On October 20, 2026, the Advanced Photonics Coalition (APC) announced a portfolio of five specifications and recommended-practice documents aimed at moving optical networking hardware from bespoke engineering toward high-volume manufacturing. The release, datelined from Sunnyvale, California, is the coalition’s first set of published standards documents. APC describes the coalition as three years old and says the portfolio was developed through collaboration of 60 member organizations.

For the photonics industry, the significance is less in any single document than in the fact of publication itself. APC is a manufacturing-oriented coalition — its stated purpose is improving business conditions for companies developing optical interconnects and subassemblies for data center, high-performance computing, and AI networking equipment. A coalition that publishes manufacturing, interface, and test specifications is declaring that the industry’s bottleneck has moved from the physics to the factory floor.

The timing matters. The announcement landed at the PECC Summit, in the same week the trade press was digesting Lumentum’s disclosure to Bloomberg that its optical components are “completely sold out” through early 2029. When the optical components for AI networking are supply-constrained on a multi-year horizon, common manufacturing specifications stop being paperwork and start being capacity planning.

What the five documents cover

The portfolio consists of five documents, as listed in APC’s announcement:

  • System Level Architecture Overview for Co-Packaged Optics — a reference framework for fitting optical engines, lasers, fiber connections, and switching into package and system designs.
  • Co-Packaged Optics Informative Outlook and Recommended Practices — guidance on integration, operations, serviceability, and design considerations that include support for optical circuit switching.
  • Silicon Photonics for Co-Packaged Optics Recommended Test Practices — recommended approaches to testing optical and electrical performance during successive manufacturing stages.
  • 4-Core Multicore Fiber Cable, Connector and Fiber Array Unit Specification — common expectations for four-core fiber, its connectors, and the optical attachment interface.
  • Silicon Photonics Manufacturing Specification — shared terminology and interfaces for the manufacturing and assembly of photonic integrated circuits.

Note the taxonomy. These documents are a mix of specifications, informative guidance, and reference architectures — they do not all carry the same compliance status. That distinction matters for anyone reading the announcement as a commitment to interoperability: buyers should examine the normative provisions and conformance methods of each document individually rather than treating the entire portfolio as a certification regime. No certification program or conformance-testing apparatus has been announced alongside the documents.

The documents reflect the coalition’s working structure: three working groups — Co-Packaged Optics, Silicon Photonics Manufacturing, and Multicore Fiber — with dedicated focus groups on system-level architecture, CPO reference design, high-speed optical interfaces, design and simulation, and test and measurement.

The rationale: variations that add no market value

APC frames the portfolio as an attack on “implementation variations that add no market value” and “supply chain silos that constrain volume.” That language is worth unpacking, because it is the honest theory of the case for manufacturing standards.

In a nascent manufacturing ecosystem, every participant invents their own interface conventions, test flows, and terminology. Some variation is genuine innovation; much of it is accidental complexity that prevents second-sourcing and forces every buyer to re-qualify each supplier’s idiosyncratic implementation. Standards convert the accidental complexity into shared infrastructure, leaving companies to compete on the things that actually differentiate their products.

“By establishing common manufacturing, interface, and test practices, APC aims to remove implementation variations that add no market value and overcome supply chain silos that constrain volume,” the coalition said. “APC standardization supports multiple sourcing and broader supplier participation, helping the industry build an open and scalable photonics ecosystem.”

“APC advances its purpose by publishing its first specifications and best practice documents,” said Jeffery Maki, chairperson of APC’s board of directors. “Our focus on manufacturing promotes and enables the common business interest of companies developing AI networking equipment.”

The open-ecosystem argument is made explicitly in the CPO document itself. As quoted in The Next Platform’s coverage of the announcement, the document states: “Without open ecosystem driven standards, the natural gravitational pull is toward vertically integrated proprietary solutions, an outcome that would fragment the market, increase customer lock-in risk, slow adoption, and ultimately diminish the opportunities of innovation for CPO’s technical benefits.” That is a candid acknowledgment of what happens to standards efforts when they fail: the market consolidates around whoever can ship vertically integrated boxes, and the open ecosystem that startups need to participate never forms.

Why now: the supply crunch and the copper wall

The context that makes this announcement feel urgent rather than bureaucratic is the demand picture. Earlier in October, Lumentum Holdings chief executive Michael Hurlston told Bloomberg in Tokyo that the company’s optical components are “completely sold out” through early 2029 — a sharp extension from the company’s April expectation of selling out through 2028. According to the Bloomberg reporting, Lumentum is unable to meet about 70 percent of demand for some products through next year, and about 30 percent for other products through 2028. Lumentum supplies advanced indium phosphide devices — EML chips, pump lasers — used in high-speed optical interconnects, and received a $2 billion investment from Nvidia earlier this year tied to multibillion-dollar purchase commitments.

Hurlston told Bloomberg that customers were already beginning to secure orders for 2030. That is not a market in which incremental demand growth can be absorbed by running factories harder; it is a market in which manufacturing capacity itself is the scarce input.

Meanwhile, the architectural case for optics keeps strengthening. Datacenter operators are confronting the limits of copper: as signaling rates rise, copper’s reach shrinks while its power and heat grow — the industry shorthand is the “copper wall.” Earlier this month, a McKinsey report flagged networking as an area requiring improvement as AI shifts toward inference workloads, advising operators to “deploy optics-integrated switches (CPO) for higher bandwidth density and lower power per bit” and warning that “a specialized architecture only earns its cost at volume.” And at GTC this week, Nvidia founder and CEO Jensen Huang demonstrated the company’s Spectrum-X co-packaged optics technology, pushing the optics-as-close-to-the-switch-ASIC-as-possible architecture from roadmap to demonstration.

Shortages of the sort Lumentum describes, and the surging prices they imply, are exactly the conditions that make common manufacturing specifications valuable — if they can unlock multiple sourcing and broader supplier participation, as APC claims. That is the causal chain the portfolio is betting on: standards lower the barriers to entry for suppliers, more suppliers mean more capacity, and more capacity relieves the bottleneck. It is a reasonable theory. It is also, at this stage, only a theory: the announcement provides no audited cost-per-bit savings, production yields, or commercial deployment totals attributable to the specifications.

The optical circuit switching thread

One of the more interesting elements of the portfolio is its treatment of optical circuit switching (OCS). APC says it is taking next steps toward including support for different forms of optical circuit switches, “including reliability and configuration requirements for rack-level OCS deployments.” The CPO Informative Outlook document is more specific. On nanosecond-level OCS, it states: “The objective is not to standardize a particular fast OCS implementation, but to define the control-plane, optical-interface and behavioral framework that allows fast, transparent OCS to operate as a first-class component of an open AI scale-up fabric.”

That framing — define the framework, not the implementation — is exactly how a standards body should handle an immature technology: it creates the conditions for competition among implementations rather than picking a winner.

The announcement drew an immediate response from one of the OCS players. Resolight, a photonics developer that APC counts on its working group, described the CPO specifications as addressing “the data-plane integration needed to make native photon routing a practical alternative to electronic data switching,” in a statement quoted by The Next Platform. Resolight’s CEO, Ofer Shapiro, offered the week’s most quotable line: “The easiest to source and the greenest transceiver is the one that you don’t need.”

The logic behind the line is straightforward. Resolight’s approach — nanosecond optical circuit switching that routes coordinated data transfers while keeping payloads in the optical domain — reduces repeated optical-electrical-optical conversions, which the company says could improve energy efficiency and cut transceiver counts in multi-rack AI scale-up networks. Where indium phosphide is supply-constrained, Shapiro’s addendum was that the sensible response is to “need less of it rather than figure out where do we get more.”

That is a vendor’s claim, and it should be read as one — Resolight is selling an architecture, and its interest in an open control-plane framework for fast OCS is transparent. But the underlying point generalizes: if the specification portfolio succeeds in making optical switching a first-class component of AI fabrics, the companies that benefit most are the ones whose architectures were waiting for exactly that framework.

What this signals for manufacturing

For readers of this publication, the signal worth extracting from the APC portfolio is what it says about where photonics manufacturing is in its maturity curve. Consider the sequence: a coalition forms around manufacturing problems; three years later it publishes its first specifications and recommended practices; the documents cover system architecture, test flows, fiber interfaces, and manufacturing terminology. That is the sequence of an industry moving from “can we build this” to “can we build this the same way twice, from two different suppliers, at volume.”

The test-practices document is arguably the most consequential of the five, precisely because it is the least glamorous. Photonic integrated circuits fail in ways that are expensive to discover late — poor wafer uniformity, bad fiber attach, unreliable lasers — and test practices that catch those failures at each manufacturing stage rather than at final assembly are what make high-volume production economically viable. Common test practices also make it possible for a new supplier — say, one integrating an excitonic modulator or detector onto a silicon photonics platform — to enter the supply chain without every buyer inventing its own qualification flow. The manufacturing playbook is what new materials have to be inserted into, and the playbook is now being written down.

The multicore fiber specification points in the same direction: four-core fiber increases spatial channel density, but only if connectors, fiber array units, and the optical attachment interface are standardized enough that cable vendors, transceiver makers, and system integrators can all build to the same mechanical and optical expectations. A specification for the fiber without specifications for the connectors and attach interface would leave the ecosystem fragmented; the APC document covers the full attachment stack.

None of this resolves the near-term supply crunch. Standards do not add capacity; they lower the cost of adding capacity and make it safer for new entrants to add it. The Lumentum sold-out horizon extends to early 2029 precisely because adding optical component capacity takes years. The APC portfolio is a bet that the capacity added over that window will land in an open, multi-sourced ecosystem rather than in proprietary silos.

What to watch

Specifications are promises about the future, and the right way to evaluate them is to watch what follows. Several things would indicate the portfolio is having its intended effect:

  • Adoption by vendors: Do transceiver, switch, and fiber vendors reference the APC documents in product specifications and datasheets? Do system vendors require conformance in procurement?
  • Conformance machinery: Does APC or a partner organization stand up conformance test methods, reference implementations, or a certification program? Recommended practices without conformance testing are aspirations; with it, they become procurement requirements.
  • Reference designs: The coalition’s focus groups include CPO reference design development. Published reference designs that build on the architecture overview would turn the framework into something engineers can copy.
  • New entrants: Do new suppliers — particularly in silicon photonics foundry services, packaging, and test — enter the AI networking supply chain citing the APC documents as reducing their qualification burden?
  • Measured outcomes: Do volume cost reductions or yield improvements get reported and attributed to common practices? The announcement’s objective is “high-volume, lower-cost manufacturing,” but the portfolio’s success will be measured in the data the industry has not yet produced.

There is also the negative case to watch for: specifications that sit on a shelf while the market consolidates around vertically integrated solutions anyway. The CPO document itself warns of that outcome. Standards efforts fail when dominant players have no incentive to interoperate — and in a supply-constrained market, incumbents are often comfortable with silos.

For now, though, the industry has something it did not have three months ago: a published manufacturing portfolio from a 60-company coalition, covering the full path from silicon photonics fabrication to fiber attach to rack-level switching.

Sources and further reading