ASML's 2025 Q4 Orders Exceed 13 Billion Euros, Indicating a Semiconductor Boom Driven by AI

Written by Niu Ske, Deep Tide TechFlow

13.2 billion euros.

This is the total order value ASML received in Q4 2025, more than doubling the market's expectation of 6.3 billion euros. This also marks ASML's highest single-quarter order record in its history.

On January 28, ASML, the giant in lithography machines, delivered a performance so impressive that it left Wall Street stunned, causing its stock price to surge 5% in pre-market trading.

This financial report serves as a signal that the AI craze is truly making its way up to the upstream semiconductor industry chain.

In Q4 2025, ASML achieved revenue of 9.7 billion euros, net profit of 2.8 billion euros, and a gross margin of 52.2%. These figures were in line with expectations and even somewhat modest.

But orders—that's the real number that drives the market crazy.

Of the 13.2 billion euro orders, 7.4 billion euros came from EUV lithography machines. These machines, each costing hundreds of millions of euros, are the only key for TSMC, Samsung, SK Hynix, and others to advance to 3nm, 2nm, or even more advanced process technologies.

Worse still, structural changes in the order volume have revealed a key signal: storage vendors have entered a frenzy.

The logic category orders amounted to 5.8 billion euros, an increase of 3 billion euros from the previous period, in line with seasonal fluctuations.

Storage class orders amounted to 7.4 billion euros, a surge of 4.9 billion euros from the previous quarter, far exceeding historical levels during the same period.

Memory giants such as SK Hynix, Samsung, and Micron are placing orders at an unprecedented, off-season pace, and the race for HBM (High Bandwidth Memory) production capacity has intensified dramatically.

CFO David Dai put it plainly: "Customers have become clearly more optimistic about the mid-term market outlook, mainly because they believe the demand related to AI will be more sustained."

This optimism has directly translated into real, hard orders. ASML's order backlog has already reached 38.8 billion euros, of which 25.5 billion euros are for EUV systems. To put it simply: the workload for the next two years is essentially fully booked.

Looking at the regional income structure, we can discover an interesting detail.

In Q4 2025, mainland China remains ASML's largest source of revenue, accounting for 36% and contributing 3.5 billion euros. This figure significantly exceeds the company's previous expectation of 25%, mainly driven by continued demand from Chinese manufacturers for ArFi (immersion DUV) equipment.

But what's truly worth noting is South Korea, where Q4 South Korean revenue share rebounded to 22%, amounting to approximately 210 million euros.

Behind this number is the aggressive expansion by SK Hynix and Samsung. SK Hynix has clearly stated its plan to purchase 12 EUV machines by 2026, fully ramping up production capacity for HBM3e, as NVIDIA's H100, H200, and B200 are eagerly awaiting supply.

Samsung is even more desperate. They have been thoroughly beaten by SK Hynix in the HBM market, with their yield rates consistently failing to improve. Now, the AI boom has given them a final window of opportunity—if they don't aggressively expand production, they will be completely out of the game.

Barclays analysts specifically named SK Hynix, predicting that it will acquire 12 EUV machines by 2026. Each EUV machine costs 260 million euros, making the total value of 12 machines amount to 3.1 billion euros.

More importantly, this wave of order peaks from storage vendors has just begun.

Micron has just announced that its 2026 capital expenditure will exceed $20 billion, representing a nearly 40% year-over-year increase. TSMC has raised its 2026 capital expenditure to between $52 billion and $56 billion, which is more than $10 billion higher than the original plan. Ultimately, all of this money will flow to ASML.

Several months ago, ASML was still saying that "revenue in 2026 could remain flat or even decline."

Now, the revenue range is directly given as 34 to 39 billion euros, with a median of 36.5 billion euros, representing a 12% increase compared to the 32.7 billion euros in 2025.

But in fact, this guideline is still conservative.

Why is this the case? Because after TSMC and Micron increased their CAPEX, the mainstream market institutions have already raised their growth forecast for ASML in 2026 to over 20%.

ASML's own guidance is for growth of "4-19%," with a median of only 12%. This kind of "below market expectations" guidance is rare among equipment stocks, as companies in this sector usually prefer to set a target that is "a bit challenging but achievable," then exceed expectations and reap a surge in stock price.

Why is ASML so conservative? There may be two reasons:

First, the uncertainty in the Chinese market. ASML expects the Chinese market to account for 20% of its business in 2026, compared to nearly 30% in 2024–2025. What does 20% mean? If ASML achieves total revenue of 36.5 billion euros in 2026, the Chinese market would contribute 7.3 billion euros—more than double the 3.5 billion euros in revenue from the fourth quarter of 2025. However, given export restrictions and the waning demand for stockpiling, it remains uncertain whether this target can be achieved.

Second, the mass production timeline for High-NA EUV systems. ASML delivered two High-NA systems (EXE:5200B) in Q4 2025 and recognized the revenue. These behemoths, each priced at 380 million euros, could easily boost revenue in 2026 if several more units are sold. However, the issue is that apart from Intel, other customers are still waiting and watching. TSMC believes that the current Low-NA EUV systems combined with multiple patterning are sufficient to support processes below 1nm, so there is no urgent need to adopt High-NA technology.

Therefore, ASML provided a "leaving room" guidance. Within the equipment sector, this is actually a positive signal, indicating that the company is confident in exceeding expectations, but simply doesn't want to overstate its projections.

Diving deep into product data can reveal some interesting insights.

EUV: Delivered 14 units in Q4, with an average selling price of 260 million euros per unit, generating revenue of 3.64 billion euros, representing a 22% year-over-year growth.

ArFi (immersion DUV): 37 units shipped in Q4, average selling price of 82 million euros per unit, revenue of 3.03 billion euros, representing a 4% year-over-year growth.

EUV and ArFi account for nearly 88% of revenue and are ASML's absolute cash cows.

However, there is a detail: ASML's EUV revenue growth mainly comes from an increase in average selling price, rather than growth in shipment volume. In Q4, 14 EUV systems were shipped, which is not a large number, but the average price increased from around 240 million euros to 260 million euros.

Why is the average price rising? Because the product structure is upgrading.

ASML's current main product for sales is the Low-NA EUV machine, model NXE:3800E. This machine has already achieved its target throughput of 220 wafers per hour, and in some customer scenarios, it can even reach 230 wafers per hour. It is a "highly mature" cash cow.

The unit price of High-NA EUV (EXE series) is 380 million euros, nearly double that of Low-NA. Two High-NA systems were recognized as revenue in Q4 2025, which directly increased the average selling price of EUV systems.

If the shipment volume of High-NA in 2026 increases from 2 units to 10 units, ASML's revenue growth rate will easily exceed 20%.

But this depends on when TSMC and Samsung actually place real orders. Currently, Intel is the most aggressive buyer, having already accepted the first 5200-type machine for high-volume manufacturing, preparing to use it in its 14A process.

TSMC is observing, and Samsung is hesitating. Whoever first adopts High-NA will gain a significant advantage in the race for sub-2nm process technologies. However, neither wants to be the first to take the plunge, as improving the yield of High-NA technology requires both time and substantial financial investment.

The root cause of this order surge is simple: AI is consuming everything.

But unlike a year ago, the transmission path of AI demand has now changed.

A year ago, people were concerned about "how many GPUs NVIDIA sold" and "whether TSMC's CoWoS production was sufficient." Now, the bottleneck has shifted to memory/storage.

The demand for computational power by large models like ChatGPT, Gemini, and Claude is insatiable. However, computational power alone isn't enough—you also need sufficiently fast memory to feed these computational beasts.

HBM3e is high-speed memory specifically designed for AI. Its bandwidth is more than six times that of regular DDR5, allowing GPUs to perform training and inference without being bottlenecked by memory I/O.

The problem is that the production capacity for HBM is too tight.

SK Hynix dominates the HBM market with an 80% market share. Samsung struggles with low yield rates and holds only about 15% of the market share. Micron has just entered the market and is not expected to achieve mass production until 2026.

NVIDIA, AMD, Google, Microsoft, and Amazon are all competing for HBM. The supply-demand gap is expected to last at least until 2027.

Therefore, storage vendors are not just expanding production, but "racing to expand production."

Micron's 2026 capital expenditure will reach $20 billion, representing a nearly 40% year-over-year increase. SK Hynix plans to acquire 12 EUV machines. Samsung has even increased its capital expenditure for its memory business by more than 50%.

This money will eventually turn into ASML orders.

From the perspective of application areas, currently, ASML's system revenue consists of 70% logic chips and 30% memory chips. However, in terms of order structure, the proportion of memory chips has already approached 60%.

This means that storage will become the largest growth driver for ASML's revenue in the next 12 to 18 months.

If there is any less optimistic news, it would be the Chinese market.

ASML expects the revenue share from the Chinese market to drop to 20% in 2026. In 2024-2025, this figure approached 30%, and even reached 36% in a single quarter (Q4).

Why did it decline? Because the Chinese orders for 2024-2025 are essentially "panic-buying" demand.

The U.S. restrictions on chips to China are tightening, and ASML's certain high-end DUV equipment (especially immersion ArFi) can no longer obtain export licenses. Chinese wafer manufacturers are well aware: while they still can buy, they are stockpiling as much as possible.

But such a stockpiling trend cannot continue in the long term.

In 2026, Chinese manufacturers will have sufficient equipment on hand, so new procurement demand will naturally decline. A 20% market share is actually quite significant—it is equivalent to a scale of 7 to 8 billion euros annually.

But for ASML, which has become accustomed to "excessive contributions" from the Chinese market, this is indeed a change that requires adjustment.

What's more troublesome is the gross profit margin.

The Chinese market mainly purchases immersion DUV tools (ArFi) with high gross margins, and the gross margin of these machines is even higher than that of Low-NA EUV systems. A decline in orders would put some pressure on gross margins.

This is also why ASML set its 2026 gross margin guidance at 51-53%, slightly lower than the 52.8% for the full year of 2025.

However, there is a detail: the Q4 gross margin was 52.2%, while the full-year 2026 guidance is 51-53%. This suggests that ASML expects the gross margin for the first half of 2026 to be below 52%, possibly due to a reduction in high-margin orders from the Chinese market.

In addition to the regular EUV, ASML is also developing a major new technology: High-NA EUV.

This is the next-generation lithography technology, with a single unit priced as high as 380 million euros. In Q4 2025, ASML has already delivered two High-NA systems and recognized the revenue.

Currently, Intel is the most aggressive buyer, having already accepted the first 5200-type machine for high-volume manufacturing, and is preparing to use it in the 14A process.

ASML is confident about the performance of High-NA: "The imaging, performance, and overlay results are all very good."

But the large-scale popularization of this technology might not happen until 2027-2028.

Why? Because both TSMC and Samsung are waiting and watching.

TSMC's logic is clear: our 2nm process is already in mass production, using Low-NA EUV combined with multiple patterning. The yield is good, and the cost is manageable. The 1.4nm process is also in development, still relying on Low-NA. Only when we go below 1nm will we need High-NA EUV.

In other words, TSMC believes that Low-NA EUV can still be used for another three years.

Samsung's situation is more delicate. They suffered significant setbacks in 3nm yield rates, and their GAA architecture progress has been too slow, falling two generations behind TSMC. Now, Samsung's most urgent task is to improve the 3nm yield and reclaim customers they've lost.

High-NA? Let Intel give it a try first.

However, if Intel's 14A process successfully achieves mass production using High-NA technology with good yield and cost efficiency, TSMC and Samsung will immediately follow suit. At that point, High-NA orders will surge like an avalanche.

This is why ASML dares to say, "Revenue of 44 to 60 billion euros by 2030."

Because High-NA is not just an equipment upgrade; it can also increase the price of a single unit from 200 million euros to 400 million euros. As long as shipment volumes do not decline, revenue can easily double.

In this impressive earnings report, ASML buried a less favorable message: it plans to lay off 1,700 employees, primarily in the Netherlands and partially in the United States.

The reason is: "The company's work approach has become less agile in certain situations."

This explanation is a bit subtle. ASML isn't short of money or orders, but rather feels that its organization has become too bloated.

Layoffs are primarily focused on the technology and IT departments, with the aim of "strengthening engineering and innovation capabilities in key areas."

Cut the peripheral departments and concentrate resources on core technology R&D.

This is actually a positive signal. ASML is preparing for larger-scale growth and requires a more flexible and efficient organizational structure.

But there's another interpretation: ASML might be concerned about the development progress and yield ramp-up of High-NA EUV technology. Laying off 1,700 employees could save money that can be redirected into more critical technological breakthroughs.

After all, Intel has already obtained the first High-NA machine. If their 14A process is successfully mass-produced, TSMC and Samsung will quickly follow. At that time, ASML must have sufficient production capacity and technical reserves to handle the surge in orders.

ASML announced a new stock buyback program of 12 billion euros, which is to be completed by the end of 2028.

The previous share buyback program (2022-2025) was originally also set at 12 billion euros, but in reality, only 7.6 billion euros was completed.

Why wasn't it completed? Because the stock price rose too quickly, making the repurchase cost too high. The company felt it wasn't worth it.

The launch of a new share repurchase program signals two things:

First, the company is confident in its future cash flow. 12 billion euros is a significant amount, and the fact that ASML is willing to spend this money on buybacks indicates that they believe their profitability will continue to improve from 2026 to 2028.

Second, the current stock price is attractive. If ASML believed the stock price was overvalued, they wouldn't rush to repurchase shares. Launching a buyback program indicates that the management considers the current valuation to be reasonable or even undervalued.

Dividends have also increased by 17%, reaching 7.5 euros per share. This is consistent with ASML's approach: when they make profits, they distribute them to shareholders.

ASML's financial report is never just the performance of a single company.

It is the barometer of the entire semiconductor industry chain, the thermometer indicating whether the AI wave has genuinely reached the manufacturing end, and the fuel gauge determining whether companies like TSMC, Samsung, and NVIDIA can continue to accelerate.

In 2025, ASML achieved revenue of 32.7 billion euros, with a net profit of 9.6 billion euros, representing a 16% year-over-year increase. The gross margin was 52.8%, and the net profit margin was 29.4%.

In 2026, the company is expected to achieve a 12% revenue growth (conservative estimate), while the market expects growth of more than 20% (aggressive estimate).

ASML's target for 2030 is revenue of 44-60 billion euros, with a gross margin of 56-60%.

This is not a conservative company. It is a company that firmly believes AI will reshape the semiconductor industry landscape and is ready to seize the biggest rewards.

The current market capitalization is $563.5 billion, with an expected PE ratio of about 39 times in 2026, which is at the central level of the historical valuation range of 30-45 times.

But this valuation may underestimate ASML's true value.

Why? Because two cycles are overlapping:

Short-term cycle (2026-2027): AI and storage demand surge, TSMC, SK Hynix, Samsung, and Micron aggressively expand production, ASML's orders are fully booked, leading to high revenue growth.

Mid-to-long term cycle (2027-2030): High-NA EUV systems begin mass production, with the unit price jumping from 200 million euros to 400 million euros, propelling revenue to an even higher level.

The overlap of two cycles could extend ASML's "super cycle" to 2030, or even longer.

The record quarterly order of 13.2 billion euros is just the beginning.

This feast has only just begun.