LONDON (Reuters) - European companies that are spending big on generative artificial intelligence need to start showing returns on their massive outlays by next year, or risk investors losing patience after they paid sky-high prices to join the market boom.

AI-exposed stocks have been caught in a down-draught with broader equity markets in recent weeks as recession fears rise, adding to pressure on the sector since January, when the launch of low-cost Chinese AI model DeepSeek spurred a tech selloff.

While many investors are bullish about Gen-AI's potential to boost productivity and profits, some are becoming more choosy, with a preference emerging for shares of companies adopting AI technology - like information group RELX ( RELX ) and software firm SAP - over those supplying chips and other hardware to the industry.

But they also say adopters need to start showing returns on the cash they've poured into the tech, or investors might cool on them too.

PENDULUM SHIFTS TO THE ADOPTERS

Chipmaker Nvidia has become almost synonymous with the AI boom. Even though its shares were hit hard by the rollout of DeepSeek's AI model, which requires fewer of the company's pricier chips, its stock is up 29% year on year.

In Europe, there are fewer AI-exposed stocks to choose from than on Wall Street, but the trend is clear.

Among hardware makers, chip equipment makers ASM International and BE Semiconductor are down a respective 25% and 20% since the January 24 DeepSeek-inspired selloff and amid U.S. recession fears. France's Schneider Electric, which provides electrical equipment to data centres, is down 14%.

Among AI adopters, meanwhile, data group LSEG is down 5.5%, while RELX ( RELX ) is just 1.6% lower. German business software group SAP is down 2.9% and on Monday overtook Novo Nordisk to become Europe's most valuable company.

"Everyone was very narrowly invested in the AI enablers," said Gerry Fowler, head of European equity strategy at UBS.

"With DeepSeek, because it cheapens AI so much, it means to some extent the markets are now looking at who are the outright beneficiaries of AI."

PATIENCE HAS ITS LIMITS

An internal survey published in January of over 100 Fidelity analysts found that almost 72% of them expected AI to have no impact on the profitability of the companies they cover in 2025.

A lot more of the Fidelity analysts surveyed saw a positive impact over five years. Yet several European portfolio managers told Reuters their timeframe is shorter than that.

"The market will lose patience with unfettered investment in AI unless it starts to see some return on investment out of the end of it," said Steve Wreford, lead portfolio manager on the global thematic equity team at Lazard Asset Management.

Wreford said companies adopting AI are likely to be given a pass if they don't deliver much in 2025, when they are likely to roll out beta testing and trials, but that by 2026, investors need to start seeing a big impact on their top lines.

AI-exposed stocks' valuations are relatively pricey. The STOXX 600 trades at an average price-to-earnings multiple of 17 times, while AI adopters like SAP and LSEG trade around 90-plus times, according to LSEG data.

Bernie Ahkong, chief investment officer at hedge fund UBS O'Connor, said investors would begin questioning some companies' multiples if they don't deliver by the end of 2025.

"During the year, management teams can always use the excuse, don't worry it's next quarter... for a multi-year theme. By the time it gets to Q4 and it hasn't come then ... people are not going to be patient anymore," he said.

KILLER USE CASE

The biggest risk in investing in AI generally for Paddy Flood, portfolio manager and global sector specialist, technology, at Schroders ( SHNWF ), is whether viable use cases emerge that people are actually willing to pay for.

"To justify continued spending, we need to see concrete applications, whether a single 'killer' use case or a range of impactful ones."

Fabio di Giansante, head of large European equities at Amundi, Europe's biggest asset manager, said the relative scarcity of European AI plays meant they already come at a premium, but that so far much of the news from the sector has been about orders and capital spending.

"At some point you need to see the effect of those things benefiting the top line and margins," he said, adding that if the benefit is not as big as inferred by valuations, multiples might be reassessed.

"This year could be the year this happens on a wide scale."