AI Is Giving Optical Computing Another Shot

IEEE Spectrum
Read full story· 18 min
Optalysys, a UK startup, is advancing optical computing by integrating optical elements into data center networking paths to perform computations more energy-efficiently. Optical computing, driven by AI inference's power demands, is rapidly growing with a market now valued at $2.22 billion, aiming to complement or compete with traditional GPUs.

What happened

Optalysys, a startup based in the UK, is developing technology to perform computations using light within data center networks. The company is inserting optical computing elements into the paths where data is already transmitted optically, aiming to make AI inference tasks more energy-efficient. The optical computing market has grown rapidly, reaching a valuation of $2.22 billion, up from $0.4 billion in 2022. This growth is largely driven by the increasing energy demands of AI inference in large data centers.

Key details

  • Optalysys uses specially engineered waveguides to compute while data is in transit.
  • The optical computing market is valued at $2.22 billion as of the latest figures.
  • Market valuation was $0.4 billion in 2022, indicating rapid growth.
  • Optalysys's core technology includes an on-chip optical Fourier transform calculator.
  • Optical computing aims to reduce the energy used for matrix-vector multiplication, a key operation in AI inference.

Why it matters

Operators of data centers and companies running large-scale AI models are affected, as optical computing could lower the energy costs of AI inference. This technology offers an alternative or complement to traditional GPUs, potentially reducing operational expenses and enabling more efficient processing. The rapid market growth means more companies may soon have access to these new optical computing solutions, which could change how AI workloads are handled in data centers.


More in Industry & Manufacturing