Semiconductor startup Volantis has raised $88 million in Series A funding to develop a new photonics-based architecture designed to address one of the most significant performance constraints in artificial intelligence computing.
The San Francisco-based company is building technology that uses optical connections to move data between AI processors and memory, replacing some of the electrical links that currently limit how quickly large models can access the information they need.
The funding round was co-led by Lachy Groom and Abstract Ventures, with participation from investors including John Doerr, VXI Capital, Triatomic and Susa Ventures. Volantis has now raised approximately $97 million in total funding.
Targeting the AI Memory Wall
As artificial intelligence models become larger and more sophisticated, raw processing power is only one part of the performance challenge.
AI accelerators must constantly transfer huge volumes of data between computing cores and memory.
The speed at which that happens, commonly referred to as memory bandwidth, can become a major bottleneck.
Current AI systems typically rely on high-bandwidth memory positioned close to GPUs. However, conventional electrical connections restrict how far memory can be placed from the processor and therefore limit the amount that can be connected efficiently.
Volantis is attempting to change that architecture by using light to move data.
Its optical fabric is designed to connect much larger pools of memory to AI processors while increasing both memory capacity and bandwidth.
Photonics Enters the AI Infrastructure Race
The company is using a technology known as vertical-cavity surface-emitting lasers, or VCSELs, to transmit information optically.
VCSEL technology is already widely manufactured and used in products including consumer electronics and facial-recognition systems, giving Volantis access to an established supply chain rather than requiring an entirely new manufacturing ecosystem.
Volantis says its architecture could enable more than 200 memory chips to operate around an AI processor, compared with the much smaller number that can typically be connected using conventional approaches.
That could substantially increase the amount of data available to processors while improving the speed at which information is delivered.
The company believes that overcoming this limitation could allow increasingly large models to operate at far higher inference speeds.
“The next AI infrastructure battle may be less about raw compute and more about how quickly data can move around it.”
Building the A-1 AI System
Volantis is developing its technology around a system called A-1.
The company says the platform is being designed to support models exceeding 20 trillion parameters while delivering inference speeds of up to 10,000 tokens per second per user.
Those figures remain design targets rather than independently verified performance results, but they illustrate the scale of improvement Volantis is attempting to achieve.
The company is targeting delivery of its first integrated inference engines to customers in 2027.
Funding from the latest round will support engineering recruitment, product development and the commercialisation of the A-1 architecture.
A New Front in AI Competition
The rise of generative and agentic AI has triggered enormous investment in processors, data centres, networking and power infrastructure.
However, as systems become increasingly powerful, attention is shifting towards the technologies that connect those components together.
Memory bandwidth has emerged as one of the most important constraints.
Simply adding more processing capacity does not necessarily improve performance if processors cannot access data quickly enough.
That is creating opportunities for companies working across advanced packaging, optical interconnects, memory and networking.
Volantis is one of a growing number of businesses attempting to use photonics to address those challenges.
Experienced Semiconductor Team
The company’s founding team includes engineers with experience at Nvidia, AMD, Broadcom and Ayar Labs.
Members of the team have previously worked on technologies including advanced semiconductor packaging and silicon photonics systems.
That experience is important as Volantis attempts to move its architecture from development towards commercial production.
Rather than designing a completely new semiconductor manufacturing process, the company is seeking to integrate photonic technology with existing industry supply chains.
This could potentially reduce some of the manufacturing barriers traditionally associated with introducing new optical technologies into computing systems.
AI Infrastructure Moves Beyond the GPU
The funding round highlights how investment in artificial intelligence is increasingly expanding beyond the processors themselves.
GPUs remain at the heart of modern AI infrastructure, but memory, networking, optical communications and data movement are becoming equally important as models grow.
For investors, that is creating opportunities throughout the technology stack.
For Volantis, the ambition is to remove one of the constraints that could otherwise limit the next generation of AI systems.
If photonics can deliver significantly greater memory capacity and bandwidth without dramatically increasing cost or complexity, it could become an increasingly important part of future AI infrastructure.
With $88 million in new funding and its first commercial systems planned for 2027, Volantis is positioning itself at the centre of that emerging technology race.