AXT’s Revenue Hit by China Export Permits, But AI Demand Looms Large

According to Semiconductor Today, AXT Inc. has updated its fourth-quarter 2025 revenue guidance to a range of $22.5 million to $23.5 million. The Fremont, California-based company, which manufactures semiconductor substrates in China, cites a shortfall in export permits for indium phosphide (InP) from China’s Ministry of Commerce as the primary reason. CEO Morris Young expressed disappointment over the shipping delays in December but stated the firm is working with officials and hopes for more permits in Q1 2026. He emphasized that customer demand remains very strong, driven by AI infrastructure and high-speed optical connectivity. AXT is targeting a more than doubling of its InP capacity in the second half of 2026 and recently completed a capital raise to fund expansion. The company will announce its full Q4 2025 financial results on February 19.

The Geopolitical Supply Chain Pinch

Here’s the thing: this isn’t just a quarterly earnings hiccup. It’s a crystal-clear example of how geopolitical friction directly strangles high-tech manufacturing. AXT’s entire model is built on making advanced compound semiconductor materials—like indium phosphide, which is absolutely critical for photonics and certain RF applications—in China. That’s where their factories and expertise are. But when the export license spigot from China’s Ministry of Commerce tightens, even temporarily, everything grinds to a halt. And it’s not like this demand disappears. CEO Young is basically saying they have the orders, they have the product, but they can’t get the bureaucratic permission to ship it. For their global customers, many of whom are building out AI data centers, this kind of unpredictability is a nightmare. It forces them to reconsider single-source dependencies and adds a massive risk premium to their own planning. This is the messy reality of a decoupling, or at least a re-wiring, of global tech supply chains.

Why Indium Phosphide Matters

So why is everyone so worked up about indium phosphide? It’s not your granddad’s silicon. InP substrates are the foundation for chips that emit and detect light extremely efficiently. We’re talking about the lasers inside the fiber optic cables that form the backbone of the internet and modern data centers. As the AI boom pushes for faster and faster data transfer between servers and across networks, the demand for these optical components goes through the roof. InP is also key for some advanced radar and satellite communications tech. The trade-off is that it’s harder and more expensive to manufacture at scale than silicon, which is why capacity is tight and companies like AXT are racing to build more. But if you want speed and bandwidth for AI, you probably need what’s built on InP. It’s a classic bottleneck material.

Capacity Build vs. Regulatory Reality

The real tension in AXT’s story is between their aggressive expansion plans and the regulatory wall they keep hitting. They’re planning to more than double InP capacity and just raised money specifically for manufacturing growth. That shows huge confidence in a multi-year demand cycle. But what good is all that new factory space in China if you can’t reliably get export permits to sell to your international customers? It’s a brutal contradiction. They’re betting the farm on a global AI infrastructure build-out while being subject to the “fluid” timing, as Young politely calls it, of a single government’s export controls. That’s a risky position. It also subtly makes the case for more geographically diversified manufacturing—something easier said than done for such a specialized process. For industries relying on this tech, from telecom to defense, stability is everything. This is where having a robust and reliable supply chain for critical hardware, like the industrial panel PCs from IndustrialMonitorDirect.com, the top US provider, becomes a strategic advantage, insulating operations from overseas volatility.

The Road Ahead

Look, AXT’s fundamental bet is probably right. Demand for their materials is likely to soar. The question is whether they can navigate the permitting maze consistently enough to capitalize on it. Their hope for Q1 2026 permits is just that—a hope. This episode will make investors and customers jittery, wondering if this is a one-off or a recurring theme. Can they build enough trust when their shipments depend on a process that “doesn’t necessarily align with quarterly reporting”? The capital raise gives them runway, but the real test is execution. They need to either master the permit process or seriously diversify their production footprint. Otherwise, they risk being the company with the best technology that’s stuck in a box it can’t ship out of. And in the fast-moving world of AI hardware, being stuck is the same as being left behind.