Nvidia Corp. has inked a technology partnership with Ayar Labs Inc., a startup developing optical chips that promise to increase the speed and efficiency of data center infrastructure.

The partnership was announced today, just a few weeks after Nvidia backed a $130 million funding round for the startup. The chipmaker was joined in the round by several other tech giants, including Intel Corp., which participated through its Intel Capital investment arm. Ayar Labs has raised a total of $194.7 million since launch. 

Nvidia will work with Ayar Labs to develop new artificial intelligence infrastructure offerings based on the startup’s optical chip technology. In particular, the companies are hoping to build “scale-out architectures enabled by high-bandwidth, low-latency and ultra-low-power optical-based interconnects.” An interconnect is a technology that links together multiple separate chips.

Linking together chips is essential for many enterprise AI use cases. With the help of an interconnect, a company can connect multiple graphics cards with one another and use them to run its AI software faster than would be possible using a single processor.

The speed of an interconnect directly influences the performance of the AI hardware that it powers. The faster data can travel between the chips in an AI environment, the faster processing is carried out.

Ayar Labs has developed a new interconnect technology that it says is significantly faster than existing products. Usually, an interconnect transmits data between the chips that it links together in the form of electricity. Ayar Labs’ technology transmits data in the form of light to increase performance.

The startup provides its technology in the form of a module called TeraPHY that can be built into processors such as graphics cards. A single TeraPHY module is capable of transmitting 2 terabits of data per second, according to Ayar Labs. The startup says that its technology can provide up to 1,000 times more bandwidth than traditional interconnects using one tenth the power. 

“Over the past decade, Nvidia-accelerated computing has delivered a million-X speedup in AI,” said Rob Ober, the chief platform architect for data center products at Nvidia. “The next million-X will require new, advanced technologies like optical I/O to support the bandwidth, power and scale requirements of future AI and ML workloads and system architectures.”

As part of the partnership, the companies plan not only to collaborate on product development but also to accelerate adoption of optical chip technology. The AI infrastructure market is only one of several areas where Nvidia could potentially apply optical interconnects. Ayar Labs says that its TeraPHY module can help improve performance in cloud data centers and supercomputing environments, which both represent major focus areas for Nvidia. 

Photo: Ayar Labs