San Francisco: Semiconductor startup Volantis has raised $88 million in a Series A funding round to develop a new optical technology designed to overcome one of the biggest hardware limitations facing modern artificial intelligence systems: the difficulty of moving data efficiently between AI processors and memory.

The company is developing a system that uses laser light instead of conventional electrical connections to link computing chips with memory chips. Volantis says the approach could dramatically increase both memory capacity and bandwidth available to AI processors.

Tackling the AI Memory Bottleneck

Powerful AI processors can perform enormous numbers of calculations, but their performance can be restricted when they cannot receive data from memory quickly enough.

Current high-end AI chips from companies such as NVIDIA and AMD use high-bandwidth memory placed close to the processor. However, conventional electrical connections have limited physical reach, restricting how many memory components can be connected around a GPU.

Volantis says its optical approach could allow up to 220 memory chips to be connected around a single GPU, compared with around eight in current high-end configurations.

Using Laser Light Instead of Electrical Wires

The company’s technology relies on vertical-cavity surface-emitting lasers, or VCSELs, to transmit information using light.

VCSEL technology is already used in consumer electronics, including systems associated with facial recognition. Volantis is adapting the technology for high-speed communication between AI processors and memory.

The company says its optical connections can travel much farther than conventional electrical connections, potentially allowing large numbers of memory components to operate as a unified, high-bandwidth memory pool.

New Funding to Accelerate Development

The $88 million Series A round was co-led by Lachy Groom and Abstract Ventures. Investors also included John Doerr, VXI Capital, Triatomic and Susa Ventures.

Several prominent technology and AI figures participated as angel investors, including AI researcher and investor Naveen Rao, AI researcher Sholto Douglas and AI podcaster Dwarkesh Patel.

The funding will support Volantis’ development of its optical AI inference system and its planned hardware products.

Targeting Extremely Large AI Models

Volantis says its first system, known as A-1, is being designed to support AI models exceeding 20 trillion parameters while targeting speeds of up to 10,000 tokens per second per user.

The company says its architecture is intended to increase memory capacity and bandwidth simultaneously rather than forcing engineers to choose between the two.

Such improvements could be particularly valuable for AI agents, coding systems and other applications that require models to process large quantities of information rapidly.

Why Optical Connections Matter

The shift from electrical to optical communication could become increasingly important as AI models continue to grow.

Larger models require more memory, while faster AI applications require that memory to deliver data to processors at extremely high speeds. Simply building more powerful GPUs does not completely solve the problem if the connection between processing and memory remains a bottleneck.

Volantis is therefore targeting the communication layer between these components rather than attempting to compete only by developing another conventional processor.

The company says its optical fabric can pool large numbers of memory chips together while maintaining high bandwidth, potentially changing how AI computing systems are designed.

Planned AI Hardware Could Arrive Next Year

Volantis aims to bring its technology to market with an AI inference chip designed to accelerate workloads such as software coding.

The company is attempting to build its system using established semiconductor and photonics technologies rather than requiring an entirely new manufacturing ecosystem. This could help it avoid some of the supply-chain challenges associated with developing completely new chip technologies.

If successful, the approach could provide AI developers with substantially greater memory capacity and faster data movement without relying solely on increasingly expensive high-bandwidth memory configurations.

Competition in AI Infrastructure

Volantis is entering a rapidly expanding market in which semiconductor companies and startups are looking for new ways to overcome AI infrastructure bottlenecks.

As AI models become larger and AI agents perform more complex tasks, the speed at which data can move between processors and memory is becoming as important as raw computing power.

The company’s photonic approach represents one possible path toward solving that problem.

The success of the technology will ultimately depend on whether Volantis can translate its laboratory and engineering claims into reliable, commercially scalable hardware. But the substantial new investment demonstrates growing interest in optical technologies as the AI industry searches for the next generation of computing infrastructure.