AXT INC (AXTI) stock intrinsic value fair value estimate & owner earnings analysis

AXT makes specialty semiconductor substrate wafers (gallium arsenide, indium phosphide and germanium) used in lasers, LEDs, 5G components and, increasingly, the optical components used to move data inside AI data centers. Nearly all of its wafer manufacturing sits in China, through subsidiaries and partly-owned raw-material joint ventures near Beijing — a structure the company has operated for decades, not a shell or a China-domiciled holding company. Since mid-2023, China has progressively restricted exports of the raw materials and, as of February 2025, the finished indium phosphide wafers that are central to AXT's business, in response to the broader US-China trade dispute. The practical effect has been dramatic: revenue that once reached $141 million (2022) fell to under $76 million in 2023, and sales destined for the United States have shrunk to roughly 2% of the total, with China itself now accounting for 62% of revenue. At the same time, AXT finished a multi-year, very capital-intensive expansion of its manufacturing campus (Dingxing and Kazuo, China); capital spending has fallen from a peak of over $40 million a year to roughly $6-10 million a year now that the buildout is largely complete. Looking at the cash the business has actually thrown off over the last ten years, after removing the swings caused by inventory and receivables timing, stock compensation, and the capital spending genuinely needed just to sustain (not grow) the current plant, AXT has been cash-negative in nine of the last ten years — including during its 2021-2022 revenue peak. This is a structurally capital-intensive, thin-margin business even before the export-control disruption is considered. Our bear case assumes the export-permit friction never meaningfully eases and AXT continues to burn cash for years while it tries to redirect sales to non-U.S. customers, eventually stabilizing at a smaller, still money-losing scale. On that path, the business is worth less than zero on a pure cash-generation basis — the ongoing losses would outweigh what the operating assets are worth, though the roughly $90 million+ the company raised in recent stock sales provides a real cushion this pure cash-flow view doesn't capture. Our base case assumes near-term losses persist for a few more years but gradually shrink and turn modestly positive by the back half of the next decade, as the genuine and growing demand for indium phosphide in AI-related data-center optics (which AXT itself reports is increasing even as realized shipments are capped by permits) slowly finds its way around the licensing bottleneck. Our bull case assumes AXT successfully reroutes a meaningful share of that AI-driven demand to markets and customers not caught by the export-permit chain, allowing revenue and margins to recover well above 2022 levels over the next decade. Because nearly all of AXT's estimated value in every scenario depends on what the business is earning a decade from now rather than on the next few years, this is a speculative situation rather than a steady compounder, and we've priced it with a wide margin of safety accordingly. The key things that would need to happen for our base case to prove too pessimistic: China permits start flowing for indium phosphide exports, or AXT builds a large enough non-U.S.-destined customer base that the permit issue stops mattering to its results. One item we could not fully reconcile from the filings: the exact split of AXT's 2017-2022 capital spending between expanding capacity and simply maintaining it; we used the company's more recent, post-buildout spending level as our best estimate of ongoing maintenance needs, which is a judgment call rather than a directly disclosed figure.