San Francisco-based semiconductor startup Volantis has raised $88 million in Series A funding to develop a photonic AI inference architecture designed to improve how computing chips communicate with memory. The round brings Volantis’ total funding to $97 million and comes as AI workloads increasingly face limits from memory bandwidth and capacity.
AI performance is no longer only about making processors faster. A major part of the challenge is moving data quickly enough between computing chips and memory.
That is the problem San Francisco-based semiconductor startup Volantis is targeting.
On October 1, 2026, Volantis announced an $88 million Series A led by Lachy Groom and Abstract Ventures. The company says the round takes its total funding to $97 million.
Modern AI systems need enormous amounts of data to move between compute and memory. Traditional electrical connections create physical limits on how many memory chips can sit close enough to a processor.
Reuters reported that Nvidia and AMD currently use high-bandwidth memory to address this challenge, but electrical interconnect limitations constrain scalability.
Volantis is approaching the problem differently: using light.
Volantis is developing an optical interconnect architecture based on photonics. Its technology uses VCSELs — vertical-cavity surface-emitting lasers — to transmit data between compute and memory.
According to the company, this architecture can connect more than 220 memory chiplets around a compute system, potentially allowing memory capacity and bandwidth to scale together.
Volantis is developing its first inference system, called A-1.
The company says A-1 is being designed for models exceeding 20 trillion parameters and up to 10,000 tokens per second per user. These are company-stated specifications and should be treated as targets/claims rather than independently verified benchmark results at this stage.
Volantis plans to begin customer deployments in 2027.
The $88 million Series A was co-led by Lachy Groom and Abstract Ventures. Participants include John Doerr, VXI Capital, Triatomic and Susa Ventures, alongside angel investors including Dwarkesh Patel, Naveen Rao and Sholto Douglas.
The founding team includes semiconductor and photonics professionals with backgrounds at companies including NVIDIA, AMD, Broadcom and Ayar Labs.
Volantis is not trying to compete simply by building another general-purpose GPU.
Instead, it is targeting an infrastructure bottleneck underneath AI computing: the connection between compute and memory.
If its architecture performs as intended, the potential value comes from allowing AI inference systems to access larger memory pools while maintaining high bandwidth.
But the important distinction is that the technology is still moving toward commercialization. The first integrated inference engines are planned for customers in 2027, so the market has not yet had independent customer-level validation of the company's claimed performance.
Volantis shows another side of the AI startup ecosystem.
The biggest opportunities may not always be another consumer AI application or model company. Some startups are building the hardware and infrastructure required underneath the AI stack.
For investors and founders, the interesting question is not simply how much money an AI startup raises. It is whether the startup has identified a genuine infrastructure constraint and can turn a technically difficult solution into a commercially deployable product.
Volantis has raised the capital. The next milestone is execution.
reuters
serrari group
