This week’s Spotlight is on Vladimir Stojanović, a Berkeley professor and co-founder of @AyarLabs whose work is helping solve one of the biggest constraints in modern AI infrastructure: moving enormous amounts of data between chips quickly enough and efficiently enough to keep up with compute. Vladimir’s research spans integrated circuits, computer architecture, and silicon photonics. In 2015, he helped lead the team behind the first processor capable of using light to communicate directly with the outside world - a Berkeley research breakthrough, in collaboration with MIT and the University of Colorado Boulder, published in Nature. That research became the foundation for Ayar Labs, which Vladimir co-founded and now serves as CTO. As AI systems scale across hundreds or thousands of GPUs, the connections between those processors become a major bottleneck. Traditional copper connections are limited in how much data they can carry, how far they can carry it, and how much power they consume in the process. Ayar Labs is addressing that problem by moving data with light. Its optical I/O technology is designed to connect thousands of GPUs across racks with the bandwidth, low latency, and power efficiency required for the next generation of AI systems. @thehousefund was proud to back Ayar Labs since the Seed round, when replacing electrical chip connections with optical ones still sounded like a highly ambitious research bet - and few were talking about AI. The company has since raised nearly $900M, reached a $3.75Billion valuation, and built partnerships across the semiconductor ecosystem as it moves the technology toward high-volume deployment. AI progress is often framed around bigger models and more powerful GPUs. But those systems only work if the underlying hardware infrastructure can move data fast enough to keep all that compute productive. Vladimir and the Ayar Labs team have spent more than a decade building for that reality. Another great example of Berkeley research becoming infrastructure the next era of technology may depend on.