Semiconductor Export Restrictions Just Got Stricter. Here Is How to Position Your Portfolio

Watching the semiconductor industry in early 2026 has an almost cinematic quality because of its hectic pace, the massive amounts of money that are exchanged, and the quiet anxiety that underlies every earnings call. Somewhere in a trading room, a portfolio manager is staring at a screen, wondering if the regulations have changed once more as Washington continues to tighten its grip on chip exports to China. Most likely, they have.

The most recent rounds of export restrictions on semiconductors, which have been gradually increasing since October 2022 and persisted into March 2025 under the Trump administration, are more than just a technical trade dispute. They stand for something more structural, a calculated attempt to impede China’s access to cutting-edge computing power while simultaneously placing a wager that American chipmakers will be able to handle the fallout. That wager appears to be getting more and more complicated.

It’s worth taking a moment to consider the numbers. A complete decoupling from China could cost U.S. semiconductor companies about $77 billion in sales in a single year, according to research from the Information Technology and Innovation Foundation. It’s not a rounding error. In 2024, U.S. companies supplied slightly more than half of the world’s chip demand, with China accounting for roughly 24%. Eliminating that market completely would be akin to operating on a patient who is still running a marathon.

The fact that the limitations are obviously having some impact—just not in a clear or predictable way—makes this especially difficult. In the years after 2022, workforce pressure increased, prices for some chip types skyrocketed, and China’s semiconductor ecosystem was upended. However, Beijing increased its efforts almost at the same time, initiating what can only be called a government-wide push for chip self-sufficiency. After the first Trump administration cut off Huawei’s access to U.S. technology in 2019, many thought the company was done.

However, it returned with domestically sourced parts, 5G-capable processors, and a new smartphone line that uses mostly Chinese chips. It’s difficult to ignore how disastrously the anticipated collapse did not come to pass.

This puts investors in a situation where accuracy is more important than usual. The Philadelphia Semiconductor Index has been rallying, and the larger supercycle narrative—driven by AI, the growth of data centers, and the seemingly endless demand for processing power—remains intact. However, it seems more and more naive to think that you can just load up on NVIDIA and wait for the money to arrive. Due to geopolitical unpredictability and worries that AI infrastructure spending may be slowing down, NVIDIA’s stock has actually dropped by almost 5% so far this year.

In the meantime, Applied Materials has increased by over 35 percent and Micron Technology has increased by about 28 percent this year. These two businesses are located a little bit away from the geopolitical hotspots.

It’s worth listening to Micron’s story. In comparison to demand, its high-bandwidth memory chips—the kind that enable AI inference at scale—are in short supply, and analysts predict revenue growth that verges on extraordinary. In its third fiscal quarter of 2026, the company anticipates revenues close to $33.5 billion, and its projected earnings growth rate for the year is approximately 600 percent. It’s not a typo. Micron seems to be on the right side of the equation, at least for the time being, as demand-supply imbalances in high-bandwidth memory tend to drive up prices.

A different kind of opportunity exists in Applied Materials; it’s quieter, less glamorous, and possibly more long-lasting. It sells the machinery used by chip makers to construct everything else. The CEO has been openly optimistic, claiming market dominance in high-performance, energy-efficient manufacturing tools and aiming for annual growth of more than 20%.

The company has more natural insulation against the disruption that any one restriction might cause because its revenues are distributed among logic chips, memory, and sophisticated packaging.

Then there is the issue of Broadcom and AMD, both of which have higher Philadelphia Semiconductor Index weights than NVIDIA. Just that fact provides insight into the current state of the market. In the first quarter of fiscal 2026, revenues from Broadcom’s custom silicon business, which produces specialized accelerator chips for hyperscalers like Google and Meta, increased 140% year over year. Revenues from AI networking increased by 60% during that time.

For its part, AMD anticipates that over the next three to five years, data center AI revenue will increase at a rate of more than 80% compound annual growth, supported by agreements with OpenAI, Meta, and other companies. These are not backup narratives. They are the main ones.

The picture of export restrictions actually shows that the semiconductor industry is splitting. As restrictions tighten, companies that rely on Chinese manufacturing partners or are heavily exposed to Chinese demand face real structural risk. Businesses that have either positioned themselves within the U.S. and allied-nation AI buildout or diversified their revenue base are in a fundamentally different position. The irony is that, despite being worrisome from a competitive perspective, China’s increasing drive for self-sufficiency might actually be making investment more clear.

NVIDIA, Broadcom, and Micron are being compelled—and in some cases choosing—to strengthen their ties with hyperscalers in the US, Europe, and Japan as Beijing purchases fewer chips from these companies.

A lot of things are still unknown. It is genuinely unclear if China will truly become self-sufficient in logic chips during this decade. It’s unclear if U.S. export regulations can keep up with Chinese smuggling networks and shell corporation schemes, like the one that allegedly tricked TSMC into creating millions of chiplets for Huawei. Perhaps the most important question of all is whether Washington can convert the revenue pressure that its own businesses are absorbing into actual domestic manufacturing capability.

The era of straightforward semiconductor investing—buy NVIDIA, ride the AI wave—seems to have ended. The wave is still genuine, strong, and worthwhile to ride. However, the underlying currents are now more intricate, influenced by both product cycles and geopolitics.

In about eighteen months, investors who know which companies are actually protected from the consequences of export restrictions and which are more vulnerable than their current stock prices indicate will likely look very wise.