The race to build out AI data centers has entered a phase where companies are scrambling for indium phosphide (InP) substrates, a key material for optical transceivers. Taiwan's Economic Daily News reported on August 17, 2026 that substrate makers are considering additional price hikes of more than 10% for the October-December quarter. Meanwhile, on the supply side, a series of long-term contracts have been signed to secure production capacity years in advance. Even if prices rise, laser epiwafers cannot be manufactured without access to substrates.
This tightness brings a surge in demand for VPEC, LandMark Optoelectronics, and IntelliEPI, but it also imposes limits on procurement costs and production capacity. Whether these companies can turn this tailwind into sales and profit growth depends on three conditions: securing InP substrates, passing costs on to customers, and achieving mass-production yields.
A Fourth Round of InP Substrate Price Hikes in Sight
According to Economic Daily News, InP substrate prices have already risen three times since the October-December 2025 quarter, and a fourth increase is now expected. While previous hikes ranged from 3-5%, the next increase is expected to exceed 10%. Epiwafers, which have a semiconductor layer grown on top of the substrate, have also seen two price increases, with a third under consideration for the October-December quarter. However, these figures are not official price lists from suppliers but numbers the newspaper attributed to industry sources.
The direction of these price increases is also confirmed by primary sources. In June 2026, JX Metals announced it would raise prices for customers alongside its plans to expand InP substrate production. In July, AXT and Lumentum signed a supply agreement running through the end of 2031, under which Lumentum plans to prepay a total of $87 million to reserve production capacity. Rather than fixing contract prices, the money is being directed toward securing the right to obtain the necessary volumes.
Policy factors restricting supply have also come into play. On February 4, 2025, China's Ministry of Commerce and General Administration of Customs added InP and two organic indium raw materials to the list of items subject to export licensing. The affected materials include trimethylindium and triethylindium. According to the 2026 edition of statistics from the U.S. Geological Survey (USGS), global refined indium production in 2025 totaled 1,100 tons, of which China accounted for 760 tons, or roughly 69%. While this figure does not represent InP substrate market share, it does indicate that the upstream supply of raw materials is concentrated in a single country.
Reuters reported in June 2026 that following the introduction of the export controls, the average price of 6-inch InP wafers rose 250% to reach $5,000. Since this is not a public price index with standardized specifications or contract terms, it cannot be mechanically compared with the observed 10%+ increase expected for the October-December quarter. Even so, the explanation that delays in export licensing and AI-driven demand are having a simultaneous effect aligns with JX Metals' price revisions and the trend toward long-term contracts.
Even With Silicon Photonics, InP Light Sources Remain Essential
The InP shortage stems not only from the increasing volume of data moving between AI servers, but also from the shift toward high-speed optical connections as the transmission method of choice. InP can efficiently emit light in the 1310-1550 nanometer band, making it useful in lasers that convert electrical signals into light and in photodetectors that convert light back into electrical signals. In 800G and 1.6T optical transceivers, it serves as a core material in electro-absorption modulated lasers (EML) and continuous-wave (CW) lasers.
The spread of silicon photonics does not necessarily mean InP demand will disappear. While silicon is well-suited for integrating optical circuits, InP-based materials are used as light sources that efficiently generate light in the wavelength bands used for communication. The co-packaged optics (CPO) prototype that Coherent unveiled in March 2026 also combines a silicon photonics circuit with a high-power InP CW laser. As the approach shifts, InP's role moves from the entire optical circuit to serving as an external light source.
Sumitomo Electric has also explained that whether EML is used at 1.6T depends on the number of wavelengths and each manufacturer's design, with some companies choosing EML and others opting for CW lasers. While InP usage does not necessarily increase at the same rate as speeds rise, the coexistence of multiple approaches means that supply capacity must be developed to support each of them.
Increased Production Plans Will Take Time to Translate Into Greater Supply
Substrate makers are moving to expand production on an unprecedented scale. JX Metals plans to invest up to 120 billion yen over the next four years, bringing its total investment—combined with previously announced amounts—to approximately 150 billion yen. In addition to its Isohara plant in Ibaraki Prefecture, the company plans to establish production capacity in the Hitachinaka area as well, aiming to increase InP substrate production capacity by 7 to 10 times compared to fiscal 2025. However, detailed dates for equipment installation and the start of mass production have not yet been disclosed.
Coherent has signed a letter of intent to expand its 6-inch InP plant in Sherman, Texas, and receive up to $50 million in CHIPS Act support from the U.S. Department of Commerce. In investor materials from August 2026, the company outlined plans to double its in-house InP output by year-end and more than double it again from that level by 2027. AXT also recorded its highest-ever InP sales in the April-June quarter, with quarterly revenue rising from $18 million in the same period a year earlier to $47.6 million.
Even so, it seems unlikely that supply will quickly become abundant. Lumentum's new U.S. plant, announced in March 2026, will use 6-inch InP, but mass production is not scheduled to begin until mid-2028. Each stage of the process—single-crystal substrate production, epitaxial growth, and device fabrication—requires improving yields and passing customer qualification after equipment installation. The scale of already-announced investments itself suggests that the supply shortage will not be resolved easily. Announcements of capital investment do not directly increase supply volumes for the October-December 2026 quarter.
The fact that major customers are locking in production capacity through 2031 reflects this anticipated time lag. What will ease near-term supply-demand conditions is not the announcement of new plant plans, but rather the utilization rates of existing lines, increased yield per wafer through the shift to 6-inch substrates, and the speed at which export licenses are processed.
Securing Substrates and Passing On Price Increases Will Determine the Fate of Taiwan's Three Companies
LandMark Optoelectronics is already beginning to convert demand into business performance. The company's revenue for the first half of 2026 reached NT$2.126 billion, up 111% year-over-year, with a gross margin of 56%. According to its annual report, the company is mass-producing EML, CW lasers, and photodetector epiwafers for 800G and 1.6T applications. In April 2026, it signed a long-term InP substrate supply agreement with Sumitomo Electric running through 2030, and increased its capital expenditure to NT$1.315 billion. Because both revenue growth and material procurement can be confirmed simultaneously, LandMark offers the clearest picture among the three companies of the benefits from supply-demand tightness.
IntelliEPI's disclosures reveal a reality in which rising sales are accompanied by growing cost pressures. The company's InP epiwafer sales grew 97.23% year-over-year in 2025. Meanwhile, purchases from Sumitomo Semiconductor totaled NT$121.234 million, accounting for 45.27% of total purchases. Overall production-related purchases rose 67.38% year-over-year, outpacing the 50.49% growth rate in sales. The company has also stated that it built up inventory in response to rising III-V material prices. Even with strong demand, if rising costs cannot be passed on through sales prices, profit margins will shrink.
VPEC handles InP HBT, InGaAs-based photodetectors, and DFB lasers, among others, using MOCVD, and its annual report indicates 6-inch mass-production capacity along with plans to increase InP product output. However, it has not disclosed its latest InP substrate procurement volumes, the scope of its long-term contracts, or its rate of price pass-through. It is necessary to distinguish between the strength of order intake and procurement constraints.
Lumping these three companies together simply as "InP-related stocks" risks misjudging the impact of higher substrate costs. The figures that should be checked, beyond revenue growth rates, include raw material inventory, purchase amounts, gross margins, and the volumes secured through long-term contracts. If the reported price increase of more than 10% is implemented in the October-December 2026 quarter, each company's ability to pass on costs will become apparent starting with its early-2027 earnings results. Furthermore, in 2028, whether JX Metals, Coherent, and Lumentum's production expansions proceed as planned will be a key factor shaping the direction of supply, demand, and pricing.