The U.S. funds xLight to tackle ASML’s bottleneck

The United States has decided to invest public funds in one of the most challenging parts of advanced chip manufacturing: the ultraviolet light source for EUV lithography. The Department of Commerce and NIST have allocated $150 million in support of xLight, a California startup led by Pat Gelsinger, former CEO of Intel, which aims to develop a free-electron laser (FEL) EUV source.

This move does not mean that the U.S. will have a full alternative to ASML tomorrow. The Dutch company remains the sole producer capable of manufacturing volume commercial EUV scanners, a position that gives it a unique role in the global semiconductor supply chain. What’s important is that Washington is focusing on a critical system component: the light that enables the printing of increasingly smaller circuits on wafers.

xLight’s approach is ambitious because it does not attempt to directly replicate current ASML technology. Commercial EUV scanners use LPP (Laser-produced Plasma) sources, based on plasma generated by laser impact on tin droplets to produce extreme ultraviolet light at 13.5 nanometers. xLight proposes using a free-electron laser, supported by particle accelerator technology. If successful at an industrial scale, this could offer higher power, greater efficiency, and a different economic model for chip factories.

The light source, the less visible bottleneck

EUV lithography is often described as an ultra-precise projector that prints microscopic patterns onto silicon. But this simplifies the problem greatly. Manufacturing advanced chips requires more than perfect optics, masks, nanometric positioning, and control software. It also necessitates a sufficiently powerful, stable, and repeatable light source to expose wafers at high throughput.

This is where xLight’s importance emerges. The company claims that current EUV sources do not deliver all the power needed for advanced lithography and that the associated costs of EUV comprise a significant portion of wafer fabrication costs. Their promise is to generate much more EUV power, reduce costs, and increase productivity for existing and future scanners.

The difference between a lab technology and a production tool is enormous. In an advanced chip manufacturing facility, the light source must operate continuously with stability, controlled maintenance, optical integration, thermal management, industrial availability, and performance metrics compatible with mass production. This is the leap that xLight still needs to prove.

Washington aims to reduce its supply chain weak point

The decision to fund xLight carries both industrial and geopolitical implications. The U.S. leads in chip design, EDA software, GPUs, cloud computing, AI, and some key semiconductor manufacturing equipment, but it does not control the scarcest tool in advanced fabrication: the commercial EUV scanner. That position belongs to ASML, with a highly specialized network of European, American, and Japanese suppliers.

For years, Washington has strategically leveraged this dependence alongside the Netherlands and Japan to limit China’s access to advanced tools. But this dependence also serves as an uncomfortable reminder: if the most critical capability is concentrated in a single foreign company, any political tension, industrial bottleneck, or supply restriction becomes a risk to U.S. strategy.

Supporting xLight doesn’t immediately break that monopoly, but it opens a second pathway. It’s not just about building “another ASML” from scratch—something that would take many years. It’s about investing in a component capable of changing the performance and economy of EUV. If an FEL source could be integrated into existing or future tools, the U.S. would gain more influence over an essential part of the system.

Pat Gelsinger and the appeal of a difficult bet

Pat Gelsinger’s involvement adds symbolic weight to the project. After leaving Intel, he has joined xLight as CEO. His name lends credibility to investors, chipmakers, suppliers, and public authorities. It also connects the project to one of the major frustrations of American industry: having lost relative manufacturing capacity despite maintaining leadership in design and architecture.

xLight had already raised $40 million in 2025, and according to information from TNW based on The Information, is now negotiating a round of $350 million with Boardman Bay Capital Management and Bain Capital as possible lead investors. The figure is significant because photonics, accelerators, and EUV sources are not software; they require heavy capital, labs, specialized equipment, precision engineering, talent, and years of validation.

Still, caution is advisable. A negotiation round isn’t a closed deal. An expected prototype is not a production line. And a FEL capable of generating EUV light does not automatically mean a ready-for-market industrial solution for clients like TSMC, Intel, Samsung, or Micron.

ASML remains active

The story isn’t that ASML is waiting for xLight to catch up. The Dutch company has already demonstrated advances in EUV source power. Researchers have shown a 1,000-watt source under relevant conditions, compared to the current 600 watts, aiming to boost productivity and reduce cost per chip by the end of the decade.

ASML estimates that this power increase could enable processing around 330 wafers per hour by 2030, compared to approximately 220 today. The company also sees a reasonable route to 1,500 watts and does not rule out reaching 2,000 watts in the future. In other words, the bottleneck xLight wants to address is exactly the same one ASML is working to improve internally.

This point is key. xLight doesn’t compete against a static picture of ASML but against a company that has spent decades solving nearly impossible engineering problems and maintains deep relationships with customers, suppliers, and research centers. The barrier isn’t just technological; it’s industrial.

More than ASML: the return of deep hardware

The xLight case is part of a broader trend. Artificial intelligence has revived capital and attention toward hardware layers that for years seemed reserved for specialists: lithography, inspection, metrology, advanced packaging, memory, interconnects, and inference accelerators. AI models depend on GPUs, but GPUs depend on factories, EUV, HBM, optics, chemicals, energy, and control tools.

TNW places xLight alongside other deep-tech ventures like Euclyd in the Netherlands, a startup seeking funding for AI inference chips and connected with figures linked to ASML. The parallel is interesting: the U.S. tries to strengthen a lithography piece where Europe leads; Europe seeks openings in AI accelerators where NVIDIA dominates.

The underlying message is that the AI race is no longer decided solely by models. It’s decided by industrial capacity. Those who control chip manufacturing tools, the memory to feed them, the energy to run them, and the infrastructure to deploy them will have an advantage that cannot be bought simply with more data or better algorithms.

Caution is mandatory

xLight may represent a significant bet, but it remains a gamble. Free-electron lasers have proven their value in scientific research and specialized applications but have not yet operated as EUV sources in advanced semiconductor fabs. Scaling this technology for manufacturing requires reliability, cost competitiveness, seamless integration with existing tools, support, maintenance, and acceptance from highly conservative customers.

Furthermore, a light source isn’t a complete scanner. ASML controls an architecture that includes Zeiss optics, precision stages, metrology, thermal management, software, process integration, and a unique supply chain. Replacing a component can significantly improve the system but does not immediately eliminate the advantage accumulated by the Dutch company.

Public funding also carries risks. Political shifts can change priorities, timelines may extend, and goals risk being caught between industrial promises and technical realities. Semiconductor cycles are measured in decades, not headlines.

A warning for Europe

For Europe, the U.S. move should be seen as a warning. ASML is one of Europe’s few structural technological assets with global influence. Its position in EUV gives the Netherlands and the European Union a rare significance in an industry dominated by the U.S. and Asia. If Washington funds alternatives—even nascent ones—it’s because they see strategic value in that dependence.

Europe should not respond complacently. ASML maintains an advantage, but it must be protected through investment, talent, suppliers, energy, industrial policy, and a smart partnership with the U.S., Taiwan, South Korea, and Japan. EUV lithography is not just a company; it’s an ecosystem of physics, optics, software, materials, manufacturing, and customers that takes decades to build.

xLight does not threaten ASML’s dominance tomorrow but points to where part of the next battle will be fought: in the light. More power, lower cost, more wafers per hour, and new wavelengths could reshape the economics of advanced chips. If AI demands more manufacturing capacity, industry will pursue any technology that increases tool performance.

The question is no longer just who designs the best chips but who controls the machines, the light sources, and the processes that make them. In that context, ASML still has the strongest answer. The U.S. has just made it clear that it wants to start building another.

Frequently Asked Questions

What is xLight?
An American startup founded in 2021 focused on developing free-electron lasers for critical applications, including EUV light sources for advanced semiconductor manufacturing.

How much public funding will xLight receive?
The U.S. Department of Commerce and NIST have finalized a $150 million incentive under the CHIPS and Science Act to support the construction and demonstration of an FEL prototype.

Will xLight replace ASML?
Not in the short term. ASML remains the sole producer of commercial EUV scanners at scale. xLight aims to develop an alternative light source—a critical component—but is not yet a fully integrated, production-ready system.

Why does the EUV light source matter?
Because its power and stability determine the productivity of EUV machines. More power can reduce exposure times, increase wafers per hour, and lower the cost per chip.

via: xlight