Jeff Bezos has a well-documented appetite for bets that look like science projects until they don’t. Blue Origin. Bezos Earth Fund. Project Prometheus. Now, through his family office Bezos Expeditions, the Amazon founder is putting weight behind CuspAI, a Cambridge-based startup.
According to a Financial Times report, CuspAI is in talks to raise at least $400 million, with investors including Silicon Valley venture firm Kleiner Perkins and Bezos Expeditions. Term sheets have reportedly been signed, though the deal has not yet formally closed. The raise would push the company’s valuation well beyond the $1 billion unicorn threshold.
What CuspAI actually does
Founded in 2024, CuspAI describes itself as a frontier AI company specialising in generative AI and molecular simulation to accelerate the discovery of breakthrough materials for industrial applications across semiconductors, energy, and climate. Think of it less as an AI company and more as a materials discovery engine: users specify the properties they need a material to have, and the platform generates the chemical composition most likely to deliver them.
Finding new materials such as advanced semiconductors, sustainable battery chemistries, or PFAS-filtering compounds has usually taken a decade or more with traditional trial-and-error methods. CuspAI’s platform claims to compress that timeline from years to months, with what the company says are success rates of up to 90%.
The company was founded by Dr Chad Edwards, a chemist and DeepTech entrepreneur who previously scaled quantum computing unicorn Quantinuum, and Professor Max Welling, an AI pioneer and former Distinguished Scientist at Microsoft Research and VP of Technology at Qualcomm. Its advisory board includes Martin van den Brink, former president and CTO of ASML, and AI luminaries Geoffrey Hinton and Yann LeCun, a lineup that signals this is not a speculative side project.
Announcing its Series A in September 2025, CuspAI CEO and Co-founder Chad Edwards said: “Next-generation AI compute, clean air and water and sustainable energy – these multi-billion-dollar global challenges share a common barrier: materials. CuspAI’s mission is to change that.”
Why this matters to the infrastructure industry
The connection to digital infrastructure is more direct than it might appear. The AI buildout’s most acute constraints are no longer primarily about land or planning permission, though those remain real, but about the physical components underpinning it all: chips, thermal materials, cooling systems, and power conversion hardware.
CuspAI’s platform has formed commercial partnerships across automotive, semiconductors, water purification, and climate tech, with partners including Hyundai, Meta Platforms and chemicals company Kemira, which is working on removing harmful PFAS compounds from water. The water purification angle is particularly pointed: as data centre operators face increasing scrutiny over water withdrawal for evaporative cooling, the ability to engineer better filtration or water-treatment materials at speed carries real operational value.
The water angle lands at a particular moment for the data centre industry. As Capacity reported this week, water rights have quietly moved from due diligence footnote to deal-breaking variable, with more than 40% of planned and existing data centres in the United States sitting in areas classified as high or extremely high water scarcity. Arizona has effectively closed new groundwater certificates to hyperscale projects. Virginia has introduced no-net-increase water clauses for new permits. Projects totalling an estimated $156 billion were delayed or cancelled in 2025 alone as a consequence of local resistance, with water concerns accounting for the single largest category of community complaints. If CuspAI can genuinely compress the timeline for discovering better filtration and water-treatment materials, as its partnership with chemicals company Kemira on removing PFAS compounds from water suggests is already underway, the commercial relevance to data centre operators is not abstract. It is an answer to a problem that is already killing deals.
CuspAI was one of the frontier AI companies named in the UK government’s AI for Science strategy and was among the first to gain significant access to the Isambard AI supercomputer. Jensen Huang has also cited the company as one of the UK startups to watch, a rare endorsement from the chief executive of the company whose chips are powering the very infrastructure CuspAI’s materials work is designed to improve.
The Bezos context: physical AI is the new frontier
Bezos Expeditions joining the CuspAI round is consistent with a clear pattern in how the Amazon founder is deploying capital in 2026. Bezos recently emerged from stealth with Prometheus, a physical-world AI lab he has led since handing over Amazon’s reins, raising $12 billion at a $41 billion valuation. Prometheus aims to do for engineering and manufacturing what large language models have done for text. The CuspAI investment sits naturally alongside that thesis: if Prometheus is building AI that designs physical systems, CuspAI is building AI that discovers the materials those systems will be made from.
CuspAI had been valued at just $520 million in September 2025, so a raise at unicorn-plus valuation represents a significant upward rerating in a short space of time. That jump reflects both new commercial contracts and, arguably, a market that is beginning to price in the strategic importance of the materials layer – the unsexy but foundational piece of the AI infrastructure stack that nobody talks about until something breaks.
RELATED STORIES
AI godfather blasts Zuckerberg: Yann LeCun breaks his silence
AI titan Yann LeCun and ex-deputy PM Nick Clegg, team up to drive Europe’s AI ambitions
AI pioneer Yann LeCun quits Meta following reported rift with Zuckerberg
Yann LeCun slams AGI hype, says human-level AI is years away






