Discovered Materials Raises Nine Million Dollar Seed Round To Solve AI Chip Overheating Problem
Two IIT Madras alumni have raised nine million dollars to build AI agents that discover new semiconductor materials in days instead of years, betting they can solve the overheating problem threatening to slow down the entire AI chip industry.
Highlights:
- Discovered Materials has raised 9 million dollars in seed funding led by Lightspeed India Partners
- The startup was founded by IIT Madras alumni Akash Ramdas and Advaith Sridhar, who met over a decade ago
- The company uses AI agents to compress semiconductor materials research from months into days
- High performance AI chips can generate heat fluxes of roughly 140 watts per square centimetre
- Discovered Materials also released a new benchmark for testing AI agents on real world materials discovery problems
Every AI breakthrough eventually runs into a wall built from something considerably less glamorous than algorithms, physical heat. Chips pushed hard enough to train the world’s largest AI models generate warmth fast enough to damage themselves, and the materials science needed to solve that problem has historically moved at a pace measured in years. Two IIT Madras alumni are betting AI itself can compress that timeline down to days.
Discovered Materials, a San Francisco based startup, has raised 9 million dollars in seed funding led by Lightspeed India Partners, with participation from Y Combinator and Peak XV Partners, alongside angel investors including OpenAI cofunder Paul Graham, Gokul Rajaram, and Thariq Shihipar.
The company was founded by Akash Ramdas and Advaith Sridhar, who first met more than a decade ago as students at IIT Madras. Between them, the founding team brings a genuinely rare combination, deep formal materials science expertise paired with frontier AI engineering experience, one cofounder holding a doctorate in materials science from Stanford University with more than a decade spent researching new semiconductor materials, the other having built AI agents and video models at companies including Persona AI and Luma Labs.
That combination shapes the company’s core approach directly. Discovered Materials builds AI agents designed to compress what has traditionally been a slow, expensive, deeply interdisciplinary research process, simulation, synthesis, and experimental validation, into a considerably faster cycle, aiming to turn months or years of materials research into a matter of days.
“AI is creating unprecedented demand for better chips, but progress is increasingly constrained by how slowly new materials reach production,” said Hemant Mohapatra, partner at Lightspeed India Partners. “Akash and Advaith bring together a rare combination of deep materials science expertise and frontier AI engineering, enabling them to compress years of materials R&D into days.”
The company has chosen a genuinely specific, urgent starting point for that broader ambition, heat. High performance AI chips, the kind powering large language model training and inference at scale, can currently experience heat fluxes of roughly 140 watts per square centimetre, a level of thermal stress that fundamentally limits how densely chip components can be packed together. Discovered Materials believes new materials could address that constraint directly, enabling denser three dimensional chip stacking while simultaneously improving how efficiently heat is dissipated once it is generated, a combination that, if achieved, could meaningfully change how much computing power can be packed into the same physical footprint.
“In the last three months, we have made new thermal materials that match the performance” of existing benchmarks, Mohapatra noted, pointing to early tangible progress from the company’s approach rather than purely theoretical promise.
Alongside the funding announcement, Discovered Materials also released something that speaks directly to how the company wants to be evaluated going forward, hundreds of new materials the company says its AI agents have already discovered, alongside Material Discovery Bench, which it describes as the first benchmark specifically designed to evaluate AI agents on real world semiconductor materials discovery problems, rather than the more abstract, synthetic benchmarks that have typically been used to measure general purpose AI capability.
That benchmark release matters for reasons beyond simple marketing. Materials science has historically been a field where genuine progress is notoriously difficult to verify from the outside, claims of breakthrough materials frequently fail to reproduce reliably once they leave a controlled laboratory setting, and a purpose built, transparent evaluation standard gives outside researchers and potential customers a concrete way to assess whether Discovered Materials’ AI agents are actually solving real problems, rather than simply generating plausible looking candidate materials that never survive contact with actual manufacturing conditions.
The founders plan to use the fresh capital in fairly direct, unglamorous ways, expanding the company’s engineering and research team while scaling up its own laboratory operations, a combination that reflects a genuine hardware and wet lab commitment alongside the AI software development, rather than a purely digital, simulation only approach to the problem.
It is worth applying real scrutiny to how far Discovered Materials has actually travelled from promising early results toward genuine industry impact, rather than treating a nine million dollar seed round as evidence the core problem has been solved. As one industry analysis of the raise noted plainly, the core risk in this category lies specifically in the transition from laboratory simulation to physical manufacturing, AI discovered materials must ultimately be proven to work reliably under real world manufacturing conditions, a process that typically involves complex chemical synthesis, extensive testing, and qualification cycles that can take considerably longer than the AI driven discovery phase itself, regardless of how quickly a candidate material was first identified.
There is also a genuinely competitive landscape sitting behind this specific opportunity. Discovered Materials is entering a field where established semiconductor manufacturers and major chemical companies bring decades of accumulated manufacturing expertise, existing production infrastructure, and deep customer relationships that a young startup, however sophisticated its AI tooling, cannot replicate quickly. TechCrunch’s own coverage of the raise captured this tension in its framing, describing the company’s task as something closer to a continuous game of whack a mole than a single, solvable engineering problem, since chip designs keep evolving, and the thermal challenges facing tomorrow’s chips will likely look meaningfully different from the ones Discovered Materials’ current materials were built to solve.
None of this diminishes the genuine importance of the underlying problem the company has chosen to tackle. Heat dissipation is a real, well documented bottleneck constraining how far chip density and AI compute performance can actually scale, and a founding team combining formal materials science depth with genuine AI engineering experience is a meaningfully more credible bet on solving it than either discipline could offer alone. Whether Discovered Materials’ AI agents can consistently produce materials that survive the considerably harder, slower journey from laboratory validation to qualified, mass manufacturable production, or whether the company finds itself perpetually chasing a moving thermal target that outpaces how fast even AI accelerated materials discovery can actually move, is a question this seed round buys the company time to work toward, but has not yet answered.
