On June 2, Jensen Huang declared at Computex, "Use light wherever you can!"
A dozen or so days later, Boss Huang boarded a plane to Texas. After getting off, he picked up a shovel, went to the site of Coherent's local expansion project, and turned the first shovel of soil to break ground for the project.
Shortly before that, Jensen Huang and Coherent's CEO had joined Trump on a visit to China, asking China to relax export scrutiny.
Two seemingly unrelated events pointed to the same thing: indium phosphide.
In the AI compute race, GPU compute is no longer the only bottleneck. High-speed interconnection between GPUs is the key constraint on data center performance. The industry generally believes copper wire has reached its physical limit, and optical interconnect is the way forward. That is why Jensen Huang shouted that phrase.
Thus, as an indispensable substrate material for 800G, 1.6T, and even next-generation CPO optical modules, indium phosphide has long escaped the category of a niche material and become a strategic cornerstone of high-speed optical interconnection.
In the past, people focused more on GPU compute competition; few noticed that this thin semiconductor wafer had quietly become the most strained link in the AI compute transmission chain.
According to Yole estimates, the global indium phosphide substrate shortfall will exceed 70% in 2026, with expansion cycles lasting more than two years.
Capacity is in critical shortage!
So the United States, Japan, Europe, and China are all trying every means to save the situation, and in the process of addressing their weaknesses, their respective predicaments have become crystal clear.
01 United States:
Lack of Capacity
On the indium phosphide track, America's greatest confidence comes from its huge downstream AI compute market. Giants such as Nvidia entered early, spending heavily to lock up substrate capacity and trying to control supply chain leverage from the demand side. But no matter how abundant capital is, it is hard to immediately create mature domestic manufacturing capability.
Even if several companies have launched expansion plans for 6-inch indium phosphide production lines, crystal growth, process debugging, and yield ramp-up all have objective time thresholds. A completely new production line often takes more than two years from construction to stable supply. During those two years, the shortfall will be hard to fill; distant water cannot quench present thirst.
In addition to the manufacturing cycle, raw materials are another hurdle. The United States has almost no high-purity indium mineral resources domestically, so the most basic raw material for producing indium phosphide must always be purchased from abroad. Even if factories are built domestically, control over upstream resources still lies elsewhere, and supply chain risks follow like a shadow.
02 Japan:
Lack of Raw Materials and Market
For a long time, Japan was the "big boss" in the world of indium phosphide substrates. Two major manufacturers, Sumitomo Electric and JX Metals, once held more than 70% of the global market share. Decades of expertise gave Japan hard power in 6-inch substrate yield and crystal growth stability that others find very difficult to match.
But Japan has a fatal weakness: there are almost no indium mines domestically. Most of the high-purity indium raw material Japan needs to produce indium phosphide depends on overseas imports. When upstream raw material supply fluctuates, even the most advanced production lines are instantly put on the back foot.
The impact of tightening raw materials has already begun to show. Many factories have been forced to cut capacity, customer delivery cycles have been repeatedly extended, and some production lines have even fallen into semi-idle predicaments. With top-notch technology but no "rice" to put in the pot, even the best processes have to stop.
At the same time, Japan does not have a sufficiently large downstream end market, so large quantities of finished indium phosphide products and optical chips need to be sold overseas. Once the external market environment changes, corporate revenue comes under pressure.
03 Europe:
Racing for the Route, but Actually Lacking Everything
Europe's approach is the most different-it is not racing for capacity, but for the route. That is, relying on its advantage in silicon photonics processes, it integrates indium phosphide as an external light source into CPO.
Many people think silicon photonics will replace indium phosphide, but the real path is exactly the opposite: silicon itself does not emit light, and silicon photonics/CPO must rely on external CW lasers made of indium phosphide as the light source. CPO switches have more optical engines, which actually significantly increases total indium phosphide usage. A single Nvidia Spectrum-X CPO switch carries 32 silicon photonics engines, more than 10 times the usage of traditional equipment.
Therefore, what Europe is competing for is the standard and ecological niche of the "silicon photonics–indium phosphide heterogeneous integration" route.
However, Europe has almost no mature large-scale indium phosphide substrate manufacturers domestically, and its own substrate supply is highly dependent on purchases from Japanese and American companies. Even though epitaxy companies such as IQE actively sign long-term supply agreements, control over upstream substrate supply still rests in others' hands, and supply chain stability is not under its own control.
In addition, like Japan, Europe lacks a huge downstream optical module end market, and its industrial closed loop is incomplete. It can export process solutions and participate in standard-setting, but without an end market to feed back into upstream manufacturing, it is very difficult to reverse-drive domestic substrate capacity building.
Emphasizing routes while neglecting capacity has left Europe in the awkward position of "having solutions but lacking materials."
04 China:
?
In the indium phosphide race, China has two uniquely advantageous chips: indium resources and the market.
China's indium reserves are prominently advantaged. According to the China Geological Survey's Global Mineral Resource Reserve Assessment Report 2024, as of the end of 2022, global identified indium resources stood at 43,689 tonnes, with China accounting for 17,538 tonnes, or 40.14% of the global total.
The downstream optical module market is China's strongest link. In the Top 10 optical module supplier list released by optical communications research firm LightCounting, seven Chinese companies are included.
Resources and market have laid an excellent foundation for China's indium phosphide industry to catch up, but resource advantages do not automatically translate into competitiveness in high-end finished products, and multiple barriers are revealing predicaments.
The first is the step from resources to "usable raw materials."
Despite huge resource advantages, sufficient upstream refined indium, and breakthroughs in 7N-grade and above high-purity indium technology, production capacity is still far from enough.
High-purity red phosphorus is even worse. High-purity red phosphorus is the only core phosphorus source for indium phosphide single-crystal growth. Every 1 kilogram of indium phosphide single crystal produced consumes 0.25 to 0.3 kilograms of high-purity red phosphorus. Its purity standards are extremely strict; trace metal impurities and arsenic impurities directly cause lattice defects inside the single crystal, severely affecting chip yield.
At present, the global high-end high-purity red phosphorus market has long been monopolized by Japanese companies-NCI and RASA Industries together account for more than 80% of the global market share.
The second is equipment. The core equipment for producing and growing indium phosphide single crystals-single-crystal furnaces and MOCVD-still mainly depends on imports. Several of the most critical pieces of equipment for domestic production lines are held in others' hands.
The third is the yield gap. It took Sumitomo Electric of Japan decades to achieve a 6-inch single-crystal yield of over 95%. The comprehensive yield of domestic 6-inch indium phosphide substrates is about 70% to 85%. This is not "about the same"; this is a "generational gap." Low yield means high cost-tolerable when substrate prices are high, but once prices fall, profit margins will shrink sharply.
Thus, various predicaments combine to create the biggest dilemma for China's indium phosphide industry: resources are in hand, but the localization of high-end products cannot keep up.