The new A‑1 unit packs 10 TB of memory and 250 Tb/s of bandwidth, positioning it as a compact solution for high‑speed AI inference workloads slated for delivery next year.
Volantis has unveiled its A‑1 photonic inference appliance, a compact system designed to deliver real‑time AI processing in data centers. The device combines 10 TB of memory with an unprecedented 250 Tb/s of bandwidth, promising to accelerate inference workloads without the power and latency penalties of traditional electronic accelerators.
Key Features of the A‑1 Appliance
The A‑1 leverages Volantis’s proprietary silicon‑photonic interconnects to move data at speeds that rival fiber‑optic communication links. By integrating photonic pathways directly onto the inference chip, the appliance reduces the number of electronic‑to‑optical conversions, cutting latency to sub‑microsecond levels.
- 10 TB of high‑bandwidth memory
- 250 Tb/s aggregate photonic bandwidth
- Modular 2‑U rack‑mount form factor
- Support for major AI frameworks via standard APIs
Target Use Cases
Volantis positions the A‑1 for workloads that demand instantaneous decision‑making, such as autonomous vehicle sensor fusion, high‑frequency trading, and real‑time video analytics. The appliance’s low‑latency profile makes it suitable for edge‑proximate deployments where every microsecond counts.
Roadmap and Availability
The company announced that the A‑1 will enter limited production in Q4 2026, with broader availability slated for early 2027. Early adopters will receive a software development kit that includes reference models and integration guides to streamline deployment in existing data‑center environments.
Volantis secured $88 million in funding to accelerate development of its photonic inference systems, underscoring investor confidence in the market potential of optical AI acceleration.
SiliconANGLE coverage of Volantis’s photonic inference funding