Optical-module buyers face a concentrated substrate supply chain, rising epitaxy requirements, and capacity additions ranging from 2-inch to 6-inch wafers, with JX planning up to ¥120 billion of additional InP investment over four years [S2][S3][S6].
Why InP remains essential for optical transceivers
Indium phosphide is a direct-bandgap compound semiconductor used to generate and detect infrared light at telecommunications wavelengths, while silicon is comparatively ineffective as a light source [S1]. InP-based components include DFB and EML lasers, PIN and APD photodetectors, semiconductor optical amplifiers, and continuous-wave laser sources [S2][S4].
Silicon photonics changes the packaging architecture, not the need for InP. Silicon can route, split, and modulate light, but many systems still require an external InP laser or bond InP laser dies to a silicon photonic chip [S1][S5]. For procurement teams, the relevant material flow therefore includes high-purity indium, crystal growth, substrate processing, epitaxy, die production, and optical-module assembly rather than finished transceivers alone [S2].
Lane counts make transceiver architecture a demand multiplier
ETU-LINK separately characterizes 400G long-distance models as having relatively low InP consumption, 800G modules as broadly using EML chips, and 1.6T or 3.2T modules as carrying more InP chips per module, with substrate and epitaxial capacity becoming major constraints [S4].
That architecture links directly to EML laser supply constraints for 2026 transceivers, because EML construction combines a DFB laser with an InP electro-absorption modulator.
Substrate specifications narrow qualification choices

Vital Materials offers VGF-grown semi-insulating, n-type, undoped n-type, and p-type InP substrates. Available diameters span 2, 3, 4, and 6 inches, with 6-inch options listed for n-type material, while semi-insulating, undoped n-type, and p-type products are listed from 2 to 4 inches [S2].
Semi-insulating InP uses iron doping, provides resistivity of at least 0.5 × 10⁷ Ω·cm, and lists etch pit density no higher than 5,000/cm² [S2]. N-type material uses sulfur or tin doping, has a listed carrier concentration of 0.8 to 8 × 10¹⁸ cm⁻³, and also specifies etch pit density no higher than 5,000/cm² [S2].
Undoped n-type InP is offered from 2 to 4 inches with mobility of 3,000 to 5,000 cm²/V·s and etch pit density no higher than 5,000/cm² [S2]. P-type InP uses zinc doping and is available in 2-, 3-, and 4-inch diameters [S2].
Qualification should therefore test carrier concentration, resistivity, mobility, etch pit density, total thickness variation, warp, surface condition, thickness, and orientation, rather than treating diameter as the only acceptance criterion [S2]. Inspection coverage across those parameters is more useful for comparing wafers than a generic “high quality” designation.
Supply concentration and capacity expansion
The InP substrate market has been described as a two-vendor duopoly involving AXTI and Sumitomo Electric, which makes qualification of an additional source strategically important [S6]. JX Advanced Metals announced up to ¥120 billion of further investment over four years on 17 June 2026, following capacity announcements in July 2025, October 2025, and February 2026 [S3].
JX plans to invest up to ¥120 billion over the next four years to expand its InP substrate production capacity, with the expansion described as being in addition to the Isohara Works investment [S3]. The announcement states that JX is engaging in ongoing dialog with customers, which have requested increased InP substrate supply for their capacity-expansion plans [S3].
New capacity is not equivalent to qualified optical-module supply. Substrates must meet device-maker specifications, support epitaxial growth, and deliver consistent defect and dimensional control before they can contribute to lasers or detectors [S2]. This makes capacity expansion trackable at three levels: wafer output, customer-qualified material, and committed transceiver shipments.
Silicon photonics reduces some consumption but not all demand

Two principal PIC approaches are identified as silicon photonic and monolithic InP [S5]. Monolithic InP combines the required active optical functions within one material platform, while silicon photonics relies on integration and may require an InP light source or bonded laser die [S1][S5].
The procurement consequence is that a shift toward silicon photonics does not automatically remove InP from the bill of materials [S1]. ETU-LINK states that a silicon-photonics approach combined with thin-film lithium niobate addresses modulation while still relying on InP light sources [S4].
Design choice can also change the risk profile. A module using an external InP laser, an EML, a monolithic InP PIC, or a silicon-photonic PIC with bonded InP will have different substrate requirements and qualification dependencies. Optical-module suppliers should identify which active components are present rather than classifying a product only as 800G or 1.6T.
Procurement priorities for constrained InP supply
The clearest purchasing signal is the combination of eight lanes in 1.6T modules, near-doubling of InP use per module, a concentrated substrate supplier base, and multiple expansion programs [S1][S3][S6]. ETU-LINK also reports orders backlogged through 2027 and a claimed production gap for InP exceeding that of DRAM and NAND [S4][S8].
Buyers should separate three risks: physical wafer supply, epitaxy and device conversion capacity, and module qualification. A 6-inch n-type substrate may improve wafer economics for a given device design, but diameter alone does not remove the need for defect-density, thickness, warp, and orientation control [S2].
For allocation, require suppliers to state wafer diameter, growth method, dopant, electrical target, etch pit density, thickness and orientation, inspection scope, qualification status, and committed volume. Dual-source qualification is a practical response to the two-vendor structure, while long-term volume agreements need delivery milestones tied to qualified material rather than unverified nameplate capacity [S3][S6].
Track JX capacity at the Hitachinaka site, additional substrate suppliers completing device-maker qualification, and the ratio of EML-equipped 1.6T modules in procurement. Those signals reveal whether InP substrate capacity is converting into qualified optical components rather than remaining at the wafer-production stage [S1][S3].