Sign in Subscribe
Western Hemisphere

Physical Superintelligence Launches With US$58 Million to Put Virtual Physicists to Work on Data Center Efficiency

The Editor

The Editor

— 4 min read
Share
Physical Superintelligence CEO Matt Pines
Physical Superintelligence CEO Matt Pines

Physical Superintelligence (PSI), a Cambridge, Massachusetts company describing itself as an AI-native physics laboratory staffed by virtual physicists, emerged from stealth on September 1, 2026, with a US$58 million seed round led by Breakthrough Energy Ventures. The company was co-founded by Chief Executive Officer Matt Pines, Alex Klokus and Dr. Alexander Wissner-Gross, and is structured as a public benefit corporation. The round also drew Dragon Global, Robot Ventures, Solari, Susa, SV Angel, Valkyrie and JUMP Capital, along with the a16z Scout Fund and individual investors including Balaji Srinivasan and Anthony Scaramucci, plus angels affiliated with OpenAI, NVIDIA, SoftBank Energy, Oracle and Hugging Face.

PSI's core product is Emmy, a platform named after the mathematician Amalie Emmy Noether that the company describes as a team of virtual physicists which breaks research problems into trees of verifiable hypotheses and tests them in parallel. Rather than sell the science directly, PSI is commercializing Emmy first on a narrower problem: optimizing the multiphysics interactions between power, cooling, network and compute systems in data centers, both for new construction and for retrofits of existing facilities, on the ground and in orbit. The company is also the founding technical partner for the Fermi Explorer Mission, billed as the first AI-planned interstellar mission to Alpha Centauri, where PSI says it validated the mission physics and identified a substantially more efficient trajectory within the program's mass and budget constraints.

Market Context

The round is unusually large for a company with no public revenue. Dealroom put it above the 99th percentile among 12,095 comparable US enterprise software seed deals, which places PSI in the same tier as the AI-for-science cohort that has absorbed enormous early checks over the past two years, including Lila Sciences, which raised a US$350 million Series A and has since been valued above US$1.3 billion, and Periodic Labs, which raised a US$300 million seed. The common thesis is that AI systems can compress discovery cycles in physical sciences the way they have in software, and that a commercial wedge can fund the longer research program.

PSI's chosen wedge is where the money currently is. Gartner expects data center electricity consumption to reach 565 terawatt-hours in 2026, up 26% from 447 TWh in 2025, with power demand rising to 132 gigawatts from 104 GW and heading toward 290 GW by 2030. AI-optimized servers alone are projected to account for 31% of data center power draw this year and to pass conventional servers by 2027. Any credible percentage-point gain in efficiency at that scale translates into large absolute savings, which is also why Breakthrough Energy Ventures, the climate-focused firm founded by Bill Gates, is the lead rather than a generalist AI fund.

What Stands Out

"PSI aims to industrialize the discovery of new physics. Today, that means giving our customers a measurable edge to design and run their data centers more efficiently." — Matt Pines, Co-founder and CEO, Physical Superintelligence

Regional Relevance

For the United States, PSI sits at the intersection of the country's two most capital-intensive current bets: frontier AI and the electricity infrastructure required to run it. American hyperscalers are committing tens of billions of dollars a year to data center construction while facing interconnection queues, transformer shortages and local opposition to new load. Software that squeezes more compute out of the same megawatt is therefore strategically valuable well beyond its own market size, because it partially substitutes for generation and transmission that cannot be built on an AI timeline. The founding team's profile also reflects a specifically American research pipeline: Wissner-Gross holds MIT undergraduate degrees in physics, electrical engineering and mathematics and a Harvard doctorate in physics, and is known for a 2013 Physical Review Letters paper arguing that intelligent-looking behavior can emerge from systems that maximize future freedom of action.

Cambridge, Massachusetts is the second layer of relevance. The Boston area has spent the past two years turning its university physics and life-sciences base into an "AI for science" cluster, with Lila Sciences also headquartered in the region. PSI adds a hard-physics node to that cluster and, unusually, one with an energy-infrastructure customer base rather than a pharmaceutical one, which pulls Massachusetts talent toward a sector traditionally centered in Texas, Virginia and the Pacific Northwest.

The orbital component matters for a third constituency. By pitching Emmy at both terrestrial and orbital data centers and attaching itself to an interstellar mission concept, PSI is positioning for public-sector and space-adjacent budgets alongside commercial ones, a route several US deep-tech companies have used to bridge the gap between research ambition and revenue.

The Other Side

Is a data center optimization product enough to justify a US$58 million seed? Physical modeling of power and cooling is a real problem, but it is also a crowded one, addressed today by computational fluid dynamics vendors, digital-twin platforms from Schneider Electric and Siemens, and in-house teams at every hyperscaler. PSI has to show that a hypothesis-generating AI layer beats a well-tuned simulation stack by a margin operators will pay for, and it has not yet published benchmarks.

Does the "discover new laws of the universe" framing help or hurt? The launch language is expansive, and investors in this cohort have shown appetite for it. The risk is that the commercial wedge and the research mission compete for the same scarce physicists and the same runway. Companies that promise both a near-term product and a foundational science program have historically had to choose within two to three years.

What is actually verifiable right now? The Fermi Explorer trajectory work is the company's main proof point, and it is a mission concept rather than a launched spacecraft. Until PSI names data center customers and quantifies efficiency gains, the round is best read as a bet on the founders and the category, not on demonstrated performance.

Sources & Transparency

Read more