By Sara Castellanos 

BOSTON -- Engineers working at a startup near touristy Faneuil Hall Marketplace are building chips that use laser beams instead of electrical signals to run artificial-intelligence applications 10 times faster than today's most advanced AI computer chips, using one-tenth of the energy.

Founded in 2017 and backed by $33 million in venture-capital funding, Lightmatter Inc. is among dozens of companies gaining interest from investors and corporations because of their novel approaches to computing. They are using light, quantum physics, molecular biology and new design methods to build chips and create data-storage techniques for future computing demands.

Quantum computing is the best-known of these new methods. Startups as well as tech giants including Alphabet Inc.'s Google and International Business Machines Corp. are developing quantum computers, which harness the properties of quantum physics to sort through a vast number of possibilities in nearly real time. The advent of quantum computing has paved the way for other experimental techniques, startup executives say.

The market for new computing technology comes as advancements in traditional chip making are hitting a physical limit under Moore's Law, the idea that every two years or so, the number of transistors in a chip doubles.

At the same time, advances in artificial intelligence, easier access to huge troves of data and the continuing digitization of business processes are putting new demands on corporate and scientific computing.

To address the challenge, some startups are making chips focused on specific software tasks. Others are pushing further, finding processing and storage solutions in new materials, including synthetic DNA.

Three miles northwest of Lightmatter's headquarters, another Boston startup, Catalog Technologies Inc., is developing a unique way of storing large amounts of data. The company recently showed it could store 14 gigabytes of data from Wikipedia.org in DNA molecules, which look like a few drops of water in a test tube.

At Catalog, machines typically found at a molecular-biology laboratory are used to "print" sequences of synthetic molecules that store and represent digital information as bits of DNA. The information is read back by using DNA-sequencing machines and a computer with proprietary software that translates those molecules into the original text, photo or video.

DNA storage doesn't require cooling the way data centers do, meaning the method holds promise for storing huge amounts of data more efficiently.

The startup's founders said in 2018 that they had raised $9 million from investors such as New Enterprise Associates. That brought total funding to $11 million since Catalog's founding in 2016. The company is set to announce more funding in a few months, said Hyunjun Park, co-founder and chief executive.

Devin Leake, the company's chief science officer, said the startup is benefiting from the momentum of cutting-edge computing methods. "The groundwork that quantum computing has done -- and the idea of neuromorphic computing -- has laid the foundation for considering alternatives," Mr. Leake said.

New computing methods are largely experimental and there are challenges to scaling them to meet corporate demands. Still, investment in computing technology startups based in the U.S. has grown to a record $477.9 billion in 2019, up from $59.8 billion in 2014, according to data firm PitchBook Data Inc.

"If you don't come up with more exotic ways of computing, computing itself [won't] progress like it has for the last 50 years," said David Moehring, formerly chief executive of quantum-computing startup IonQ Inc.

Mr. Moehring co-founded venture-capital firm Cambium Capital Partners in 2018 to invest in cutting-edge computing startups. Among them is Vorticity Inc., founded last summer by Chirath Neranjena, who previously worked for Alphabet Inc.'s research arm X. Vorticity is revamping chips by reorganizing components such as memory in order to make them more efficient. The goal is to speed up complex scientific calculations for industries including mining and aerospace.

The rise in alternative computing methods is partly driven by companies' growing demand for powering processes such as artificial intelligence. Current AI models running on traditional chips require massive computing power from data centers.

"There's a lot of high-performance compute requirements coming our way....The question is, are there better ways of doing it?" said Martin Hofmann, chief information officer of Volkswagen AG, the world's biggest auto maker by sales.

For more than three years, the Germany-based company has been experimenting with using quantum computing for various applications, including speeding up the time it takes to train neural networks, one of the key AI technologies underpinning self-driving cars.

Quantum and other exotic computing technologies are "fraught with risk" for investors because they rely on new architectures, said Adam Fisher, partner at Bessemer Venture Partners.

Still, Bessemer is looking to make an investment in the quantum-computing sector. "We're looking very seriously," said Mr. Fisher, who led the first major investment in AI chip-making startup Habana Labs in 2016. The four-year-old company was recently acquired by chip giant Intel Corp. for $2 billion.

Intel is ramping up investments in novel computing methods by developing quantum chips and neuromorphic chips. Neuromorphic chips, which are modeled after the body's nervous system, can process large amounts of data and learn from it in real time to make "good enough" predictions automatically, similar to the way humans learn about the world, said Mike Mayberry, Intel's chief technology officer.

Write to Sara Castellanos at sara.castellanos@wsj.com

 

(END) Dow Jones Newswires

February 04, 2020 12:38 ET (17:38 GMT)

Copyright (c) 2020 Dow Jones & Company, Inc.
Intel (NASDAQ:INTC)
Historical Stock Chart
From May 2020 to Jun 2020 Click Here for more Intel Charts.
Intel (NASDAQ:INTC)
Historical Stock Chart
From Jun 2019 to Jun 2020 Click Here for more Intel Charts.