Volantis raises $88 million to connect AI chips and memory chips using laser technology

Volantis Secures $88 Million to Address AI Hardware Bottleneck

San Francisco-based semiconductor startup Volantis announced it has raised $88 million in Series A funding to solve a fundamental challenge facing AI chips from Nvidia and Advanced Micro Devices1

1

. The funding round was led by Stripe veteran Lachy Groom and Abstract Ventures, with participation from prominent investors including John Doerr, who backed Google and Amazon in their early days, along with VXI Capital, Triatomic, and Susa Ventures1

1

. Angel investors include AI podcaster Dwarkesh Patel, AI chip veteran Naveen Rao, and Anthropic researcher Sholto Douglas1

1

.

Tackling the Memory Bandwidth Challenge

The fundamental limit for AI chips lies in how efficiently the computing parts can communicate with memory chips where AI models reside and feed data to the computing chip1

1

. Current solutions from Nvidia and AMD involve encircling computing chips with expensive high-bandwidth memory chips, but even Nvidia's best GPU offerings can hold only eight memory chips due to the limited reach of tiny electrical wires connecting them1

1

. This constraint creates a significant AI hardware bottleneck that limits performance for AI inference tasks.

Laser-Based Communication Breakthrough

Volantis is developing a photonic memory layer that sends data between computing chips and memory chips using beams of laser light, eliminating the reach problem inherent in electrical connections1

1

. This laser-based communication approach would enable the company to pack 220 memory chips around a GPU, a dramatic increase from current limitations1

1

. The technology provides more than 30 times the memory bandwidth of current accelerators, fundamentally changing data transfer speeds for AI applications3

3

.

VCSELs: Proven Technology for New Applications

To achieve this breakthrough, Volantis is utilizing vertical-cavity surface-emitting lasers, or VCSELs, which already power facial recognition features in hundreds of millions of iPhones1

1

. Apple invested heavily in bolstering the VCSEL supply chain over the past decade, making these components relatively accessible1

1

. VCSELs are easier and cheaper to manufacture than lasers typically used in optical networking devices because they can be made from gallium arsenide, a more readily available material than alternatives3

3

.

Strategic Material Choice Amid Global Supply Concerns

Volantis deliberately uses gallium arsenide lasers to sidestep constraints in indium phosphide, of which China holds 70% of global supply2

2

. Indium phosphide wafer prices rose approximately 250% by June, with industry leaders calling access to them a key risk for the semiconductor industry2

2

. This material choice positions Volantis to avoid supply chain bottlenecks that have plagued other AI chip developers1

1

.

The A-1 System: Targeting Frontier AI Models

Volantis plans to ship its photonic inference systems with a data center inference appliance called the A-1, which is approximately one-third the size of a standard server rack3

3

. The system features 10 terabytes of memory with 250 terabits per second of memory bandwidth3

3

. The A-1 system targets models above 20 trillion parameters at up to 10,000 tokens per second per user2

2

. According to CEO and co-founder Tapa Ghosh, this capability will "enable real-time frontier inference, restart scaling laws & enable entire code bases in context windows," allowing coding agents to complete tasks in 30 seconds rather than 30 minutes3

3

.

Experienced Team with Proven Track Record

Volantis' founding team brings strong industry credentials, with engineers who previously worked at major chipmakers including Nvidia and Broadcom3

3

. Their technical achievements include the first commercial implementation of CoWoS interconnect, widely used in graphics processing units3

3

. Co-founder Roy Meade previously ran Micron's high-bandwidth memory programme and served as vice president at Ayar Labs2

2

. CEO Tapa Ghosh is a Thiel fellow and former Y Combinator founder2

2

.

Timeline and Market Implications

Volantis plans to deliver its first chip next year that would accelerate tasks like AI coding, with first customer deliveries of the A-1 system scheduled for 20271

1

2

2

. Ghosh emphasized the pragmatic nature of their approach, stating, "Advanced packaging is always to be respected - it's never trivial - but it's not necessarily a new thing to do. No one is going to win a Nobel Prize if our project works, but the good news is, they won't need to"1

1

. By bringing together relatively well-known components, Volantis aims to avoid the supply chain bottlenecks that have slowed other AI chip developments1

1

. The company's ability to dramatically increase memory bandwidth while using proven technologies positions it to address one of the most pressing limitations in current AI infrastructure, potentially reshaping how frontier AI models are deployed and scaled.