Supplier diligence

Evidence a photonics fab needs before marketing a larger wafer format

Atom Industries proposes four evidence levels a specialty photonics fab's leadership should require before marketing a larger wafer format: PDK release scope, alpha lot acceptance, volume reservation terms, and a named packaging acceptor, each tied to a different buyer commitment.

Technicians in a semiconductor cleanroom

The decision and the direct answer

Before a specialty photonics fab markets a larger wafer-format line, Atom Industries proposes, as a diligence framework rather than a published standard, that its CEO, COO or business development lead require evidence at four levels: PDK release scope, alpha lot acceptance, volume reservation terms, and a named packaging acceptor. Each level supports a different buyer commitment, from a design start to a module quote.

The framework is assembled from supplier-published paths at SMART Photonics, AIM Photonics, PHIX and GlobalFoundries, and the ordering and the mapping of levels to commitments are Atom's proposal, not a finding about any supplier.

Why a PDK port is not a qualification

SMART Photonics, a pure-play photonic integrated circuit manufacturer, publicly frames its customer path as three steps: design with a process design kit, alpha production, and volume production. That vocabulary is useful because it separates what a design kit enables from what a production line has demonstrated. A PDK lets a designer tape out; it does not, by itself, show that wafers of a given size have been run, measured against acceptance criteria, or accepted by a customer. This is an inference from SMART's own published path, not a universal requalification standard: the SMART page is undated and describes only its own process. A buyer or operator should treat a PDK port to a larger format as design enablement, not process qualification, and ask the fab what alpha-level evidence exists for that format before relying on it.

A forward-looking wafer-format claim, read carefully

GlobalFoundries' announcement of its acquisition of Advanced Micro Foundry, datelined November 17, 2025 and published November 18, 2025, shows how wafer-format claims reach the market. GF states it will address long-haul optical communications, computing, LiDAR and sensing on AMF's 200 mm platform in Singapore, with plans to scale to 300 mm as market needs grow. GF also describes AMF as offering manufacturing, prototyping and testing services supported by proprietary technology platforms and in-house process design kits. The 300 mm statement is a plan, and the announcement excerpt provides no qualification data or customer acceptance for it. SMART's page offers a parallel case: the company states it is investing in new 6-inch indium phosphide wafer production capacity and that, when fully activated in 2028, the new production is expected to manufacture more than 12,000 wafers per year. That figure is forward-looking on an undated page. The lesson for a fab's leadership is not about either company specifically; it is that a public statement of a planned format scale-up is not independent proof that the larger format is qualified. A CEO or COO should ask what evidence would sit behind their own equivalent statement before making it.

The commercial reservation boundary

AIM Photonics publishes the agreement sequence that marks when its process becomes commercially reservable. AIM states that purchasing wafer fabrication or test, assembly and packaging services involves disclosure of confidential information, so the first step is an agreement covering that disclosure. Non-members need an executed non-disclosure agreement, then a services agreement setting payment and delivery terms, plus an export control form required under U.S. export control regulations. AIM states that a fully executed services agreement is required before it can reserve space on a multi-project wafer run or a private run, or provide test, assembly and packaging services. This describes AIM's own process, not a universal rule. Atom proposes that a buyer adapt it into a question for any fab: can the fab execute a comparable services agreement for the larger format today, and what delivery commitments, if any, will it state? Because AIM's sequence exists to protect confidential information, detailed acceptance data may require a direct engagement rather than a public review; that possibility is an inference, not something AIM's page states about qualification data.

The module acceptor boundary

PHIX Photonics Assembly, which describes itself as the photonic equivalent of an outsourced semiconductor assembly and test company, states that a photonic integrated circuit by itself is not a functional product: it needs connection to optical fibers, other photonic integrated circuits, cooling solutions and electronics to become part of a photonics-enabled module. PHIX states it builds optoelectronic modules on indium phosphide, silicon photonics, silicon nitride and planar lightwave circuit platforms, with facilities in Enschede, the Netherlands and Bozeman, Montana. The implication, and this is an inference rather than a PHIX conclusion, is that a wafer-format qualification decision that stops at fab process qualification leaves the module acceptor unresolved. Atom proposes a corresponding diligence question: can the fab name an assembly and test provider that has accepted packaged parts from the larger format, and does that acceptor's own process step documentation cover the format in question? Whether a larger wafer changes anything at the package level is precisely what the named acceptor, not the announcement, should be asked to answer.

A proposed evidence pack for the go/no-go

Atom Industries proposes, as a diligence step rather than a published requirement, that a specialty fab's leadership require a four-part evidence pack before marketing a larger wafer format as qualified. First, the PDK release scope, including any stated wafer-size limitations and which process steps the kit covers. Second, alpha lot acceptance criteria and results for the larger format, with the party that accepted the lot identified. Third, volume reservation terms: whether a services agreement can be executed for that format today, and what delivery commitments, if any, the fab will state in writing. Fourth, a named packaging acceptor, meaning an assembly and test provider that has accepted parts from the larger format. SMART's three-step path, AIM's agreement sequence and PHIX's packaging description supply the vocabulary for these levels; the pack itself, and its use as a go/no-go gate, are Atom proposals.

A proposed comparison table

Atom also proposes the following decision aid, which is a proposal, not a finding about any supplier. A PDK port only fits a design start: the buyer can enable a layout but should not expect production commitments. A PDK plus an accepted alpha lot fits prototype work, following SMART's published step structure; Atom proposes extending that position to multi-project wafer participation, since the MPW reservation mechanism is described by AIM, not SMART, so that link is an Atom proposal rather than a supplier statement. A PDK, an alpha lot and an executable volume reservation fit a production quote, following AIM's stated boundary that a services agreement precedes a reserved run. That package plus a named packaging acceptor fits a module quote, since PHIX states a photonic integrated circuit is not functional until packaged. A buyer should ask which position the fab actually occupies before quoting volume by quarter against it.

What would change this approach

Three conditions would change the proposed approach. If a fab publishes nonconfidential alpha or volume acceptance data for the larger format, the evidence pack collapses toward reviewing that data directly. If the buyer's requirement is a substrate or bare die rather than an assembled module, the named packaging acceptor level drops out. If the fab's own acceptance records document the larger format, the review should verify against those records rather than the announcement. One scope note: the supplier examples here are silicon photonics and indium phosphide; none of the cited pages states a GaAs-specific transfer or requalification rule, so a buyer with a GaAs process step should check the fab's own qualification records for that requirement. Atom Industries is development-stage and has no manufacturing capacity to offer; this page is a diligence aid, not a supplier qualification finding. The next useful question for a fab leadership team is which of the four evidence levels it can support today, in writing, for the format it intends to announce.

Sources & further reading

This perspective supports research and relationship development. It does not represent available manufacturing capacity, a committed transaction or an investment offering.

The next step is a conversation

Bring a question worth investigating.

Bring us an opportunity