Mega-cap stocks tumbled Monday around fears the improvement from AI stocks might be ending.

The biggest names in the US stock market pulled back sharply to kick-off trading this week. This move was initiated by fears surrounding a new AI model announced by DeepSeek, a Chinese AI company that released a paper alongside their R1 model last week. This model appears to have comparable performance to the top models from domestic companies like OpenAI, but at a fraction of the cost (source: Yahoo Finance). The DeepSeek R1 model also utilizes an open-source framework that originated from Meta’s Llama models, meaning that any developer can use the model and see how it works. The mobile version spiked in popularity on Monday, overtaking ChatGPT for the top free app on Apple’s App Store before new registrations were halted (source: cnbc.com).

The public was led to believe that Chinese AI companies were not capable of building models like those built in the U.S. due to restricted access to the most powerful processors. Assuming the information released on the R1 model is accurate, DeepSeek proved that it could circumnavigate the chip restrictions by building on top of readily available open-source models. This was the source of investor panic on Monday; if DeepSeek could build a competitive model with a fraction of the investment in processing power, should we expect other companies to reduce their investments in processing power moving forward?

We are not in the game of predicting the future by attempting to figure out an answer to that question. All we can do is evaluate investor’s reactions and how those reactions affect the relationship between supply and demand for a given security. However, understanding the context around the AI supply chain is helpful. For the purposes of today’s discussion, we will divide the space into two parts; companies that build the processors, and companies that buy the processors. That is not meant to represent the space in its entirety. Electric utility companies are also a big part of this supply chain, as the companies that buy the processors need increased energy production to power their expanding purchases. The utilities sector (XLU) showed the second-worst declines of any broad sector Monday, behind technology (XLK). Those companies will also be important to monitor, but today’s conversation is geared toward the mega-cap companies that make up most of your major market indices.

Much of the AI-related stock appreciation has been centered around semiconductor manufacturers, as the companies driving AI innovation need high-powered processors to run their systems. Monday’s market action saw the VanEck Semiconductor ETF SMH drop by about 10%. Investors seem to be assuming that more efficient AI models will mean less reliance on high powered processors from the companies behind the models. This moved SMH to a sell signal but leaves it at support from November 2024. The decline also led SMH to reverse down into a column of O’s on its market relative strength chart, now sitting one box away from giving a sell signal against SPXEWI. Using this fund as a proxy, the broad semiconductor space still maintains a favorable long-term technical picture but is showing near-term weakness. We have seen SMH oscillating between $235 and $265 since last October. This range has produced some stagnation in relative strength as the improvement experienced over the last two years has cooled. Additional downside could turn that stagnation into relative weakness. On the other hand, some consistent improvement after this pullback (i.e. a series of higher lows) could be an opportunity to add exposure. Further support below this range can be seen at $215 and $205, while overhead resistance is seen at $265 and the all-time highs of $280 from July.

Next in the supply chain is the companies that have been investing in those high-powered processors, namely Microsoft MSFT, Meta Platforms META, Amazon AMZN, Alphabet GOOGL and Tesla TSLA. Most of those companies opened the day with initial declines that did not produce many changes to the stock’s long-term technical pictures. META climbed back into positive territory intra-day and printed a new all-time chart high Monday at $656. MSFT and AMZN saw declines but saw no changes to their default charts. TSLA dropped by over 3%, leading to a few more Os printed on the default chart, but the stock remains on a buy signal. GOOGL saw the sharpest decline of any of these names, dropping over 4% to reverse lower into a column of Os from the all-time chart high of $200. The stock still sits on two consecutive buy signals and shows initial support close by at $188 and $186.

Sharp pullbacks for these growth-oriented equity areas are common. For perspective, we have examined the number of 5% pullbacks experienced by a variety of representatives below, including from the Nasdaq-100 Index NDX, broad semiconductors (SMH), and the eight companies that had a market cap north of $1 trillion through Friday. Each of these individual companies have benefitted notably from the AI-related hype over the past two years. Our study simply examines each time the security in question experiences a 5% pullback that is followed by a 5% rally in share price (used to separate each individual instance) based on data through last Friday (1/24). Some of these representatives have not experienced a 5% pullback in the recent decline, while some have seen amplified weakness with the market action on Monday. Regardless, we see some consistent themes emerge that are outlined above the table.

Key Takeaways:

• The Nasdaq-100 Index NDX experiences about six pullbacks of at least 5% each year on average, dating back to 1993. We only saw three such pullbacks in 2023 and another three in 2024, well below the average.
• Semiconductors (SMH) tend to see more pullbacks, averaging nearly nine a year since 2000. We saw six such events in 2023 and 11 pullbacks in 2024.
• The individual stocks tend to show more substantial price movement than the broader representatives. Five of the eight stocks examined average double-digit counts of pullbacks each year.
• The average 5% pullback leads to a 10% correction for each security in our examination.
• The higher number of pullbacks for these names also produces quicker short-term rallies, raising the probability of a snap-back in the near-term.