ASML’s Richard Kemkers: “We can’t let the AI power wall happen”
"Innovation and collaboration across chip design, manufacturing, and academia are the only answer to AI's energy hunger."
Published on October 30, 2025
Richard Kemkers, ASML © TU/e / Bart van Overbeeke
Bart, co-founder of Media52 and Professor of Journalism oversees IO+, events, and Laio. A journalist at heart, he keeps writing as many stories as possible.
At the launch of TU/e’s Casimir Institute, ASML’s Richard Kemkers warned that AI’s soaring compute demand could soon outpace global energy supply. His message: innovation and collaboration across chip design, manufacturing, and academia are the only way forward.
A decade ago, few imagined that semiconductors would define every aspect of modern life. Now, says Richard Kemkers, Head of Research at ASML, the same transformation is happening with artificial intelligence; only faster, deeper, and more energy-hungry.
“Only a few years ago, we were surprised that semiconductors were penetrating so deeply into our daily lives,” Kemkers said during his keynote at the launch of the Casimir Institute at Eindhoven University of Technology (TU/e). “Today, AI is doing exactly the same, and while it brings a fantastic economic opportunity, it also brings a huge responsibility.”
That responsibility, he argued, lies in making AI’s growth sustainable for the industry, for the economy, and for the planet.
The power wall problem
Semiconductor technology, Kemkers reminded the audience, has made remarkable strides in energy efficiency: chip performance per watt roughly doubles every two years. “Energy reduction is there by roughly 40% every two years,” he said. “That’s good news.”
But there’s a catch. “If you look at the demand side - the growth of AI and the compute needed to support training models - this is tremendous,” Kemkers explained. “If you just extrapolate Moore’s Law, it doesn’t match. In fact, within about ten years, we would run into a power wall, an electricity supply problem.”
That scenario, he emphasized, “will not and cannot happen.” The solution lies in innovation: smarter algorithms and more efficient chips.
Innovation on two fronts
Kemkers sees two main levers to close the growing gap between compute demand and energy supply.
First: smarter AI models. These need to become more efficient, consuming less energy while maintaining or improving performance. “That’s already happening, and it needs to continue,” he said.
Second: semiconductor innovation itself. Here, ASML and its ecosystem have a crucial role to play. “The monolithic chip design needs to be compartmentalized into sub-sections, so-called chiplets,” Kemkers said. “That means more efficiency, higher yield, and better integration through heterogeneous integration and 3D bonding.”
Alongside 3D stacking, traditional 2D scaling - the continued shrinking of transistors - remains essential. “Of course, ASML is playing its part in pursuing that path,” he added. But Kemkers stressed that progress can’t come only from the chips themselves. Every step in semiconductor manufacturing must become more cost-effective and energy-efficient.
Watt Matters in AI
Watt Matters in AI is a conference that aims to explore the potential of AI with significantly improved energy efficiency. In the run-up to the conference, IO+ publishes a series of articles that describe the current situation and potential solutions. Tickets to the conference can be found at wattmattersinai.eu.
Holistic lithography and EUV breakthroughs
As the global leader in lithography, ASML is targeting efficiency gains in its own domain. "Of course, ASML is focusing on lithography and metrology, so we play our part, but every vertical in this line will have an impact." Kemkers pointed to three key areas:
- Yield optimization – “We focus on yielding as many good transistors per wafer as possible,” he said. ASML’s “holistic lithography” approach leverages advanced metrology and feedback to correct deviations in real time.
- Simplified processing – Through EUV (extreme ultraviolet) lithography, ASML can achieve single-pattern exposures, reducing the need for multiple passes and cutting energy use.
- Shared responsibility – “Every element in the industry needs to optimize for cost and efficiency,” he emphasized.
These steps, combined with cross-industry collaboration, are vital to meeting AI’s accelerating compute demands without exhausting the planet’s power resources.
Brainport’s ecosystem advantage
Beyond technology, Kemkers underlined the importance of the Brainport Eindhoven ecosystem, with its dense network of suppliers, researchers, startups, and global companies all working in sync. “ASML is known as a system integrator,” he said. “We work with many partners, our customers, suppliers, academia, and governmental bodies. Especially here in Brainport, this ecosystem is vital to everything we do.”
One of the strongest examples is the long-term partnership between ASML and TU/e. “We’re very proud of this collaboration,” Kemkers said. “It’s a ten-year agreement that will eventually support around 100 PhD students per year.”
The collaboration is structured around seven innovation roadmaps, co-owned by ASML and TU/e, spanning topics from mechatronics and optics to sustainable supply chains. “It’s an excellent interchange of knowledge,” Kemkers noted, “going from the university to ASML and vice versa.”
Talent: the invisible infrastructure
No matter how advanced the technology, it ultimately depends on people. Kemkers dedicated part of his talk to developing talent across all levels, from PhD students to MBO graduates.
He highlighted the ASML PhD Masterclass, a biannual event where hundreds of students apply for a three-day workshop at ASML. “We typically select about 60 participants and hire 10 to 15 of them,” he said. “They do very well, they thrive compared to their peers. It’s a great way to connect academic talent to the needs of industry.”
Such initiatives, he added, are part of ASML’s broader diversity and education programs, ensuring a continuous influx of skilled professionals into the semiconductor ecosystem.
A shared mission
Kemkers closed on a note of urgency, but also optimism. “AI is growing tremendously,” he said. “But accommodating that demand requires many innovative steps.”
He sees the future of semiconductors as a layered evolution:
- 2D geometric scaling will continue.
- 3D integration and bonding will drive new gains.
- Holistic lithography will ensure yield and efficiency.
And, above all, collaboration will make it possible. “The ecosystem here in Brainport is fantastic,” Kemkers said. “It’s a great breeding ground for new enterprise. I would very much welcome more entrepreneurial activity in the semicon space, especially together with TU/e.”
In short, the message from ASML’s frontlines was clear: AI’s future depends on energy efficiency, and that future will be built right here, in the heart of Eindhoven.