Volantis Raises $88 Million to Break the Memory Wall with Lasers Borrowed from iPhones
The San Francisco startup is developing optical interconnects that would let a single GPU talk to 220 memory chips instead of eight, using vertical-cavity surface-emitting lasers already manufactured at scale for Apple's facial recognition systems.
SAN FRANCISCO: The fundamental constraint on AI chip performance is no longer the speed of computation but the distance data must travel between a processor and its memory. Volantis, a semiconductor startup based in San Francisco, has raised $88 million in venture capital to solve that problem using beams of laser light, tapping a technology that already sits inside hundreds of millions of iPhones (Reuters, 2026).
The bottleneck is well understood across the industry. Nvidia and AMD both encircle their GPUs with high-bandwidth memory chips, but even Nvidia's best current offerings can accommodate only eight memory chips per processor because of the limited reach of the tiny electrical wires connecting them. The computing power is there, yet the memory cannot feed it fast enough. Volantis says its optical approach eliminates that reach problem entirely, potentially allowing 220 memory chips to be packed around a single GPU.
The technology relies on vertical-cavity surface-emitting lasers, or VCSELs. While the name sounds exotic, these tiny lasers are already mass-produced for facial recognition features in Apple devices, a supply chain the iPhone maker spent a decade strengthening. Volantis is betting that by assembling relatively well-known components rather than inventing new ones, it can sidestep the supply chain bottlenecks that have slowed other AI chip ventures and deliver a product next year targeting workloads like AI coding.
"Advanced packaging is always to be respected - it's never trivial - but it's not necessarily a new thing to do," said Tapa Ghosh, CEO and co-founder of Volantis. "No one is going to win a Nobel Prize if our project works, but the good news is, they won't need to."
The funding round was led by Stripe veteran Lachy Groom and Abstract Ventures. John Doerr, an early backer of Google and Amazon, also participated, alongside VXI Capital, Triatomic, and Susa Ventures. Angel investors included AI podcaster Dwarkesh Patel, AI chip veteran Naveen Rao, and Anthropic researcher Sholto Douglas.
BuiltWorld AI Operational Take: The memory wall has become the defining constraint on AI infrastructure economics, and Volantis is proposing a solution that swaps electrical traces for optical ones. For data center operators, the implication is significant. Optical interconnects could dramatically increase the compute density achievable within a single rack, but they also introduce new requirements around thermal management for laser components and optical signal integrity across dense packaging. If Volantis delivers on its 2027 timeline, the design assumptions baked into current AI factory reference architectures will need to be revisited.