To deliver the world’s first
optical supercomputer with
of 100x energy efficiency
Scaling CMOS-based processors is increasingly challenging due to the slowdown of Moore's Law and the breakdown of Dennard scaling; data movement with electronic wires (capacitive loss) limit clock speeds and lead to thermal dissipation.
Opticore patented optical processing units (OPUs) are designed to handle memory-intensive AI tasks with on-chip data movement and computation with photonic integrated circuits. Our prototype achieves 100x higher energy efficiency and 100x greater computing density than digital electronics. Fully integrated with foundry-based, co-packaged optoelectronics, the system promises to lead breakthroughs in the next era of high performance computing in data centers.
Throughput
Power
Chip area
Memory bandwidth
AI Mode scalability
1,000 POPS (single chip)
5 KW
1560 mm2
480 TB/s (300 HBMs)
Trillion matrix elements
Opticore team proposed (PRX 9, 2019) and demonstrated (Nat. Photonics 17, 723–730) the first photonic logic based on photoelectric multiplication in coherent detection, which is the fundamental building block to enable large-scale, low energy computing with O(N) devices for O(N2) computes per clock cycle. As photonic technologies have become a necessity in datacenters for chip-to-chip communications, computing in the optical domain is an ultimate solution to overcome the electronic bottlenecks in energy efficiency and computing density with minimum change to the existing data center infrastructure.
The “memory wall” in AI computing stems from the capacitive resistance of metallic wires used for memory data movement, which limit clock speeds (1-3 GS/s) and lead to energy consumption. Opticore OPUs utilize the same high-bandwidth memories (HBMs) for their scalability and performance, but here the OPUs convert memory data to optical beams for on-chip movements with waveguides and computations with optical devices. Without capacitive resistance in optical waveguides, memory no longer needs to be physically close to processors, enabling potentially unlimited memory capacities.
Photonic devices are significantly larger than electronic transistors, which makes scaling existing optical computing approaches that map neural parameters onto photonic hardware (1 device/parameter) challenging. Opticore addresses this limitation by leveraging innovations in temporal mapping, where neural data is encoded using optical pulses. For example, an optical modulator can activate tens of billions of parameters per second, and through parallel channels, trillions of parameters can be processed simultaneously. Furthermore, all parameters can be dynamically programmed for training tasks, enabling highly flexible and scalable computing.
The Opticore computing chips are fabricated with standard foundry services, with co-packed optolelectronics and hyperbonding of HBMs.
Alex was CIO at Keshik Capital, a multi-strategy fund based in Singapore and now runs a family office. Previously, Alex worked at leading global investment banks and hedge funds. He has close to two decades of experience managing multi-asset portfolios and investing in public and private corporate securities and derivatives. Alex also conducts research on energy and security with the Australian National University and has published in top journals. Alex has a degree in economics, and he speaks English and Mandarin.
ECE Professor at MIT,
PhD in Applied Physics, Stanford,
B.S. Caltech,
Awards: Presidential Early Career Award, Humboldt Professorship, DARPA Young Faculty Award.
Co-founder of DUST Identity (Security company), QuEra Computing (quantum control), Advisor of Lightmatter (optical computing)
First demonstration of integrated photonics for deep learning, spinning off Lightmatter and Lightelligence
Postdoc, UC Berkeley
PhD, KAIST (Korea)
Bachelor: Seoul National University
Skills: photonics and microwave simulation, chip layout designs, and fabrication
MIT Postdoc fellow,
PhD, Electrical Engineering, Boston University
Awards: Charles L. Newton Prize, John F. Flagg Prize, for outstanding students. Wilder Trustee Scholarship, Xerox Engineering Research fellowship
Adam was the Partnership Development Manager and Tax Advisor at Carta, a leading capitalization table management company backed by prominent investors such as Andreessen Horowitz, Tribe Capital, and Union Square Ventures. Prior to Carta, Adam honed his expertise at EY, where he played a pivotal role in scaling multiple tax startups and practices to achieve seven-figure ARR after the first year of joining. Recognized for his innovative approach, Adam has been featured in Berkeley Haas News and a16z articles for his outreach and impact. He is a proud graduate of UC Berkeley’s Haas School of Business.
Assistant Professor, EECS, UC Berkeley
Postdoc, Harvard University
PhD, Cornell University & Columbia University
Awards: 2023 DARPA Young Faculty Award, 2023 Optica Foundation Award, 2020 The Optical Society (OSA) Ambassador, 2019 The Rising Stars, 2019 Caltech’s 2019 Young Investigator Lecturer, 2016 Maiman Student Paper Competition at CLEO, 2016 Emil Wolf Student Paper Competition
Assistant Professor, EECS, UC Berkeley
Postdoc researcher, MIT
PhD, Max-Planck Institute-Quantum Optics,
Awards: 2023 DARPA NaPSAC, 2024 DARPA INSPIRED, Sony Faculty Research Award, 2023 SPIE AI/ML best paper Award, 2023 Optica foundation award
Demonstration of the Opticore computing with 100x more energy efficiency, Nat. Photonics 2023
Research Scientist, MIT
Senior Scientist: NTT PHI lab
PhD, Stanford University
Bachelor: Caltech
Awards: Intelligence Community fellowship
Theory proposal and analysis for the OptiCore computing architecture, R. Hamerly, PRX 2019
Development
Fremont (CA)/Remote
Development
Fremont (CA)/Remote
Development
Fremont (CA)/Remote
- MIT news
Find Out More