To The Moon

Axcelis Technologies is an awkward stock to analyse — which is precisely why I find it interesting.

The easy story says Axcelis is a semiconductor-equipment company whose silicon-carbide boom collided with slowing electric-vehicle investment and Chinese overcapacity. Revenue peaked at $1.13 billion in 2023, slipped to $1.02 billion in 2024 and fell again to $839 million in 2025. Trailing twelve-month revenue has recovered slightly to $866 million, but earnings have continued south.

That looks suspiciously like a cyclical peak.

Yet underneath it, the business mix is changing. Memory is returning, AI infrastructure is creating demand for decidedly unglamorous power-management chips, aftermarket revenue is expanding, and $Axcelis(ACLS)$ is attempting to combine with Veeco Instruments.

The market is therefore not simply deciding whether Axcelis recovers. It is trying to decide what Axcelis actually becomes.

The old cycle is fading. The next one looks different

The EV Hangover Is Real

I would not minimise the bear case.

Power devices represented 55% of Axcelis's system shipments in 2025. That extraordinary exposure worked beautifully when silicon-carbide manufacturers were racing to build capacity for EVs. When customers subsequently slowed capital expenditure, Axcelis discovered one of investing's oldest lessons: a fabulous growth market can still buy too many machines.

The financial damage is visible.

Operating income has fallen from $265.8 million in 2023 to $119.3 million in 2025 and $89.4 million on a trailing basis. TTM net income is $92.8 million, down more than 60% from the 2023 peak. Operating margin has compressed from 23.5% to 10.3%.

Revenue bent. Profits broke harder. That distinction matters

This is not merely Wall Street becoming grumpy. Profitability has genuinely deteriorated.

But Q2 2026 offers an intriguing wrinkle. According to Axcelis's Q2 presentation, power represented 39% of shipped system revenue while general mature applications contributed 44%. Silicon-carbide bookings improved sequentially, while Axcelis added two new Chinese SiC customers.

The patient is hardly sprinting, but neither is anyone measuring it for a coffin.

Memory Changes the Plot

The more interesting development is DRAM.

In its H1 filing, Axcelis reported that DRAM represented 22% of shipped systems revenue, while mature-process applications accounted for 77%. That is a meaningful shift for a company whose recent identity has been overwhelmingly power-heavy.

And it matters because AI's semiconductor appetite extends far beyond GPUs.

HBM requires substantially more complex manufacturing than conventional memory, while AI data centres consume power-management ICs, analogue devices, microcontrollers and optical-connectivity silicon. Many are manufactured on mature processes where implant intensity remains important.

Axcelis's Q2 presentation specifically identified growing AI data-centre demand for devices produced at 28nm and above. The company also completed a successful Purion XEmax evaluation at a leading foundry for next-generation power-management IC production.

That is an underappreciated angle: Axcelis does not need to manufacture bleeding-edge AI equipment to benefit from AI infrastructure. The electricity surrounding the GPU may prove almost as interesting as the GPU itself.

Silicon gets the headlines. Power delivery quietly sends the invoice.

Financial Deep Dive: The Balance Sheet Buys Time

The financials contain both reassurance and a warning.

TTM revenue stands at $866 million, gross profit at $372.5 million and free cash flow at $65.6 million. Gross margin remains 43.0%, respectable considering the revenue contraction, although below 2025's 44.9%.

Free cash flow is where enthusiasm needs restraint. It has fallen from $204.9 million in 2022 to $136.2 million in 2023, $128.6 million in 2024, $107 million in 2025 and $65.6 million TTM. FCF margin has consequently compressed to 7.6%.

Working capital explains part of the pressure. Inventory has climbed to $338 million from $195 million in 2021, while inventory turnover has declined to 1.52. That is one number I would watch particularly closely. Semiconductor-equipment inventories can represent preparation for recovery — or expensive optimism sitting on a warehouse floor.

$Axcelis(ACLS)$ nevertheless has considerable financial resilience. Cash and short-term investments total approximately $402 million, with another $175 million of long-term investments. Against roughly $70 million of reported debt, net cash is approximately $332 million.

It has also been shrinking its share count. TTM repurchases reached $57.8 million, following $121.1 million during 2025. Diluted shares have fallen from roughly 34 million in 2021 to around 31 million.

One less obvious positive is the aftermarket business. TTM aftermarket revenue reached $307 million, up from $235 million in 2024. That expanding installed-base revenue provides a useful counterweight to volatile new-system orders.

Machines already sitting inside fabs can keep generating economics long after the original champagne cork has disappeared.

Earnings weakened. The balance sheet refused to join them

The Moat Is Narrow — And Deep

Competitive analysis requires some nuance.

$Applied Materials(AMAT)$ remains the formidable incumbent in ion implantation, backed by vastly greater scale, R&D resources and customer relationships across semiconductor fabrication. Axcelis cannot win a spending contest with Applied, nor should investors pretend otherwise.

Its defence is specialisation.

Ion implantation is extraordinarily difficult engineering: dopants must be introduced into wafers at precisely controlled doses, energies and angles while maintaining purity, throughput and yield. Once a system is qualified for a production process, replacing it is neither casual nor cheap.

Axcelis has concentrated its R&D around Purion rather than trying to become another broad-line equipment conglomerate. Its common platform spans high-current, medium-current and high-energy applications, giving customers platform familiarity while allowing Axcelis to attack specialised niches.

That creates something better described as a technical oligopoly than a monopoly. Applied has scale; Axcelis has focus.

China: Customer and Risk Department

China remains the uncomfortable part of the story.

Asia-Pacific revenue is $652 million TTM, roughly three-quarters of total sales, although that figure includes major semiconductor markets outside China. Axcelis itself acknowledges China has represented a significant portion of recent sales and that US export restrictions can affect which customers and fabs it may serve.

China's mature-node capacity build may persist as domestic manufacturers pursue semiconductor self-sufficiency. Equally, local overcapacity, tighter US restrictions or eventual Chinese equipment substitution could turn today's demand into tomorrow's air pocket.

There is an additional irony. The proposed Veeco transaction still requires final regulatory approval from China's State Administration for Market Regulation. Axcelis therefore has China simultaneously occupying the customer list, risk register and merger timetable.

Diversification has rarely had such a sense of humour.

The Valuation Story Has Flipped

This is where I depart most sharply from the conventional value-trap argument.

Using the 28 September prior close of $123.11, rather than the subsequent intraday quote, Axcelis carries a market capitalisation of approximately $3.83 billion and enterprise value around $3.50 billion. On trailing numbers it trades at 41.6 times earnings, 58.4 times free cash flow and 35.2 times EBITDA. Forward P/E is approximately 25.7.

That is no longer distressed valuation territory.

Investors buying Axcelis today are already paying for meaningful earnings recovery. The margin of safety therefore has to come from future operating improvement, memory penetration, aftermarket growth and successful strategic execution — not from a bargain-bin multiple.

$Veeco Instruments(VECO)$ complicates that equation. Because the proposed merger is all-stock, $Axcelis(ACLS)$ is not emptying its cash reserves to fund the deal, and management expects the combined company to retain substantial liquidity. But it would assume Veeco's $230 million of 2029 convertible notes while attempting to extract $35 million of annual run-rate cost synergies.

In other words, shareholders would swap Axcelis's unusually pristine standalone balance sheet for greater diversification, greater scale — and greater complexity. The promised first-year non-GAAP EPS accretion sounds attractive; delivering it is what counts.

Watch What Changes Next

For me, Axcelis has moved beyond the simple value-trap debate. At 25.7 times forward earnings, the market is no longer offering investors the luxury of being right eventually. The operating recovery needs to arrive.

That makes the next evidence unusually important. The forthcoming Q3 results should show whether the improving SiC bookings and broader mature-node demand are translating into revenue, while memory's share of shipments will reveal whether diversification is becoming structural rather than merely quarterly noise.

Then there is Veeco. Chinese regulatory approval remains the final major hurdle, and the merger agreement allows the regulatory timetable to extend considerably if necessary. Completion would transform both Axcelis's competitive footprint and its financial profile.

So I would watch two clocks: the earnings cycle and the merger clock.

Two clocks are running. Neither answers to the other

If memory, aftermarket and mature-node demand strengthen while Veeco broadens the business without diluting Axcelis's financial discipline, today's valuation may ultimately look justified.

If not, investors will have learned an expensive semiconductor lesson: sometimes the cycle recovers before the valuation gives you permission to celebrate.

Wall Street will presumably discover this afterwards.

@TigerStars @Daily_Discussion @Tiger_comments @Tiger_SG @Tiger_Earnings @TigerClub @TigerWire