The frontier sibling of Optical Circuit Switching (OCS) — switching at packet granularity rather than dedicated lightpaths.
Mechanism
OPS switches individual packets in the optical domain, rather than holding a dedicated lightpath (the Optical Circuit Switching (OCS) approach). This is the harder, holy-grail end of optical switching: it needs fast all-optical logic and, historically, optical buffering — and the absence of practical optical RAM is the blocker that has stalled OPS for ~two decades. Because it is so much harder, the field is far less crowded than OCS, which is the bull-case framing for a startup that can make it work.
- Does optical packet switching (NOT circuit switching of the Google-OCS kind) for scale-out networking; claims ~1,000× lower latency and lower power than electro-optic conversion via a passive optical approach.
- Tech: Lithium Niobate (TFLN) used for frequency mixing (the non-linearity), not modulation; poling is the hard step. First patent is an all-optical passive ALU (Non Linear Photonic Logic).
- GTM: a passive-switching PIC module / custom NIC IP block, partnering with NIC OEMs rather than competing on silicon.
The framing that matters: packet-vs-circuit is the differentiator. Anyone can do OCS; packet-granularity all-optical switching is the harder, less-crowded problem.
Companies using
Connected ideas
Frontier (open questions)
- Practical all-optical packet switching without optical RAM/buffering?
- Venture-scale exception to the pattern where no standalone OCS company has ever reached venture scale (they get acquired or stay small)?
- Real switching speed + port count of leading approaches?