Ayar Labs bags $150M in additional late-stage funding to help make bigger AI chip clusters

Networking technology startup Ayar Labs Inc. said today it has extended its recent late-stage funding round to $650 million after raising an additional $150 million from an unnamed investor.

The company originally disclosed it had raised $500 million in its Series E round back in March, which was led by Neuberger Berman. Nvidia Corp., Advanced Micro Devices Inc. and MediaTek Inc. also participated in that round, which brought its total amount raised to $870 million. With today’s Series E extension, Ayar Labs has now raised just over a billion dollars in outside funding.

The enormous amount of money being funneled its way underscores the massive potential of its optical interconnect technology, which enables artificial intelligence chips to be linked to fiber optical cables and speed up data transmission in AI clusters. Ayar Labs says its big ambition is to help AI systems scale even bigger than they already are by replacing the copper interconnects that are currently widespread in AI data centers. By doing that, it believes it can accommodate the growth of even bigger chip clusters and larger foundation models.

Copper interconnects have become a major bottleneck in AI data centers, because they suffer from three major problems – distance limitations, power and temperature. Copper cables only work well over short distances, because the data signals start to degrade the further the cable extends.

In addition, copper cables are power-hungry, and when AI clusters grow to thousands of graphics processing units, which must all be linked with copper wires, the cost of the energy requirements becomes unsustainable, the startup says. The third problem is that copper wires often overheat when connected to multiple high-performance processors, causing workloads to break down.

Ayar Labs’ solution is a product called TeraPHY, which swaps out the copper for optical interconnects that have greater bandwidth, performance, power efficiency and lower latency, meaning they can potentially link much larger clusters of GPUs across data centers. The company has also developed co-packaged optics technology to place its TeraPHY Optical Engine directly into the ASIC socket of servers, alongside the GPU, further enhancing connectivity.

“Copper interconnect is becoming the limiting factor for AI scale-up,” said co-founder and Chief Executive Mark Wade. “This funding allows us to move faster on manufacturing-ready co-packaged optics at scale. We’re deepening our work across the foundry and packaging ecosystem and growing our engineering capacity to support customer programs worldwide.”

Ayar Labs has been doing this for a while, but it’s only in recent years that the startup has started getting real momentum. Founded way back in 2015, it didn’t raise any significant capital until five years later, when it finally received $35 million in funding from outside investors. But with the AI boom, development of optical interconnects has suddenly taken on a real sense of urgency, which is why some of its biggest backers are the chipmakers that stand to benefit from bigger clusters of chips.

Wade told Reuters the company now hopes to bring its technology to the mass market by early 2028. The money from the Series E round has already helped to pay for a new chip design center in Bengaluru, India, and the extra cash from today’s extension will help it to start preparing its products. But it needs to work fast.

“We build the optical chip, and our optical chip gets fed into the customer’s product,” he explained. “If my customer’s products are looking to ramp in the 2028-2029 time frame, we have to have all of our stuff qualified for volume production by the end of 2027.”

The startup can at least expect some help from its new partner Wiwynn Corp., a Taiwanese data center infrastructure company that has also invested in it. Ayar Labs said Wiwynn made a strategic investment earlier this year, but it’s not clear if that’s where the Series E extension money comes from. It has been working with Wiwynn to deliver optically connected, rack-scale AI systems that can support hyperscale workloads. The systems leverage SuperNova remote light source technology to integrate directly into Wiwynn’s rack-level server architecture.

“Co-packaged optics is a foundational technology for the next generation of AI and cloud data centers,” said Wiwynn CEO William Lin. “Through our collaboration and investment with Ayar Labs, we see a clear path to new architectures that reduce the power and complexity limits of copper interconnects, powering rack-scale deployment for the next wave of hyperscale AI data centers.”

Photo: Ayar Labs