\n\n
Key developments:
- A Tsinghua University team unveiled the Optical Feature Extraction Engine (OFE2), a 12.5 GHz light‑based processor that moves AI’s most demanding tasks from power‑hungry electronics to ultra‑fast photonics, promising real‑time, low‑energy inference.
- India’s Ministry of Electronics and Information Technology approved a ₹99.94 crore (≈$10.3 M) Phase‑II grant for IIT Madras and izmo Microsystems, targeting 1.6 Tbps silicon‑photonic transceivers, co‑packaged optics, and quantum‑ready QRNG/QKD chips, underscoring national commitment to photonic data‑center infrastructure.
- Bengaluru startup Quanfluence raised $10 M (led by Chiratae Ventures) to build a full‑stack photonic quantum computer, aiming to ship its first processors by 2027 and a scalable fault‑tolerant system by 2029, marking a significant infusion of capital into the photonic quantum ecosystem.
- Aegiq and the University of Sheffield announced QGATE, a photonic quantum architecture that tolerates up to 25.98 % photon loss—four times the previous best—paving the way for more robust photonic qubits and reducing the error‑correction overhead that has hampered photonic quantum scaling.
- JEDEC released JESD264, the first reliability standard for silicon photonics, providing a common testing framework that will help manufacturers meet the stringent uptime demands of AI and high‑speed networking markets.
- Neurophos secured $110 M to commercialize silicon‑photonic optical compute units (OCUs), claiming massive on‑chip optical parallelism that can break the power wall limiting GPUs for AI workloads.
Why they matter:
These announcements collectively signal a transition of photonics from experimental lab demos to industry‑scale platforms. High‑speed optical engines like OFE2 demonstrate that light can directly perform AI inference at gigahertz rates, cutting energy use and latency. Government backing (India) and standards (JEDEC) create the ecosystem and trust needed for large‑scale deployment. Capital inflows into photonic quantum startups (Quanfluence, Aegiq) accelerate the path toward fault‑tolerant photonic quantum computers, while massive funding rounds for optical compute (Neurophos) aim to address the impending end of Moore’s law for AI.
What to watch next:
- Commercial roll‑out of OFE2‑style optical AI accelerators in data‑center edge nodes.
- Prototype shipments of the 1.6 Tbps silicon‑photonic transceiver from the IIT‑izmo partnership and its impact on co‑packaged optics adoption.
- Demonstrations of Quanfluence’s first photonic processor and benchmark comparisons against electronic quantum processors.
- Real‑world validation of QGATE’s loss‑tolerance in multi‑node photonic quantum networks.
- Industry uptake of JESD264, tracking how quickly major foundries adopt the standard.
- Deployment of Neurophos OCUs in AI training clusters and measurement of power‑efficiency gains. [[1]]