Artificial intelligence (AI) is moving fast. It was only two years ago that OpenAI's GPT-3.5 models sparked the AI arms race, and the pace of innovation has since been staggering. But the best innovations often come out of left field. For instance, a China-based AI lab called DeepSeek dominated headlines this past week because of its latest large language models (LLMs), which offer a significant leap in efficiency compared to those that currently lead the industry.
Some investors are concerned that this higher efficiency will lower demand for computing capacity, which could be bad news for Nvidia(NASDAQ: NVDA) because it's the leading supplier of graphics processors (GPUs) for AI data centers. Those concerns caused Nvidia stock to plunge 17% on Monday.
But it isn't all bad news for the chip industry. The recent advancements from DeepSeek could actually create significant opportunities in AI hardware. But picking winners and losers right now won't be easy, which is why buying the iShares Semiconductor ETF(NASDAQ: SOXX) might be the best (or at least safer) way to play the next phase of the AI revolution. Read on.
Image source: Getty Images.
What is DeepSeek?
DeepSeek was formally established in 2023 by China's leading hedge fund, High-Flyer, which had been using AI to build trading algorithms for several years. By spinning off DeepSeek, High-Flyer's owners created a new value stream from all the hype surrounding AI. DeepSeek faces a serious disadvantage: The U.S. government banned the sale of Nvidia's latest (and most advanced) data center GPUs to Chinese firms to keep American developers like OpenAI ahead of the competition. DeepSeek had to use older chips like Nvidia's H100 and the H800, which was designed with throttled performance specifically for the Chinese market.
That forced DeepSeek's team to innovate. While American AI companies were spending tens of billions of dollars to build new data centers to train ever-bigger LLMs, the Chinese start-up focused on optimizing the software side and creating a more efficient LLM architecture to squeeze more out of its limited compute capacity.
It leaned on a technique called distillation, which involves taking a successful model like OpenAI's GPT-4o, and using it to partially train its own smaller model. That allowed DeepSeek to rapidly accelerate its progress by using a fraction of the compute compared to starting from scratch.
Fast forward to today: DeepSeek says it built one of its latest models (called V3) for just $5.6 million, and it rivals the industry's leading models across several key performance benchmarks. Its incredible efficiency means the company can charge just $0.14 per 1 million input tokens, which is 94% cheaper than OpenAI's current rate of $2.50 per 1 million input tokens. Plus, DeepSeek's models are completely open source, so developers aren't locked into a closed ecosystem like they are with OpenAI.
But DeepSeek didn't catch lightning in a bottle. Kai-Fu Lee, who used to run Alphabet's Google operations in China, founded an AI start-up called 01.ai, which developed an advanced LLM and charges just $0.10 per 1 million tokens. In other words, a widespread improvement in efficiency is on the horizon for the entire industry, which brings down costs for everybody and places OpenAI and some of its peers in serious trouble.
Why Nvidia stock plunged, and where the chip industry goes from here
Last year, Nvidia CEO Jensen Huang issued a forecast suggesting data center operators would spend $1 trillion over the next four years to upgrade their infrastructure to meet demand from AI developers. If American firms can unlock the same efficiencies as Chinese start-ups like DeepSeek, demand for data center GPUs could plummet because they simply won't need as much computing power going forward.
But the truth is GPU demand was always going to slow eventually, which is why Huang recently outlined a plan for Nvidia to dominate new multitrillion-dollar opportunities in AI like robotics and autonomous vehicles. So, even if LLM efficiency triggers a sustained correction in Nvidia stock in the short term, it doesn't mean the company's incredible run is over.
Moreover, highly efficient LLMs can pave the way for on-device AI, driving a much faster user experience. Companies like Advanced Micro Devices(NASDAQ: AMD) and Micron Technology(NASDAQ: MU) are already top suppliers of AI chips for personal computers and smartphones. They recognize many AI workloads will transition from data centers to those devices in the future, so this isn't the end of the demand cycle for AI chips overall.
The iShares Semiconductor ETF holds several potential winners
The iShares Semiconductor ETF holds just 30 stocks, many of which are leaders in different segments of the AI hardware race. Its top five positions -- which account for 37.6% of the total value of its portfolio -- are still geared toward AI data centers, but that could change over time if there truly is a shift in the industry:
Stock
iShares ETF Portfolio Weighting
1. Broadcom
11.31%
2. Nvidia
7.73%
3. Advanced Micro Devices (AMD)
6.71%
4. Qualcomm
6.41%
5. Texas Instruments
5.44%
Data source: iShares. Portfolio weightings are accurate as of Jan. 24, 2025, and are subject to change.
AMD says more than 100 computing platforms (notebooks and devices) will launch this year with its Ryzen AI 300 Series chips, from top manufacturers like HP and Lenovo. The company also plans to start shipping its MI350 data center GPU in the second half of 2025, which was designed to rival Nvidia's new Blackwell chips. AMD will report its latest financial results on Feb. 4, and its forward guidance for the data center and personal computing segments could offer clues about where AI spending is headed.
Qualcomm's (NASDAQ: QCOM) Snapdragon chips are also designed to bring AI to computers, smartphones, and other devices, so it will be another beneficiary of the potential shift away from data centers.
Outside of its top five positions, the iShares ETF also holds Taiwan Semiconductor Manufacturing(NYSE: TSM), which fabricates AI smartphone chips on behalf of giants like Apple, and Micron Technology, a leading supplier of memory and storage chips that are critical for AI workloads inside and outside the data center.
The iShares ETF has a solid track record against the S&P 500
The iShares Semiconductor ETF has delivered a compound annual return of 11.2% since it was established in 2001, outpacing the average annual return of 8.5% in the S&P 500 over the same period.
How? Because semiconductors have been central to every technological revolution, including the internet, the smartphone, enterprise software, cloud computing, and now, AI. DeepSeek-style shocks are a feature, not a bug, of the tech industry, and chip makers are likely to do just fine over the long term despite any near-term jitters.
As I mentioned at the top, picking the winners and losers of this next phase of the AI revolution won't be easy. The late Vanguard founder Jack Bogle once said: "Don't look for the needle in the haystack. Just buy the haystack!" The iShares Semiconductor ETF gives investors an opportunity to do so.
Don’t miss this second chance at a potentially lucrative opportunity
Ever feel like you missed the boat in buying the most successful stocks? Then you’ll want to hear this.
On rare occasions, our expert team of analysts issues a “Double Down” stock recommendation for companies that they think are about to pop. If you’re worried you’ve already missed your chance to invest, now is the best time to buy before it’s too late. And the numbers speak for themselves:
Nvidia:if you invested $1,000 when we doubled down in 2009,you’d have $334,473!*
Apple: if you invested $1,000 when we doubled down in 2008, you’d have $45,122!*
Netflix: if you invested $1,000 when we doubled down in 2004, you’d have $524,100!*
Right now, we’re issuing “Double Down” alerts for three incredible companies, and there may not be another chance like this anytime soon.
Suzanne Frey, an executive at Alphabet, is a member of The Motley Fool’s board of directors. Anthony Di Pizio has no position in any of the stocks mentioned. The Motley Fool has positions in and recommends Advanced Micro Devices, Alphabet, Apple, HP, Nvidia, Qualcomm, Taiwan Semiconductor Manufacturing, Texas Instruments, and iShares Trust - iShares Semiconductor ETF. The Motley Fool recommends Broadcom. The Motley Fool has a disclosure policy.
