Google's OCS Apollo[0] achieved roughly the same losses and switching latency using MEMS, but with 136x136 connectivity. They expect it to scale beyond 1000x1000.
What's the unique selling point for this new solution? Cost? Something quantum-related MEMS can't do?
[0]: https://arxiv.org/abs/2208.10041