Summary

• ASML's critical role in semiconductor manufacturing, driven by AI advancements, makes it a key player with significant growth potential, despite geopolitical challenges.

• Open-source AI models like DeepSeek-R1 democratize AI, increasing demand for advanced chips and benefiting ASML's specialized EUV lithography tools.

• ASML's robust financials, including a strong installed base services segment, provide stability and growth opportunities amid cyclical semiconductor market fluctuations.

• High-NA EUV technology promises to push the boundaries of chip performance, ensuring ASML remains central to future AI and semiconductor advancements.

Artificial intelligence is at a fascinating crossroads. On one hand, open-source organizations like DeepSeek are driving down the cost of AI model development. On the other hand, the infrastructure to run these free or inexpensive AI models remains incredibly costly. At first glance, there may be a paradox. If AI is becoming “cheap,” why are incredibly expensive advanced chips still in extraordinary demand? The answer is that “cheap” AI doesn’t translate to low computational requirements. Instead, it means that AI software is widely available or open-source. Those advanced generative systems still need to run on hardware with billions of transistors, specialized memory, and parallel processing.

The process of manufacturing it is also so specialized that only a few facilities worldwide can manage it. These facilities, in turn, depend on ASML’s (NASDAQ:ASML) EUV lithography tools. Because of this relationship, ASML’s fortunes rise with each wave of AI breakthroughs. Every new model that sets a higher bar for performance eventually forces a wave of hardware upgrades. The progression from 7 nm to 2 nm processes indicates that ever more precise lithography will be necessary moving forward.

ASML's recent downturn presents a great entry point for investors.

Data by YCharts

ASML’s Critical Role in the Hardware Ecosystem

Semiconductor manufacturing is a sprawling universe of foundries, chip designers, equipment suppliers, and material vendors. With that being said, ASML is arguably the most specialized link in the chain. Its lithography systems shape the actual circuit patterns on silicon wafers. When the industry transitioned to deep ultraviolet lithography for advanced nodes, ASML arose as a leader. The shift to EUV ensures that ASML will be virtually the only player able to scale and commercialize the technology.

EUV machines work by generating photons at a 13.5 nm wavelength, which is much smaller than the 193 nm light used in DUV. This enables the accurate patterning of sub-10 nm features in fewer steps. The complexity and sophistication at which these machines operate is immense. For instance, these machines require specialized mirrors that reflect almost all incoming light to vacuum conditions needed for consistent high-energy photons. This exclusivity, combined with demand from top chipmakers, gives ASML a virtually insurmountable moat. The margin for error at such miniature scales is razor-thin, so customers look ASML as a sure-fire bet.

The EUV machine consists of over 100,000 unique parts.

High-NA EUV Lithography Machine (ASML)

Stunning Success of Open-Source AI and its Implications

DeepSeek's stunning success with DeepSeek-V3 and more recently DeekSeek-R1 has likely changed the course of AI development. DeepSeek's latest model R1 is basically equivalent to OpenAI's o1 across popular benchmarks while being 95% cheaper. Consider the typical path of an open-source AI project. Developers start with a baseline model for natural language processing, image recognition, or reinforcement learning. They then refine the architecture, incorporate new techniques, or train on bigger datasets. The code is published publicly, enabling others to replicate or further modify the approach. This free sharing accelerates the improvement cycle of AI, often with thousands of contributors worldwide.

As these models increase in complexity, training times and inference loads skyrocket. Data centers might quickly re-equip themselves with the latest GPUs or AI accelerators that promise greater performance gains. In the best-case scenarios, these upgrades also cut power consumption per operation. Either way, the push to jump to the next advanced node remains strong. This drives chip designers like Nvidia (NVDA), AMD (AMD), and Intel (INTC) to push TSMC (TSM) or Samsung (OTCPK:SSNLF) to adopt the latest EUV nodes. Each push leads back to new or upgraded lithography machines, funneling business to ASML.

Importantly, the open-source movement magnifies the user base for AI. Start-ups, academic labs, and smaller enterprises can now do proof-of-concept AI projects that were impossible a decade ago. Success in these modest pilot projects often triggers larger investments in AI compute. The wave is rising from all corners, not just from large hyperscalers with deep pockets. Again, this growing adoption would translate into increased demand for ASML's machines.

DeepSeek's stunning success with its latest model releases could further democratize AI and increase AI hardware demand. Despite the fact that DeepSeek-R1 is competitive with the industry-leading proprietary model o1, it is 95% cheaper.

Benchmark Comparisons (DeepSeek)

ASML’s Financial Footing and Outlook

With Q3 2024 net sales at €7.5 billion, ASML appears to be on track to close the year around €28 billion. Net income stands at €2.1 billion for Q3, reflecting a solid gross margin near 50.8%. These figures show the resilience of ASML’s business model, even amid broader semiconductor market cycles. ASML’s 2025 guidance of €30–€35 billion in total net sales might appear conservative given the growing need for AI processors. However, the firm is factoring in geopolitical complications and possible near-term cyclical headwinds. Export restrictions are likely to curb shipments of the most advanced machines to certain regions, notably China.

On a more optimistic note, AI’s relentless appetite for advanced logic and memory also provides a new stabilizing force. For instance, a downturn in smartphone demand might have created notable headwinds for equipment makers. Today, strong AI demand can offset weaker categories, creating a more balanced profile. Should that pattern continue, ASML could see less volatility than in past cycles.

Export Controls and the China Factor

One cannot discuss ASML without addressing the export-control elephant in the room. The motivation to restrict access to EUV machines in certain markets stems from national security concerns and trade tensions. For ASML, that environment means it cannot sell the most advanced EUV machines to China without explicit export licenses. China’s share of ASML’s business is expected to moderate to around 20% of total revenues.

While these developments are meaningful in the short term, they have not derailed ASML’s overall growth. Global demand for advanced chips is too broad. China remains a substantial market for older-generation tools and potentially less advanced DUV systems. However, ASML still maintains a tight grip on the market for the most advanced machines.

The Installed Base Protects Against Cyclicality

ASML’s business includes not just the sale of new systems, but also a sizable installed base services segment. Each machine in operation requires constant calibration, maintenance, and upgrades. Software updates can enhance overlay accuracy or improve throughput, each sold to the customer as part of an ongoing service agreement. This recurring revenue helps stabilize ASML’s financials. For AI-intensive applications as chipmakers need to squeeze out every last fraction of yield or performance from each wafer. This means that service contracts and upgrades grow even more critical.

Sub-2 nm Ambitions Through High-NA

While current EUV machines already push the limits of optical lithography, High-NA EUV raises the bar further. “NA,” or numerical aperture, determines the resolving power of the lithographic lens. Higher NA translates to the ability to print smaller features more cleanly. This technology could be key to reaching or surpassing the sub-2 nm threshold. For AI-focused chip designers, such a capability can have a monumental impact on the performance-per-watt ratio.

If High-NA lives up to expectations, it might ensure that Moore’s Law-like scaling continues for the foreseeable future. This allows for entirely new AI designs that incorporate specialized compute blocks, massive caches, and advanced memory-on-logic integration. While the timeline for High-NA’s full production deployment remains a few years out, there is already notable interest for these machines. Given the premium pricing of these next-gen machines, they could become an even larger driver of ASML’s revenues.

Conclusion

The relationship between democratized AI software and specialized hardware is clear. Each algorithmic breakthrough increases performance requirements, pushing device makers to adopt smaller transistor nodes, which in turn rely on ASML’s lithography machines. While geopolitical and cyclical factors may create momentary bumps, the secular trend remains bright. As AI further spreads throughout the global economy, the demand for advanced semiconductors should remain strong. This leaves ASML squarely at the center of the next era of computing. ASML is currently valued at $288 billion and has a relatively high forward P/E of 36. Even still, ASML has a great deal of growth potential given its status as the ultimate bottleneck in the AI supply chain.

Editor's Note: This article discusses one or more securities that do not trade on a major U.S. exchange. Please be aware of the risks associated with these stocks.