發生了什麼事
San Francisco-based startup Volantis has raised $88 million in a Series A funding round led by Lachy Groom and Abstract Ventures, with participation from John Doerr, VXI Capital, Triatomic, and Susa Ventures. The company, founded by Tapa Ghosh and Roy Meade, aims to develop a photonic memory layer designed to accelerate AI by using light to connect compute and memory components. Volantis intends to begin customer deliveries in 2027.
Volantis has secured $88 million in Series A funding to develop a photonic link between AI compute and memory. The company is led by Tapa Ghosh, a Thiel fellow and former Y Combinator founder, and Roy Meade, a former vice president at Ayar Labs who previously managed Micron’s high-bandwidth memory program.
The company's A-1 system is designed to support AI models with more than 20 trillion parameters, with a stated goal of achieving speeds of up to 10,000 tokens per second per user. These figures are currently internal design targets rather than verified performance metrics from a functional prototype.
The funding round included participation from Lachy Groom, Abstract Ventures, John Doerr, VXI Capital, Triatomic, and Susa Ventures. The company plans to initiate its first customer deliveries in 2027.
為什麼這很重要
The funding highlights a strategic divergence in the semiconductor industry regarding the materials used for optical interconnects in AI infrastructure. While European public initiatives are investing heavily in indium phosphide—a material currently facing supply chain risks and price volatility—Volantis is designing its systems around gallium arsenide. This choice reflects a broader industry effort to mitigate reliance on materials where supply is heavily concentrated, specifically noting that China currently controls approximately 70% of the indium phosphide supply. By targeting high-performance metrics for models exceeding 20 trillion parameters, Volantis is positioning its technology to address the bandwidth bottlenecks inherent in scaling massive AI models, though these performance figures remain design goals rather than verified benchmarks.
The core of the industry tension lies in material science. Indium phosphide, which is central to European industrial photonics strategy, has seen its wafer prices rise by approximately 250% as of June, with industry leaders identifying supply access as a significant risk.
Volantis is explicitly designing its architecture to avoid indium phosphide, opting instead for gallium arsenide. This decision is framed as a strategic move to bypass supply chain constraints, given that China controls roughly 70% of the global indium phosphide supply.
The broader context involves a clash between public-sector industrial policy in Europe—which is investing hundreds of millions of euros into indium phosphide pilot lines—and private-sector American startups that are exploring alternative materials and architectures to solve the 'memory wall' in AI computing.
互動機制：它實際上是如何運作的
以互動方式探索這項發展背後的基礎技術。
Which component of an AI application is the machine-learning model itself?
接下來看什麼
The primary development to monitor is the 2027 timeline, which aligns with the scheduled operational launch of the Netherlands' EUR 153 million indium phosphide pilot line in Eindhoven. Observers should track whether the industry shifts toward the gallium arsenide approach favored by Volantis or if the European-backed indium phosphide ecosystem achieves the necessary scale and cost-efficiency to become the standard. Additionally, the company's ability to meet its stated performance targets—10,000 tokens per second for 20-trillion- models—will be a critical indicator of the viability of its photonic architecture compared to existing copper-based signaling solutions.
The 2027 delivery date is a critical milestone, as it coincides with the operational launch of the European indium phosphide pilot line in Eindhoven. The success of Volantis will depend on whether its gallium arsenide-based approach can provide a more reliable or cost-effective alternative to the state-backed European efforts.
Market competition remains intense, with other firms like Kandou advocating for the continued use of copper signaling, arguing that existing data center infrastructure makes copper a more practical long-term solution than a full transition to optical interconnects.
Investors and industry analysts will be watching to see if Volantis can transition from its current design goals to a functional, scalable product that can handle the massive bandwidth requirements of future 20-trillion- AI models.