The app for independent voices
$TSEM
<Tower Semiconductor × Oriole Networks: Nanosecond Optical Circuit Switching for AI>
Tower Semiconductor (TSEM) and Oriole Networks announced a collaboration (March 16, timed to OFC 2026) to commercialize nanosecond-speed optical circuit switching (OCS) on Tower's SiPho foundry platform — targeting AI cluster networking.
The problem: As GPU clusters scale beyond 16,000 nodes, multi-tier electrical packet switching creates severe tail latency and power bottlenecks. Communication overhead in existing networks typically runs in the tens of percent, meaning GPUs spend significant time waiting for data rather than computing.
Oriole's solution — PRISM/PRISM Ultra:
-Replaces all tiers of electrical switching with edge-based nanosecond OCS + fully passive glass core (zero power, zero cooling)
-PRISM Ultra delivers 51.2 Tb/s per xPU, 180ns xPU-to-xPU latency, and 1-hop contention-free connectivity scaling to 1 million nodes
-Reduces training communication overhead to under 1%
Tower's role: Provides the high-volume SiPho fab for Oriole's XTR modules (integrated photonic switch + transceiver), integrating lasers, optical amplification, switching, modulation, and detection on a single platform.
Strategic context: This is part of Tower's aggressive SiPho platform diversification in early 2026 — $NVDA (1.6T transceivers), Scintil (DWDM CPO lasers), Salience Labs (OCS), LightIC (LiDAR), Xanadu (quantum). Tower is positioning itself as the " $TSM of silicon photonics."
Watch for: Commercial deployment timeline, hyperscaler adoption signals, and how Oriole's radical full-stack OCS approach competes with $GOOGL's slower OCS and incremental CPO strategies.
