Tech Sell-Off: This All-Star Artificial Intelligence Stock Is Down 50% but Up 127,300% Since Its IPO. Here's Why It's a Screaming Buy Heading Into the Midterms.

The four-month sell-off in artificial intelligence (AI) chip stocks has been severe. Since June 30, the iShares Semiconductor ETF (SOXX +2.85%) has fallen more than 25%, and many of the stocks it holds have fallen much more.

Recently, one of the biggest winners of the AI boom fell sharply after reporting its latest earnings. Shares of this company, which has performed better than the market for many years, are now almost 50% below their record highs in June.

But this successful stock has suffered similar drops before. Even so, its shares have risen more than 127,000% since its initial public offering in 1984. That figure does not even include its growing dividend. Company leaders just raised the dividend by 27%, bringing the current dividend yield to 0.57%.

Lam Research plays a key role in AI

Semiconductor equipment maker Lam Research (LRCX +3.22%) plays an important role in AI. Its advanced machines, which remove and add thin layers of material, are essential for making the latest logic and memory chips. Lam specializes in building chip structures that go upward in layers. Making these structures requires carefully stacking materials at the atomic level.

Years ago, NAND flash memory makers began stacking memory cells on top of one another. Lam went on to dominate sales of machines used in NAND production. Today, advanced logic chips are also using more vertical structures, including gate-all-around transistors. DRAM, another type of memory, is also becoming more vertical through high-bandwidth memory (HBM) and three-dimensional DRAM, known as 3D DRAM.

Why this quarter's guidance should not matter

The reason for the supposedly weak forecast is not falling demand. It is a lack of supplies. Company leaders said there is not enough cleanroom space for its machines, and shortages of certain machine parts will stop Lam from reaching its full potential this quarter. Chief Financial Officer Doug Bettinger said the forecast would have been "a bit higher" without these limits.

Meanwhile, company leaders expect the semiconductor industry's spending on wafer front-end equipment (WFE) to rise to more than $200 billion in 2027, up from about $155 billion this year. Keep in mind that Lam's management also believes this forecast is limited by supply shortages. In addition, management did not say how far above $200 billion spending might go. Since it is still October 2026, this figure is likely a very cautious estimate.

Spending of more than $200 billion on WFE would suggest that Lam could grow by about 30% next year if it grows at the same rate as the industry. But its growth could be much higher. Lam should also grow faster than the overall WFE market because its machines are important for making vertical chip structures. Finally, Lam has a history of increasing its profit margins when the industry is growing, and it is also buying back its own shares. These factors should help earnings per share rise even more, as they did last quarter.

Lam has another possible source of growth: new NAND technology

Much of Lam's growth over the past couple of years has come from advanced logic chips and DRAM, which have benefited most from the growth of AI infrastructure. But NAND flash memory, where Lam has its strongest market position, may soon become a bigger source of growth.

Over the past decade, NAND flash makers have tried to fit as many bits of information as possible into each memory cell. They have also stacked as many memory layers as they could onto each chip.

Lam has benefited from the process of stacking more NAND layers, often by selling upgrades for existing machines. But making production more efficient this way has limited the overall growth in the number of wafers produced by the NAND industry. As a result, manufacturers have needed fewer new NAND machines.

However, companies building AI infrastructure may soon start using a type of memory called SLC NAND, or single-level cell NAND. It stores just one bit of information in each memory cell. AI companies may choose it because it has lower latency, which means it can respond faster. For tasks where speed is especially important, such as agentic AI, SLC NAND may be the best choice, even though it stores less information.

If manufacturers start using NAND that stores less information in each cell, they may need to produce more wafers to meet demand. Lam's management said that shifting just 1% of NAND production capacity to SLC NAND would require a 2% increase in the number of NAND wafers produced.

During a recent conference call with analysts, Lam CEO Tim Archer said, "NAND is yet to emerge as to its full potential from our perspective, and I think you are starting to hear about it and the strategic role it can play in AI."

The stock may be a good buy again

After its recent price drop, Lam trades at just 23.9 times the average earnings per share analysts expect for its 2027 fiscal year, which ends on June 28, 2027. If we use earnings per share for the full 2027 calendar year, which should be higher, the stock's price-to-earnings ratio could be below 20. That would put it below the overall market's valuation.

Calendar year 2027 should be a very strong year for the semiconductor equipment industry. It is harder to know what will happen in 2028 and later. But industry leaders have recently talked about supply shortages, and investment in NAND may soon begin growing faster again. Given these factors, Lam's relatively low valuation may be an attractive price for this leading chip equipment company. After the stock's decline this summer, Lam Research is a buy once again.