It hasn’t been that long since investors have seen the reality of declining stock prices. The last time one of the main market indexes fell more than 10% from its all-time high, which is called a correction, was way back in October 2023. Sometimes these corrections are slow, and sometimes the losses happen swiftly. Between February 19 and March 10 of this year, the Nasdaq Composite dropped nearly 13%. This sell-off was largely due to former President Donald Trump’s trade policies and concerns about additional tariffs on Taiwan, a major supplier of chips used in AI data centers.
As a result, some of the biggest movers in the AI sector saw their stock prices drop. Nvidia (NVDA -0.75%), Broadcom (AVGO 0.55%), and Taiwan Semiconductor Manufacturing Company (TSMC -0.56%) all experienced significant price declines. Combined, these companies lost $1.16 trillion in market capitalization during that period. While it might be tempting to buy shares in all three companies at these lower prices, one stands out as an incredible value with a sustainable competitive advantage.
The Factors Driving AI Stock Declines
Several factors have contributed to the drop in AI stocks over recent weeks. Increased economic uncertainties have impacted consumer confidence, as U.S. trade policies have heightened geopolitical tensions. Markets generally dislike uncertainty.
Perhaps the most significant uncertainty hitting chipmakers is the potential for the Trump administration to enact new tariffs on Taiwan, the home of TSMC. Nearly all the biggest chipmakers, including Nvidia and Broadcom, depend on TSMC to produce and package their chips. TSMC attracts nearly two-thirds of all spending on chip fabrication.
A tariff on Taiwan would significantly increase costs for Nvidia and Broadcom, and practically every other chipmaker. Subsequently, they would have to increase their prices or see a hit to their profit margins—or likely both. Higher prices could also lead to decreased demand for their chips.
While the big tech companies buying Nvidia and Broadcom chips have ample budgets, their financial resources are not limitless. Those companies are also under increasing pressure to provide meaningful returns on their investments. With the increased pressure, there may not be much room in their budgets to increase spending. Decreased demand cycles back to TSMC, which faces the challenge of substantial fixed costs. Lower facility utilization rates could result in a significant drag on profitability if tariffs go into effect or demand declines for any other reason.
TSMC has taken steps to appeal to the Trump administration. It has committed to invest an additional $100 billion in the U.S. in addition to its plans to expand its facilities in Arizona over the next two years. If the current administration pushes to have more chip manufacturing occur in the U.S., TSMC is signaling its willingness to fulfill that initiative.
Long-Term Investment Considerations
In the midst of the current sell-off, investors should consider the long-term potential of any investment. Nvidia’s position appears most precarious in the long run. Higher costs for its chips could accelerate a shift from its biggest customers to more cost-efficient alternatives.
Meta Platforms is already developing a custom AI accelerator chip for training its Llama foundational models. It is reportedly aiming to use those chips for training by 2026. It currently uses its own chips for machine learning and plans to expand its use to AI inference this year.
The other three hyperscalers have expressed similar goals and have seen positive results with their custom silicon. It is worth noting that Meta and Alphabet both rely on Broadcom’s technology to create custom chips. Thus, rising costs could end up benefiting Broadcom’s custom AI accelerator business. Management expects that business, in conjunction with its network solutions, to reach a serviceable addressable market of between $60 billion and $90 billion by 2027. However, its networking chips are still a major portion of that business, so the effect could be muted.
TSMC may not be as impacted over the long term as some might think. Its technology lead is substantial. Nvidia CEO Jensen Huang called TSMC “the world’s best by an incredible margin.”
Switching from TSMC to another foundry isn’t a viable option for Nvidia, Broadcom, or most of its other key clients. First, there aren’t many options with the manufacturing scale they need. Expanding production capacity takes a long time. Secondly, these chips are designed using TSMC’s processes. In some cases, they are designed with custom TSMC processes, like Nvidia’s Blackwell platform. Switching to a competitor would require months of redesign and validation.
Invest in the AI Stock with Staying Power
TSMC likely has the most sustainable long-term competitive advantage, which is a self-reinforcing phenomenon. As TSMC attracts more revenue for high-end chip designs than any other foundry, it can invest more in research and development, new equipment, and capacity expansion, which strengthens its position to win even more contracts in the future. Though the foundry could see a downward blip in demand, it doesn’t face a significant competitive threat.
Moreover, demand should remain relatively stable as hyperscalers switch to lower-cost alternative GPUs or their custom silicon. TSMC has contracts with all of them, including Meta for its new custom AI chip. Most importantly, the stock’s valuation is currently very attractive. After the recent sell-off, investors can purchase it for less than 20 times forward earnings estimates at the time of this writing. Even if the company experiences some margin contraction and slower demand growth in the short term, that’s a price that can easily absorb any impact. Its company growth prospects and robust competitive advantages remain notable.
