It's a good time to be an AI chip startup, especially if you happen to specialize in silicon photonics.

On Tuesday, Nvidia-backed startup Ayar Labs raised $500 million to accelerate the mass production of its co-packaged optics (CPO) tech.

The cash infusion comes a day after Nvidia said it would inject $4 billion into photonic networking providers Coherent and Lumentum ($2 billion each) to scale up their manufacturing capacity in anticipation of demand.

Founded in 2015, Ayar Labs' TeraPHY chiplets provide an alternative to copper for chip-to-chip communications that's capable of supporting higher bandwidths over longer distances.

Above 800 Gbps, copper interconnects are limited to a couple of meters and often require retimers to keep error rates from getting out of hand. Because of this, higher-speed copper interconnects, like those found in Nvidia's NVL72 systems, are usually constrained to the rack while pluggable optics are used for rack-to-rack communications.

Pluggable optics support much longer ranges, but that reach comes at the expense of higher power consumption and latency. This is the problem that silicon photonics providers, like Ayar, are trying to address.

By integrating its TeraPHY chiplets directly into the GPU or accelerator, Ayar says its designs can support significantly higher bandwidth while using a fraction of the power required by pluggables.

Over the past few years, Ayar has worked to validate its tech with multiple prototypes including one built in collaboration with Intel and the Defense Advanced Research Projects Agency (DARPA).

More recently, Ayar began working with Taiwanese semiconductor design services provider Global Unichip Corp (GUC) to develop reference designs based on its optical I/O chiplets.

One of these reference designs was developed in collaboration with Alchip and uses eight of Ayar's next-gen TeraPHY chiplets. Combined, the company claims it can support more than 200 Tbps of aggregate bandwidth per package.

For reference, that's about 5x more bandwidth than we see on Nvidia's Rubin GPUs, which top out at 28.8 Tbps (3.6 TB/s) of bidirectional bandwidth. What's more, since Ayar's interconnects are optical, the links aren't limited to a single rack.

"We want to be able to scale up to 10,000 GPU dies connected in a scale-up domain, while keeping the rack power and power density to around 100kW," Ayar CTO Vladimir Stojanovic told El Reg in an exclusive interview late last year.

• Nvidia burns $4B to light up American photonics manufacturing
• Microsoft's Project Silica promises eternal storage. It can't get there from here
• Let there be light! DARPA seeking physics-defying photonic computers to supercharge AI
• Positron: we don't need no fancy HBM to compete with Nvidia's Rubin

With datacenters increasingly facing power constraints and the performance advantage of scale-up systems, like Nvidia's NVL72 or AMD's Helios, becoming clearer, it isn't surprising to see venture capitalists lining up to push photonics startups like Ayar over the finish line.

In addition to scaling up high-volume production and test capacity, Ayar says the funding will enable it to expand its global operations beginning with a new office in Hsinchu, Taiwan.

The $500 million Series E funding round was led by Neuberger Berman but included a whole host of investors, including MediaTek and (again) Nvidia, which dropped its first cash into the startup back in 2022, according to Pitchbook.

Ayar is one of many photonics startups trying to carve out a niche for themselves.

Lightmatter is another startup playing in the CPO space. Last year the company launched a photonic interposer alongside an optical I/O chiplet in a similar vein to Ayar's TeraPHY.

Meanwhile, others have already found their exit. In February, IP giant Marvell Technology completed its roughly $3.25 billion acquisition of Celestial AI. ®