HPC

Photonics startup Luminous Computing bags $105m

Bill Gates among those plowing funds into AI supercomputing tech


Luminous Computing, a startup using photonics to drive artificial intelligence, has raised venture capital backing, pulling in $105m in Series A funding from a range of investors that includes Microsoft co-founder Bill Gates.

The Bay Area upstart, founded in 2018, announced the funding round, which included such firms as Gigafund, 8090 Partners, Third Kind Venture Capital, Alumni Ventures Group and Strawberry Creek Ventures. It adds to the $1m in pre-seed money the company received in 2018 and the $9m in seed funding pulled in a year later.

Luminous officials said the new cash will be used to double the size of the company's engineering team and the build-out of its custom chips and software, as it ramps toward commercial-scale production. It also is continuing to recruit photonics designers, digital and analog very large-scale integration (VSLI) engineers, packaging and system integration engineers and machine learning experts.

The company is among a growing number of vendors that are turning to photonics to solve the gnarly problems of compute, latency and memory when it comes to AI and machine learning technology, which are quickly becoming cornerstones of modern applications and industries from autonomous vehicles to pharmaceuticals.

The goal is to use light rather than electricity running over wires to send signals, which proponents argue will enable the development of more sophisticated and less expensive AI systems.

"AI has become superhuman," Marcus Gomez, CEO and co-founder of Luminous, said in a statement. "We can interact with computers in natural language and ask them to write a piece of code or even an essay, and the output will be better than most humans could provide. What's frustrating is that we have the software to address monumental, revolutionary problems that humans can't even begin to solve. We just don't have the hardware that can run those algorithms."

According to Luminous, the issue is that current computers use electrical signals, which drags on performance as signals travel. Electrical signals consume more energy and carry less information over longer distances and companies now are leaning on software to compensate for the bottlenecks in communications caused by the hardware.

In addition, today's "AI supercomputers" (high-end CPU plus accelerator) can't keep up with the computing power needed to train models and the software techniques used in the models are increasingly complex. The systems themselves also don't scale well enough: even those machines with more than 1,000 processors often run at less than 20 per cent efficiency, with the rest of the time sitting idle due to communications limits, the company says.

Luminous' plans come from years of research done by the founders and other company executives. Gomez was a research scientist at dating app company Tinder in 2018 and work in research roles at Google (in machine intelligence) and the Mayo Clinic, a software engineer at Bloomberg and a researcher of network biology at the Harvard Medical School. He dropped out of Stanford's MS program to launch Luminous.

Mitch Nahmias, co-founder and Luminous' CTO, received his BS, MA and PhD in electrical and electronics engineering from Princeton. While at the school, Nahmias researched the relationship between a laser and a biological spiking neuron, which helped create the field of neuromorphic photonics.

Matt Change, the company's vice president of photonics, also received a PhD in electrical engineering from Princeton and spent two years at Apple designing hardware to reduce interference between co-existing wireless radios on the Apple Watch, leaving in 2019. Before that, he was CTO at Rebeless, a startup creating photonic integrated circuit technology for wideband signal processing.

Luminous is not revealing details of its plans, though a 2019 report about its $9m in seed money said the goal is to create a single photonics chip that can replace 3,000 TPUs, which are small ASICs created by Google designed for machine learning inference in low-power devices.

Luminous has more than 90 employees and has working prototypes of its chips in its labs, with plans to ship development kits to organizations within two years.

While there is a lot of work being done in the area of silicon-photonics computing, it's likely still while companies address such hurdles as cost, scalability, power consumption and reliability. ®


Other stories you might like

Biting the hand that feeds IT © 1998–2022