Photonics capacity diligence
Testing a customer volume forecast before releasing a photonics capacity tranche
Convert a customer volume forecast into qualification-gated capacity tranches, with the demand evidence, accepting party, deliverable and reservation trigger required before each release, using SMART Photonics, PHIX and AIM Photonics statements as worked examples.

The decision this article supports
Before committing capacity, treat the customer's volume forecast as a hypothesis and release capacity only in tranches, each tied to a qualification milestone rather than to the forecast number itself. For each tranche, require named demand evidence, an identified accepting party and exact deliverable, and a reservation trigger you can verify. The evidence table below is an Atom Industries proposal, assembled from three public supplier pages; it has not been validated with an operator or a customer, and no source here describes forecast testing directly.
What the supplier pages actually establish
Three public documents frame the problem. SMART Photonics' Start Production page states that the company supports customers from proof of concept to high volume production and presents a three-step path: design with a process design kit, alpha production, and volume production. The same page states that SMART Photonics 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, corresponding to millions of photonic chips. That is a forward-looking company statement about expected annual output. Atom Industries' inference from it is that the figure is a supply-side expectation to be tested, not evidence of capacity allocated to a particular customer, at a particular qualification stage, in any particular quarter; an operator who reads it as customer-specific availability would be treating a supplier announcement as a customer commitment.
PHIX's About Us page states that PHIX offers assembly services and contract manufacturing for photonic integrated circuits and MEMS, builds optoelectronic modules on indium phosphide, silicon photonics, silicon nitride and planar lightwave circuit platforms, and describes itself as the photonic equivalent of an outsourced semiconductor assembly and test company. PHIX states that the design and manufacturing of the package connecting a PIC to optical fibers, other PICs, cooling solutions and electronics is its core business, and that it can produce working prototypes from the first chips and then optimize PIC and module design, process and equipment to scale module production to high volumes. The page also describes facilities in Enschede, the Netherlands, and Bozeman, Montana. Any utilization figures, customer commitments or qualification results would have to be requested from PHIX or its customers directly, since the page text used here is a services and company description.
AIM Photonics' MPW and TAP Agreements page 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. The page states that a fully executed services agreement is required before AIM Photonics can reserve space on a multi-project wafer run or a private run, or provide test, assembly and packaging services, and that an export control form is required to meet U.S. export control regulations. All three pages are undated live pages with no reported document version, so no publication date should be assigned to them.
The forecast-to-capacity evidence table
The following table is an Atom Industries proposal, not a supplier document. For each capacity tranche an operator considers releasing, record one row with these columns: customer demand evidence; accepting party and exact deliverable; qualification milestone; service boundary; reservation trigger; and the capacity tranche itself.
Demand evidence should be specific to the tranche. A forecast stated as volume by quarter is stronger than an annual number, and a signed purchase commitment is stronger than a forecast; none of the three pages used here states customer commitments of this kind, so the operator must request them directly from the customer. For the accepting party and deliverable, require the customer to state whether the purchased item is a substrate, a fabricated die, an assembled module or a test service, and who accepts it at handoff. Atom Industries uses this same framing in its own discovery work: a useful first discussion starts with the item being purchased and the party who accepts it.
The qualification milestone column should name the technical event that must occur before the tranche releases, such as a qualified process step or an accepted prototype build. The service boundary column should state which stage of the supply chain the operator's own scope covers, so a packaging forecast is not silently converted into a fabrication commitment. The reservation trigger column should record the administrative gate: AIM Photonics' page shows what one looks like, since a fully executed services agreement and export control form precede any run-slot reservation there. Finally, mark each cell as customer-supplied, operator-verified, or still missing; a tranche with missing milestone evidence stays unreleased.
Separating development builds from volume service
PHIX's language, producing working prototypes from the first chips and then scaling to high volumes, suggests that development-build output and volume output may need separate evidence before a forecast becomes actionable. This is an Atom Industries inference from service-stage language on the PHIX page, not a statement that PHIX separates those stages commercially. The practical implication for the operator: a customer pointing to successful prototype builds has evidence of technical feasibility, not evidence of volume demand, and the two should occupy different rows in the table with different release conditions.
Administrative readiness is not technical qualification
AIM Photonics' sequence, non-disclosure agreement, then services agreement and export control form before any reservation, indicates that a run slot is gated by executed commercial and export-control documents. An operator testing a customer forecast should keep this administrative readiness separate from technical qualification: a customer that has signed agreements has cleared paperwork, not demonstrated that its part passes the operator's process at the claimed volume. Atom Industries' inference from the AIM Photonics page is that the operator must define the technical qualification milestones itself and write them into the services agreement, since the page documents the agreement sequence as the gate it describes.
Conditions that would change this approach
The approach would change if the customer provides different evidence. A nonconfidential, customer-identified volume commitment tied to acceptance criteria would let the operator collapse tranches; a forecast that shifts materially between quarters would argue for smaller tranches with shorter review intervals. If the operator's own technical diligence shows the process step is not yet qualified, no customer evidence justifies releasing volume capacity, and the first tranche becomes a qualification tranche. If the deliverable in question is a test service rather than a physical part, the milestone column should be redefined around measurement correlation rather than process yield. Atom Industries is itself development-stage: it has no operating facility, qualified process, signed customer or available production capacity, and this article is commercial discovery, not a statement of operating capability.
The next question to ask the customer
Ask the customer directly: which qualification milestone, if missed, would cause you to reduce or defer the forecast volume? The answer reveals which milestone the customer itself treats as load-bearing, and it converts the forecast from a single number into a map of the commitments that actually gate demand. If the customer cannot name one, that is itself evidence about how firmly the forecast should be weighted when the operator decides which tranche, if any, to release. Where an additional source is needed, the operator should ask the customer for the acceptance records and run documentation that sit behind the forecast, not for confidential process data.
Sources & further reading
This perspective supports research and relationship development. It does not represent available manufacturing capacity, a committed transaction or an investment offering.