Sanjai Kohlichases photons for a living.

The chip was called the GPS Spread Spectrum Receiver. The company built around it was SiRF Technology. By 2007 it sat inside more than seven out of every ten consumer GPS devices on Earth - the cheap ones, the ones in the dashboards of rental cars and the cargo holds of container ships and, eventually, the back of every smartphone.

Kohli is not a man you have heard of. He prefers it that way. The Times of India called him the father of mass-market GPS. The European Patent Office gave him their Inventor of the Year award in 2010, jointly with Chen. The IEEE made him a Fellow in 2012. And then he went and did three more startups.

He grew up in India and trained, of all things, as an aeronautical engineer - a B.Tech from IIT Bombay in 1979. Two years later he had an M.S. in Systems Science from Washington University in St. Louis. The aeronautics never quite stuck. The radios did.

What followed was thirty years of small bets in hard places: WirelessHome in 1999, which by 2003 owned roughly 40% of the last-mile wireless market. TrueSpan, building multi-standard chips for digital TV on cellphones, a problem most people in 2005 didn't yet know existed. Inovi, a millimeter-wave mesh networking company that Facebook bought in 2014 and turned into the bones of its Terragraph platform. Geoverify. Visible, working on integrated sensors for autonomous robots.

Then in 2019 the phone rang and it was a board. Wave Computing - a custom AI silicon firm that had acquired the storied MIPS architecture - needed a CEO. Kohli took the job. He kept it through the company's 2020 Chapter 11, kept it through emergence the following March, kept the MIPS brand alive long enough to see it relaunched.

In 2021 he turned up as Entrepreneur in Residence at Sutter Hill Ventures, the unflashy but ferociously successful venture firm just up Sand Hill Road. EIRs at Sutter Hill don't tend to stay EIRs for long.

By May 2023 he was CEO of CSpeed Inc., a Palo Alto company with about fifty-four employees and a public-facing tagline that says, simply, the future of datacenter connectivity. CSpeed works in silicon photonics - the discipline of routing information not through copper wires but through laser light moving across a chip. It is the field every hyperscaler is now obsessed with, because the bottleneck for an AI training cluster is no longer the GPU. It is the cable.

Kohli, who has spent his life squeezing signals through hostile media - GPS through urban canyons, wireless through walls, TV through 3G - is now squeezing photons through silicon waveguides. Same instinct. Different wavelength.

A GPU in 2026 can move more data than the wire connecting it to its neighbour can carry. Hyperscalers building AI clusters are running into a wall that has nothing to do with silicon transistors and everything to do with the copper traces between them. Light is the obvious answer. Light is also famously hard to wrangle inside a chip.

That is the field Kohli is now working in. Silicon photonics has been ten years away for thirty years. It is finally, quietly, arriving - because the economics finally demand it, and because the manufacturing tooling finally exists. CSpeed sits in the middle of that arrival.

It is a fitting third act. GPS taught the world where things were. The wireless mesh business taught it how to talk without wires. The next problem is teaching machines to talk to each other faster than copper allows. Kohli has done the first two. The third is the hardest. He is, by all accounts, characteristically uninterested in being loud about it.