Ayar Labs raises $35 million to develop light-based chip interconnects

Enterprise

Ayar Labs, a startup leveraging novel silicon processing techniques to develop high-speed, high-density, low-power optical-based interconnect “chiplets” to replace traditional I/O, today raised $35 million. The company says the proceeds will be put toward product development and commercialization of its in-package optical interconnect solution as well as growing Ayar’s footprint internationally.

Making transistors — the tiny on-off switches inside silicon chips — smaller and smaller has enabled the computer revolution and the $1 trillion-plus electronics industry. But the copper interconnections between transistors on electronics are a major impediment to progress. The resistive-capacitive delay, or RC delay, within interconnects has prevented transistors from becoming faster as they shrink, as is normally the case. Moreover, copper remains unreliable at small dimensions even with insulators made of shielding materials.

Ayar, which was cofounded in 2015 by Alex Wright Gladstein, Chen Sun, and Mark Wade, aims to develop products that move data from chip to chip using light instead of electricity transmitted through copper wires. Based on a decade of research and collaboration between MIT, the University of California, Berkeley, and Colorado University Boulder, Ayar claims to have created a solution that overcomes the power and performance scaling challenges of semiconductors as well as the interconnect bandwidth bottleneck between those devices.

Ayar Labs

Photonic circuits like Ayar’s, which are also being investigated by startups like Lightelligence, LightOn, and Lightmatter, require only a limited amount of energy because light produces far less heat than electricity. They also benefit from reduced latency and are less susceptible to changes in temperature, electromagnetic fields, and noise.

Ayar’s chiplets and multi-wavelength lasers replace electrical-based I/O. The chiplets can drive a 1000 times improvement in interconnect bandwidth density at 10 times lower power, the company says, enabling new system architectures for AI, cloud, high-performance computing, 5G, and lidar.

For instance, Ayar’s TeraPhy chiplet is theoretically capable of tens of terabytes per second of bandwidth thanks to a modular multiport design with ports carrying eight channels of light. TeraPhy features a wide, high-bandwidth electrical interface to connect to partner silicon and a light supply called SuperNova, a photonic integrated circuit that generates eight or sixteen wavelengths of light, multiplexes them, splits the power, and amplifies it to eight or sixteen output ports. Ayar says that SuperNova can supply light for 256 channels of data, which translates to 8.192 terabytes per second.

“These smart optical I/O chiplets enable system-on-chip companies and system-integrators to focus on their core function integration and process scaling while offloading the I/O tasks to low-power, high-throughput and reach optical I/Os,” Ayar Labs’ website reads. “This enables logically connected and physically federated systems.”

Santa Clara-based Ayar’s series B round, which was co-led by Downing Ventures and BlueSky Capital alongside new investors Applied Ventures, Castor Ventures, Downing Ventures, and SGInnovate, brings the company’s total raised to date to over $60 million. Existing investors including BlueSky Capital, Founders Fund, GlobalFoundries, Intel Capital, Lockheed Martin Ventures, and Playground Global also participated.

Sign up for Funding Weekly to start your week with VB’s top funding stories.

Products You May Like

Articles You May Like

Square Enix Says Sales of AAA Games Like Final Fantasy 7 Rebirth, Final Fantasy XVI Fell Short of Expectations
ChatGPT Integrates Google Drive and Microsoft OneDrive For Paid Users With Connect Apps Feature
#Tesla is rehiring some #supercharger workers it fired #technology #shorts
Google Pixel 8a Gets AI Wallpaper Generator With First Software Update: Report
MITRE Unveils EMB3D: A Threat-Modeling Framework for Embedded Devices

Leave a Reply

Your email address will not be published. Required fields are marked *