Credo Technology Group is trying to turn micro-emitter optics from a company-specific product idea into an industry standard for AI data centers. The company has launched the Short-Reach Micro-Emitter Optics, or SRMO, Workstream inside the Open Compute Project, bringing together an unnamed hyperscaler plus AFL, ams OSRAM, BizLink, Foxlink, Heraeus Covantics, SENKO and YOFC.
The new workstream is focused on a problem that is getting harder as AI clusters expand: how to move very large amounts of data across short distances without relying entirely on bulky copper cables or higher-power conventional optical links. Credo’s October 12 announcement says the group intends to standardize link power, signaling, parallelism, connector and module footprints, link budgets and compliance requirements.
Community discussion ahead of OCP focused on Credo’s micro-emitter optical backplane and whether optics can begin replacing copper in shorter AI links.
The target is the 3–50 meter gap
Credo’s argument is that copper remains attractive inside racks because it is reliable, inexpensive and low power at very short distances, while traditional optics becomes necessary once distances increase. The company says micro-emitter optics can fill the middle ground by using very large numbers of lower-speed optical lanes in parallel instead of forcing a small number of lanes to run at extreme signaling rates.
Credo CEO Bill Brennan explains the “wider, slower, better” micro-emitter strategy at the 2026 Open Compute Project Global Summit.
Credo’s first product built around that approach is its 1.6T ZeroFlap Active Light Cable. The company says the cable uses micro-VCSEL emitters, reaches 50 meters, consumes 12 watts per cable end and is about 10 times thinner and lighter than the copper approach it is targeting. Credo says production is planned for 2027. Those are vendor claims and will still need to be validated at scale by customers.
Why going wider could reduce power
The most interesting technical claim is not simply bandwidth. Credo argues that a dense array of micro-emitters can run individual lanes at lower energy per bit while still delivering enormous aggregate throughput. In Brennan’s OCP presentation, Credo estimated that a future micro-emitter backplane could use less than 2 picojoules per bit versus about 5 picojoules per bit for the SerDes driving a comparable copper backplane, implying a potential power reduction of roughly 60% at the link layer.
Credo’s Don Barnetson demonstrates the company’s 1.6T ZeroFlap optical platform and predictive link telemetry at AI Infra Summit 2026.
That power argument matters because networking is becoming a larger part of the rack-level power budget as accelerator clusters scale. A watt used moving data is a watt that cannot be assigned to additional compute. Credo is effectively arguing that optics should be judged not only on cable cost but on the opportunity cost of the power required to drive each link.
Industry discussion has increasingly focused on the transition from pluggable optics to near-packaged and co-packaged designs as AI bandwidth density rises.
The standard matters more than one cable
The biggest long-term implication is interoperability. A proprietary optical system can work technically and still struggle to become infrastructure if emitters, receivers, connectors and modules cannot be sourced across multiple suppliers. The SRMO workstream is intended to prevent that by defining common electrical, optical and mechanical requirements across the supply chain.
That makes this a logical follow-up to BitcoinVersus.Tech’s earlier coverage of Credo’s 1.6T ZeroFlap light cable, Coherent’s 6.4T near-packaged optics, and JEDEC’s first reliability standard for silicon photonics.
The confirmed news is the formation of the OCP workstream and the participation of a broad optical-supply ecosystem. The more ambitious performance figures—60% lower link power, copper-class reliability and a future 30-petabit optical backplane—remain Credo’s engineering targets and demonstrations rather than an industry-wide production baseline.
If the workstream succeeds, the bigger shift may be architectural: optics could move closer to the accelerator and replace copper across distances that historically belonged to electrical links.
BitcoinVersus.Tech Editor’s Note:
We volunteer daily to ensure the credibility of the information on this platform is Verifiably True. If you would like to support to help further secure the integrity of our research initiatives, please donate using the QR code provided
For best QR-code scanning results, please scan this donation code from a desktop or larger screen. Mobile screens may prevent camera recognition.
Leave a Reply