Former Intel CEO Pat Gelsinger Targets Deep Tech and Lithography Breakthroughs at Playground Capital
The former Intel chief argues that free-electron lasers and expanded nuclear capacity are critical to sustaining semiconductor progress and powering artificial intelligence infrastructure.
Pat Gelsinger, who stepped down as chief executive of Intel in late 2024, has joined Playground Capital as a general partner, marking a definitive shift from corporate management to deep technology venture capital. Following a period of extensive consultation involving 100 meetings over 100 days, Gelsinger selected the role to focus on fledgling technologies built on new science, specifically aiming to help semiconductor startups reawaken Moore’s Law.
Central to his strategy is a board seat at xLight, a portfolio company developing novel lithography techniques using nanometer-scale beams of light. Gelsinger believes that advancing beyond the current 13.5-nanometer wavelength standard, which is utilised by Dutch manufacturer ASML, is essential to overcoming physical limits in transistor shrinking. He advocates for the use of free-electron lasers to potentially achieve wavelengths of 5, 4, 3, or 2 nanometres, positioning xLight as a partner to improve existing ASML machines rather than a direct competitor.
The venture capital landscape is shifting rapidly in response to the disruption of the software industry by artificial intelligence, with Gelsinger noting that the semiconductor industry is now projected to reach a trillion-dollar valuation next year, ahead of previous 2030 estimates. He argues that while many firms are moving toward deep tech, few possess the technical rigour required to identify winners in fields where physics has not yet been fully proven, emphasising the need for investment teams comprising engineers and PhDs to conduct rigorous due diligence.
Beyond hardware innovation, Gelsinger identified energy capacity as a critical bottleneck for AI infrastructure, describing the low-single-digit expansion in US energy capacity over the last decade as insufficient. He called for reigniting nuclear build-out and leveraging existing nuclear footprints to support the computational demands of the sector, suggesting that efficiency improvements in chips and power distribution could eventually allow one gigawatt of power to produce the equivalent of 10 gigawatts worth of processing tokens.
Gelsinger also addressed the regulatory environment, advocating for either industry-led review or government intervention to ensure foundational models meet security and values alignment requirements. He stressed the importance of rigorous benchmarking and visibility into training data, arguing that Western nations must maintain leadership in foundational models to ensure they enhance human experience and solve complex global problems.
