This article collates the work of our colleagues in the US and Europe on Monday's AI stock sell-off with added reaction from Australia.

US chip stocks and other artificial-intelligence-related names rallied on Tuesday after selling off sharply on Monday following Chinese AI lab DeepSeek’s announcement that a new version of its AI model has less need for semiconductors.

The news was seen as having potentially significant implications across a wide range of stocks, from semiconductor designers and makers to utilities who investors believed would benefit from booming electricity demand for data centers.

Following a selloff in Europe, key US-listed stocks such as Nvidia (NAS: NVDA) and Broadcom (NAS: AVGO) fell sharply after the US market open. Both semiconductor giants were down more than 16% on Monday before rebounding by around 8% and 2% respectively on Tuesday.

Losses from Monday’s rout spread into sectors including energy and utilities, with the Morningstar US Utilities Index falling 4.8% and power giant Vistra Corp (NYS: VST) posting losses of more than 25%.

In Europe, Dutch semiconductor stock ASML (AMS: AMSL) lost 7% on Monday, while BE Semiconductor Industries (AMS: BESI) was down 11%, shares in ASM International (AMS: ASM) fell 12% and shares in Arm Holdings (NAS: ARM), a UK-based company listed in New York, were also down nearly 10%.

Why did US AI stocks sell off on Monday?

“The investment case for the AI supply chain until now was that more spending led to better outcomes for AI models,” says Morningstar equity analyst Javier Correonero.

“Big tech firms Microsoft (NAS: MSFT), Alphabet (NAS: GOOG), Amazon (NAS: AMZN) and Meta Platforms (NAS: META) have deployed hundreds of billions to purchase GPUs from Nvidia and ensure chip supply to satisfy the insatiable demand for AI.”

Morningstar thematic ETF expert Kenneth Lamont explains how DeepSeek could signal a move away from “brute-force computing power” and rewrite the investment case for AI. “Until now, the conventional wisdom has been clear: The best AI models rely on massive datasets and immense computational power, rewarding scale and favoring hardware giants like Nvidia and ASML,” he says.

“DeepSeek’s latest innovations are turning that assumption on its head. The startup’s new models demonstrate how efficiency gains in AI development can reduce reliance on brute-force computing power. This breakthrough slashes computational demands, enabling lower fees and putting pressure on industry titans like Microsoft and Google to justify their premium pricing.”

Less power-hungry AI weighs on grid companies

“Shares in Siemens Energy (ENR) fell 20% on Monday, part of a broader selloff for companies with exposure to electricity consumption and AI,” observes Morningstar analyst Matthew Donen. News about DeepSeek “has raised questions about the true cost of AI models, which initially appeared to be extremely energy-intensive. Should the cost decline, as Deepseek’s model implies, the market’s expectations of electricity consumption will decline.”

Donen adds: “Siemens benefits from increasing electricity consumption through the sales of its gas turbines and high-voltage electricity equipment. French electrical infrastructure groups Legrand (LR) and Schneider Electric (PAR:SU) also saw their share prices take a hit.”

ASX shares and ETFs affected

Given what happened overnight in the US, the fate of Australian listed shares and ETFs with exposure to the AI theme was hotly anticipated ahead of Tuesday’s market open. In the end, data centre operators and uranium shares were among the biggest losers.

Uranium mine developers Bannerman Energy (ASX: BMN), Lotus Resources (ASX: LOT) and Deep Yellow (ASX: DYL) all fell by more than 15%, while the Betashares Global Uranium ETF (ASX: URNM) lost 10%. Meanwhile in the data centre camp, Goodman Group (ASX:GMG) shares fell by 8%, NextDC (ASX: NXT) lost 7% and DigiCo’s Infrastructure REIT (ASX: DGT) slumped 11.5%.

“For Goodman, the main question looming behind all this is whether enormous spending on AI infrastructure remains justifiable” says our Australian REIT analyst Yingqi Tan, who went on to list three major risks she sees to the data-centre segment of Goodman’s business.

“First, funds under management growth could be slower as capital partners become more cautious about investing in AI infrastructure. Second, development income could fall as data center take-up moderates. Third, Goodman may not be able to raise rents as much.”

“More than anything, the DeepSeek launch sends a reminder of the dangers of high expectations” says Tan, who maintained her $27 per security Fair Value estimate for Goodman. This is almost 25% below its current market price, even after Tuesday’s fall.

In the ETF space, funds focused on AI and semiconductors fared poorly. Global X’s Semiconductor ETF (ASX: SEMI) fell by over 8% on Tuesday, led by its two biggest holdings Broadcom and Nvidia.

ETFs tracking the US’s Nasdaq technology index, such as Betashares Nasdaq 100 ETF (ASX: NDQ), posted a smaller loss of around 2% and are set to recoup much of this after the index bounced overnight due to the rally in Nvidia.

Fair Value for big US names left unchanged

Morningstar will maintain its Fair Value estimate of USD 130 per share for Nvidia, says equity strategist Brian Colello.

“Despite DeepSeek’s promise, we doubt the leading cloud vendors and AI builders will pause their plans, although it’s a risk that certainly bears watching. We believe AI GPU demand still exceeds supply, so while slimmer models may enable greater development for the same number of chips, we still think tech firms will continue to buy all the GPUs they can as part of this AI ‘gold rush,’” he explains.

We have also held our Fair Value estimates for Microsoft (USD 490 per share), Amazon (USD 200), and Alphabet (USD 220). Our analyst Dan Romanoff sees these wide-moat firms benefiting from a commodified LLM layer as increased spending on AI boosts their public cloud businesses.

“R1’s launch and its dramatically lower pricing (more than 90% below OpenAI’s latest reasoning model) go hand in hand with our broader “commodification of complements” view of the large language model space” he says.

“We believe that as the price of LLMs (the complementary good) goes down, the value and usage of the public cloud vendors’ primary good, cloud infrastructure, increases. To that end, we believe Amazon, Microsoft and Google benefit from reduced LLM pricing in the long run."

Romanoff also expects these large US tech companies to replicate some of the AI techniques DeepSeek used to reduce the costs of their own model training and inference. This could potentially lower their medium- and long-term capital expenditures.

Terms used in this article

Star Rating: Our one- to five-star ratings are guideposts to a broad audience and individuals must consider their own specific investment goals, risk tolerance, and several other factors. A five-star rating means our analysts think the current market price likely represents an excessively pessimistic outlook and that beyond fair risk-adjusted returns are likely over a long timeframe. A one-star rating means our analysts think the market is pricing in an excessively optimistic outlook, limiting upside potential and leaving the investor exposed to capital loss.

Fair Value: Morningstar’s Fair Value estimate results from a detailed projection of a company's future cash flows, resulting from our analysts' independent primary research. Price To Fair Value measures the current market price against estimated Fair Value. If a company’s stock trades at $100 and our analysts believe it is worth $200, the price to fair value ratio would be 0.5. A Price to Fair Value over 1 suggests the share is overvalued.

Moat Rating: An economic moat is a structural feature that allows a firm to sustain excess profits over a long period. Companies with a narrow moat are those we believe are more likely than not to sustain excess returns for at least a decade. For wide-moat companies, we have high confidence that excess returns will persist for 10 years and are likely to persist at least 20 years. To learn more about how to identify companies with an economic moat, read this article by Mark LaMonica.