Harnessing photonics to lead the next wave of advancements in AI computation.
“Semiconductors are arguably humanity’s greatest achievement to date,” says Jimmy Goodrich, Vice President of Global Policy at the Semiconductor Industry Association. Over the last half century, what has fueled the digital semiconductors’ pervasiveness in society is described by Moore’s Law, built on the observation that the number of transistors on a microchip doubles every two years, and the cost per transistor is halved.
AI, at the cusp of its own revolutionary impact to humanity, is driving an unprecedented demand for computation right at the time that the physics of digital semiconductors is failing to continue to support year over year speed ups that follow Moore’s law. The challenges facing AI and all high-performance computing is not just the end of Moore’s law, but the end of Koomey’s law - the energy efficiency law of CMOS scaling. Dimensional scaling in electronics has ceased to contribute to on-chip power reductions over the course of the past ten years. Increasing the amount of compute and memory in chips has resulted in significant increases in power, the primary contributor being data movement.
Celestial AI’s novel approach has raised $56 million in Series A funding to go beyond the limits of traditional transistor scaling and disrupt the AI chipset industry. The company’s mission is to fundamentally transform the way computing is done through their proprietary Photonic Fabric™ technology platform that uses light for data movement, both within chip and between chips. Celestial is breaking the digital semiconductor mold by integrating photonics into accelerators for AI workloads, thereby enabling step-change advancements in AI computation. Harnessing light for data transmission, within chip and chip-to-chip, is faster, more power efficient, and lower in cost than conventional electronic-only semiconductor solutions.
The company’s Photonic Fabric™ enables optically addressable memory and compute that decouples their technology from the slowdown of Moore’s Law. Their proprietary architecture allows for highly efficient data mapping without the need for complex optimizations. This software advantage extends to multi-chip exascale systems as the Photonic Fabric democratizes optical access to effectively limitless memory and compute.
Hybrid photonic-electronic computing solutions redefine what is possible in the field of artificial intelligence. Photonics has been primed for use in computing applications through decades of advances of silicon photonics design and fabrication, driven by the data communications industry. “Celestial AI’s hybrid photonic-electronic platform allows us to leverage the complementary strengths of electronics for high-performance, high-precision computing and photonics for high-speed, low-power, high-bandwidth data movement. The result is transformational performance advantages relative to electronic-only systems,” says Celestial AI founder and CEO Dave Lazovsky. “Our competitive differentiation will increase with time, as AI model complexity increases, driving increased data movement.”
Lazovsky and his team have a shared belief that there are three pillars for a successful AI architecture. It must:
- Solve real customer application problems
- Be designed for reliability, manufacturability, and low-cost from the start
- Have a robust developer and deployer ecosystem that integrates seamlessly with customers’ existing flows
Celestial AI’s leadership is the driving force delivering on these three pillars and exceeding the bounds of the industry. Lazovsky formerly managed over $1 billion in semiconductor manufacturing equipment business at Applied Materials before founding his first startup, Intermolecular, and leading it through its IPO. Preet Virk, the company’s COO, brings deep technical knowledge from managing engineering teams in the semiconductor and data communications sectors. He previously served as SVP at MACOM and Mindspeed Technologies where he drove the initial adoption of electro-optical solutions enabling 100G transceiver deployment data centers. Celestial AI’s CTO, Phil Winterbottom, brings expertise that spans across photonics, AI / machine learning hardware and software, and ASIC system architectures. Following his 10+ years at Bell Labs, he served as VP of Engineering at an optical AI computing startup, and co-founded and led the technical teams at Entrisphere (acquired by Ericsson) and Gainspeed (acquired by Nokia).
The Celestial AI team plans to use their newly secured capital to expand the engineering team, product development initiatives, and strategic supplier engagements for the construction of their Orion AI accelerator products, serving a projected addressable market of over $70 billion by 2025. As Lazovsky and Virk note, humanity has only scratched the surface of AI’s potential. The digital domain has opened the door for AI but is not able to continue to deliver the same level of performance gains that has put the computational power of supercomputers of yesterday into the smartphones of today. Photonics promises to be the technology to usher in the next era of rapid growth, and Celestial AI has the know-how to take the industry there.