2026-10-11 17:12 UTC

Further disclosures will determine whether Coherent’s use of all available indium-phosphide laser output for internal demand materially constrains outside AI data-center optical-networking supply.

state: expiredheat: lowuncertainty: highnovelscott: highai-infrastructure optical-interconnects supply-chainCoherent

What is this?

Coherent Corp. (NYSE: COHR) is a major photonics manufacturer that produces indium phosphide (InP) lasers critical for AI data center optical interconnects (transceivers at 800-Gig and 1.6-Terabit speeds). During an earnings call for FY2026 Q1, a Coherent executive stated the company does not sell its InP output on the open market because all capacity is consumed internally for its own products. This internal-capture policy, combined with tight supply of refined indium (China controls ~70% of production, with tightened export controls since 2025) and concentrated manufacturing, has created a supply-demand gap for InP substrates and lasers that industry observers (Lumentum, TrendForce) describe as a critical bottleneck for scaling AI data center networking. The web snippets confirm the factual assertion (internal allocation) and contextualize it as part of a broader InP shortage, but they do not independently verify the exact quote attributed to 'Anderson' or whether Coherent later adjusted the policy.

Why it matters to Scott

Coherent's internal capture of all indium phosphide laser output creates a verified bottleneck in optical interconnects, a layer Scott has identified as a potential limit on AI scaling beyond compute or memory. This is not merely an example of a known pattern — it is a concrete, documented supply constraint at a specific company that directly tests whether networking interconnect supply (which Scott has written about as a possible scaling ceiling) becomes the binding variable for frontier AI data center buildout. The case opens a dated-receipts opportunity: if Coherent's internal allocation materially constrains outside supply while Scott has argued that interconnect bottlenecks might limit scaling, subsequent disclosures will either confirm or falsify that claim.
dev:concept.boundary-case-compressiondev:concept.evidence-inference-falsifier-reportingwork:concept.ai-consulting-practiceradar:concept.ai-infrastructureradar:concept.ai-hardwareradar:lumilens-700m-optical-interconnect-financing
queries asked of Scott's wikis
  • Scott's framework on AI infrastructure commodity bottlenecks — when vertical integration by a supplier constrains an entire layer
  • Scott's analysis of supply-chain concentration risk in AI hardware (e.g., indium, gallium, silicon carbide)
  • Scott's position on optical interconnects vs. electrical alternatives for AI data-center scale-up
  • Scott's writings on whether frontier model scaling is limited by networking interconnect supply (vs. compute or memory)
  • Scott's takes on China export controls on specialty metals and their impact on AI buildout timelines
  • How Scott distinguishes between real capacity constraints and strategic vertical-integration narratives from component vendors

Measured heat

no measured readings yet — the hourly heat pass fills this in

How the heat travelled

no chain yet — the hourly chain pass fills this in

Evidence (2) — ⭐ canonical anchor

sourceobjectauthorscorecomments
🟠 redditOne of the few companies making the lasers that move data inside AI data centers just said it's keeping all of them for itself
singularity
Novel-Lifeguard649112612
🟧 echo.other ⭐During Coherent’s FY2026 Q1 earnings call, Anderson said: “we do not sell indium phosphide in the open market” because “all of our capacity Coherent Corp.; statement by CEO James Anderson——

Interpretation history

Decision trace