanics: Diversified Customers Base & Benefitting from China’s Surge

Unlike AEM and UMS, Micro-Mechanics’ business model doesn’t rely on just a few anchor customers.

Instead, they manufacture high-precision, consumable tools, such as capillary nozzles and die-attach pick-up tools, that are essential for semiconductor packaging and assembly.

Because these tools wear out during normal production, revenue scales directly with global chip assembly volume.

Those who dump the stock out of fear over China’s aggressive domestic chip expansion clearly don’t understand the fundamental business model.

For 9M 2026, China contributed roughly 35% of group revenue and delivered the strongest YOY growth at 25%! China isn’t a threat to Micro-Mechanics — it is their largest and fastest-growing market.

Timely Correction: De-risking Stretched Valuations

While all three semiconductor enablers remain fundamentally sound, the recent plunge serves as a healthy reality check.

Earlier this year, market sentiment was sitting at peak euphoria. Share prices were appreciating “too fast, too furious,” pricing in an uninterrupted, linear super-cycle.

This pullback has effectively squeezed out speculative momentum, resetting valuations back to far more reasonable levels grounded in earnings realities.

While it’s easy to look back and wish I had sold at the peak to buy back today, the truth is no one has a crystal ball. I couldn’t predict short-term price swings — and I don’t try to.

Losing some of those profits is a trade-off I am willing to make, rather than risking future upside by selling out of a secular AI thesis that is still in its early stages.

That said, discipline comes from structure, not emotion.

The semiconductor rally in the previous quarter prompted me to set a guidepost: capping my exposure to 30% before evaluating any portfolio rebalancing.

Thus far, it has stayed comfortably within the limit — moving from a high of 28% in May down to the current 22%.

Crucially, this structure is precisely why I am not losing sleep over the recent drop.

Structural Risk Management: Protecting Overall Portfolio Return

By bounding sector exposure, I naturally limit the impact of any single-industry drawdown on my overall portfolio.

I’m glad to share that this defensive structure is working as intended.

Buoyed by the continued strong performance of Singapore banks and a solid recovery across my US core holdings, overall YTD portfolio returns have only dipped slightly from a peak of 18% down to 16.7%.

That is the beauty of appropriate diversification: you intentionally forgo a fraction of maximum upside during a speculative surge, in exchange for total peace of mind when the market inevitably recalibrates.

Mixed Feelings: Valuation Sanity over Momentum

Unless you are trading purely on short-term technicals, no long-term investor enjoys seeing the market value of their holdings contract.

Yet paradoxically, I feel far more comfortable holding AEM, UMS, and Micro-Mechanics at current price levels than I did at their mid-year highs.

When prices were soaring, every tick upward felt like driving at 120 km/h on an expressway — thrilling, but keeping you in a heightened state of anxiety because there was little margin for error.

At current levels, the market has eased off the pedal. The car is still cruising along at a solid pace, but within the speed limit.

With all three companies scheduled to report their earnings next month, I look forward to seeing how they perform – checking to confirm their operational engines are still humming along.