The pros give their take on whether AI stocks are back in the game

The bull run in tech stocks hit the skids at the start of the year when Chinese start-up DeepSeek revealed it had developed an AI chatbot to rival ChatGPT on a much smaller budget, raising doubts over whether the US would continue to dominate the sector. Technology shares plummeted, led by GPU manufacturer and artificial intelligence (AI) bellwether Nvidia (NASDAQ:NVDA) – the engine behind ChatGPT – which fell almost 17% on the day of the announcement.

Before the market could catch its breath, US President Donald Trump began rolling out a series of tariffs on US imports, culminating in the announcement of further sweeping tariff hikes on Liberation Day, sending global markets into a tailspin. Tech stocks were swept up in the turbulence as investors fretted over the potential disruption to supply chains for AI products and tariffs weighing on the cost of AI development.

• Invest with ii: Open an ISA | ISA Investment Ideas | Transfer a Stocks & Shares ISA

But as trade wars loomed large Trump hit the pause button, announcing a 90-day suspension on tariffs for all countries except China, soothing the battered stock market. There was further relief for the markets on 12 May when the US and China finally agreed to slash tariffs for 90 days.

In another boon for AI companies last week, the US and the United Arab Emirates signed an agreement for the Gulf country to build the largest artificial intelligence campus outside the United States, which included a pact for the UAE to import 50,000 of Nvidia's AI chips per year, powering Nvidia’s share price up 14% over the course of the week.

AI stocks may be on the ascendent once again, but will AI challengers continue to disrupt the market and is a re-escalation of trade tensions on the horizon?

• Is US stock market exceptionalism over?
• Have investors been selling, buying, or sticking with the US?

'DeepSeek misundersanding' 

According to Axa Investment Managers’ James Dowey, co-lead portfolio manager of the AXA Framlington Global Technology fund, much of the reaction to DeepSeek was the result of the market overestimating its abilities.

“There was an element of misunderstanding around what DeepSeek was able to do,” he says. “DeepSeek were able to complete the last mile of the marathon for $6 million because they stood on the shoulders of other AI models; they tweaked those models and came up with something far more efficient.”

Dom Rizzo, portfolio manager of the T. Rowe Price Global Technology Equity strategy, says that while DeepSeek was seen as a threat to more established players, it could actually prove beneficial to US tech firms.

“Innovations like those from Chinese Al company DeepSeek could bring a level of utilisation efficiency to AI hardware,” he says. “It remains to be seen, but a real possibility is that US AI labs could adopt many of these efficiency gains.”

Storm Uru and Clare Pleydell-Bouverie, co-lead managers of the Liontrust Global Technology fund, agree the emergence of DeepSeek will drive the demand for compute power.

Uru says: “DeepSeek-R1 – the first open-source model to employ chain-of-thought reasoning, and therefore to require 100-200x more compute (resources) than first-generation systems – has transformed both compute demand and hyperscalers’ capital-expenditure plans.

“Although DeepSeek is arguably cheaper to train than its closed-source rival, OpenAI’s o1, and remains less costly to run, the shift to ‘thinking’ models still entails vastly greater compute requirements than earlier, non-reasoning versions. The net result is a broader range of practical AI use-cases and a structurally higher level of aggregate compute demand, not lower.”

'An active manager’s paradise’

Uru described the past couple of months as “an active manager’s paradise”, with the managers taking advantage of the opportunities presented by the market volatility.

“Companies tied to the AI trade have faced a double whammy of negative sentiment this year; first, with the release of DeepSeek, and secondly with Trump’s announcements of tariffs. In both instances, the market overreacted and created attractive entry points for companies whose earnings power has little altered – or indeed strengthened – from the start of the year,” Uru says.  

The pair have actively been adding to leading infrastructure businesses – the companies providing the hardware and software needed to create AI-powered solutions – that were hardest hit by the post Liberation Day sell-off, such as Nvidia, Broadcom (NASDAQ:AVGO) and Amphenol (NYSE:APH).

Pleydell-Bouverie noted: “Just three weeks ago Nvidia was trading on forward earnings multiple of 19x, versus Procter & Gamble (NYSE:PG) at 23x. The former is growing its earnings 80% year on year, the latter 2%. These are the types of dislocations we have taken advantage of.”

Axa Investment Managers’ Dowey and his co-manager Matt Ward are also positive on the outlook for Magnificent Seven member Nvidia, having ramped up the fund’s holding in the stock when they assumed control of the portfolio in January, making Nvidia the fund’s largest position at 7%.

“There was some scepticism surrounding Nvidia after its stellar run, but we maintained a positive view on the company because it’s executing well on its strategy,” Ward says. “What sets Nvidia apart is its pace of innovation; the gap between Nvidia and its competitors is getting bigger.”

However, Hyun Ho Sohn, portfolio manager of the Fidelity Global Technology fund, is avoiding “high-momentum mega-cap AI semiconductor names” such as Nvidia and Broadcom.

“Despite recent share price weakness, my view remains cautious as I see limited upside given elevated capital expenditure levels among hyperscalers and AI-focused start-ups reliant on external funding,” he explained.

Instead, Sohn has active positions in Amazon (NASDAQ:AMZN), Workday (NASDAQ:WDAY), and Adobe (NASDAQ:ADBE), which he says have resilient balance sheets, and names Taiwan Semiconductor Manufacturing (NYSE:TSM) as a top active position in the fund, praising its strong pricing power.

“As a leading semiconductor foundry with control over the most advanced manufacturing technologies, TSMC benefits from a wide economic moat supported by scale and reliability.”

On Trump’s protectionist stance, Sohn notes that similar tariff risks emerged during the President’s first term.

“Both broad and targeted measures disrupted supply chains,” he says. “While some businesses experienced cost pressures, resilient companies successfully navigated the environment. We should not forget that technology companies and their management teams are used to dealing with constantly changing market dynamics in a fast-paced sector always trying to beat obsolescence risk.”

• Ian Cowie: tariff truce boosts technology investment trust duo
• Does this fund currency trick boost returns?

Political tailwind?

Liontrust’s Pleydell-Bouverie remains optimistic on the outlook for tech stocks despite the shadow of impending tariffs, following Trump’s trip to the Middle East earlier this month.

“Although the situation is still fluid, and we expect distinct restrictions to remain on China, there has been a marked pivot in the Trump administration's stance towards exporting AI technology in the past week,” she says.

“The dismantling of Biden’s AI diffusion rules and the deal struck with Saudi Arabia to build out 500MW data centre capacity powered by Nvidia chips underscore a significantly softened stance. The US is now encouraging the world to use the US tech stack, not restrict it.”  

The question investors continue to wrestle with is which part of the AI stack – or ecosystem – presents the best opportunities.

Axa IM’s Dowey and Ward have been adding to infrastructure software, an AI-adjacent sector.

Dowey says: “The companies that provide the infrastructure to manage and analyse companies’ data will capture revenues from AI earlier than the application software that's right at the end of the end of the process.

“These are not household names but are smaller companies in the background, but if they do good things with AI you can really move the dial in terms of stock returns. We’ve bought companies such as JFrog (NASDAQ:FROG), Cloudflare (NYSE:NET), Nutanix (NASDAQ:NTNX), GitLab (NASDAQ:GTLB), and Confluent (NASDAQ:CFLT) this year, which we believe will be good exposure for the next leg of the AI journey.

Richard Clode, portfolio manager on the Janus Henderson Global Technology Leaders team, says that while the team is still focusing on the infrastructure build-out phase of GenAI, such as semiconductors and networking companies, he foresees a pivot from training AI to inferencing – utilising trained AI models to draw conclusions from data.

“We want to be more exposed to the inferencing names in networking, to companies that are helping the hyperscalers develop their own chips, for example Amazon’s Trainium and Google’s TPUs,” he says.

“In summary, we are thinking of pivoting our hardware exposure and being more selective, having more on the platforms and then starting to have some selective software exposure, potentially in the next few years.”