AMD announced the acquisition of startup Enosemi based in Silicon Valley that specializes in development of photonic integrated circuits. The move, announced Wednesday, is designed to accelerate AMD’s co-packaged optics for AI systems, AMD said.
A photonic integrated circuit is a microchip that often combines multiple optical components like lasers, detectors and modulators on a single chip. Unlike electronic integrated circuits with transistors and capacitors, photonic integrated circuits use photons, also known as light particles, to transmit data. They are designed to provide high speed and low power consumption. They are used in high-speed data transmission, sensing and optical computing, which opens them for use in medical, automotive and communications applications.
Financial terms of the deal were not disclosed. Pitchbook lists Enosemi as employing 16 workers with 11 investors that was founded in 2023. The company’s website lists several products, including chiplets that include one chip with a photodetector, TIA and digital control. Custom silicon is also available.
Co-packaged optics for next-gen AI will be crucial as more complex AI models emerge, requiring faster movement of data. AMD sees the move as bolstering its role as a full-stack AI provider, along with leading CPUs, GPUs and SoCs combined with enhanced networking, software and systems integration expertise.
In a blog, Brian Amick, SVP of technology and engineering at AMD, wrote that the elite team and top talent at Enosemi “has a proven track record of building and shipping photonic integrated circuits in volume, a unique feat that few select teams have accomplished.”
The Enosemi team at AMD “will help us immediately scale our ability to support a variety of photonics and co-packaged optics solutions across next-gen AI systems.”
Previously, AMD has integrated Xilinx and Pensando as well as Silo AI, Mipsology and most recently ZT Systems.
Co-packaged optics will deliver higher bandwidth and better power efficiency, Amick claimed. He called that ability a “transformative step in system architecture where tighter integration between compute and networking is enabled to support the performance and scale that advanced AI workloads require.”
Indeed, the move to full-stack across compute, networking, architecture, software and other elements is how AI providers are hoping to distinguish themselves. GPU leader Nvidia has taken a similar approach, while Intel—trailing in the GPU game—markets itself in similar manner.
In March, Enosemi and Jabil announced a collaboration in advanced packaging process technology.