News

Rout or reset? What the AI sell-off means for tech trusts

The AI trade has been under pressure since the summer, and the autumn has turned into a rout. Since early October, Nvidia has fallen by about a third from its $239 high. On 5 November, the day after the US midterms, the Nasdaq Composite fell 5.5% and the Philadelphia Semiconductor Index fell more than 12%. A week later Jensen Huang, Nvidia’s co-founder and chief executive for 33 years, resigned after reports about his private life raised national-security questions.

Yet investors have not lost their appetite for AI altogether. On 19 November Anthropic, the developer of the Claude models, listed on the Nasdaq in one of the largest flotations on record. It raised $81.3bn at $83 a share. The shares closed the day at $97.17, 17% above the offer price, valuing the company at around $1.75trn on a fully diluted basis.

The Nasdaq fell 2.2% the same day, suggesting that investors will still pay up for AI businesses growing revenue at exceptional rates but they are no longer prepared to buy everything with an AI label.

It has been a painful few months for the tech-focused investment trusts. The share prices of Manchester & London (MNL), Allianz Technology Trust (ATT) and Polar Capital Technology Trust (PCT) are all down around 20% since 6 October. In October alone, PCT’s NAV fell 15.1% and ATT’s 13.2%, against falls of 9.8% in their respective benchmarks.

For three trusts that each more than doubled shareholders’ money over three years, the obvious question is whether this is a buying opportunity or the start of something worse.

What caused the AI sell-off?

The sell-off appears to have been caused by three separate shocks landing together. Firstly, money got more expensive after the Federal Reserve (Fed) raised rates in September for the first time since July 2023, to 3.75%-4%, and the 10-year Treasury yield went above 5% for the first time since 2007. AI infrastructure is capital-intensive and increasingly funded by debt, so it suffers most when the discount rate rises. Meta’s $97.5bn fundraise and Microsoft’s $25bn bond sale reminded the market how much of the build-out now relies on borrowed money.

Secondly, AI has become a political issue, with voters concerned over data centres’ environmental impact and their impact on rising power bills. After the US midterms, Democrats control both houses of Congress, and new governors who campaigned on restricting data centre approvals took office in states that account for around 45% of planned US capacity.

And finally, Huang’s resignation removed one of the people most closely identified with the AI build-out, along with his relationships with customers and with Washington.

Reports of hedge funds deleveraging and of forced selling made the declines indiscriminate. So much so, that recent good results from AMD, Arista Networks and Coherent were largely ignored.

Is the AI sell-off just a blip?

Although the higher cost of capital and the political risk look likely to last, the evidence on demand is that it is strong and growing. Nvidia’s results, published on 17 November, showed revenue of $115.8bn, up 103% year-on-year, with data centre revenue up 110%. Its guidance for the current quarter was around $130bn. The hyperscalers are seeing the same, with Microsoft’s Azure growing 48%, Google Cloud 103% and AWS 39%, and Microsoft says that demand still exceeds the capacity it has available. Whatever is driving the sell-off, it is not a lack of customers.

What is changing is who pays for the build-out, and on what terms. PCT’s manager, Ben Rogoff, argues that the data centre backlash may change who ultimately pays for infrastructure, not whether it gets built. The hyperscalers have already pledged to cover the full cost to power their data centres. That protects volumes but squeezes returns, and some projects planned for 2028 may slip.

Higher borrowing costs add to the pressure. With 10-year Treasury yields at about 5%, a hyperscaler that funds new capacity from its own cash flow has a clear advantage over a debt-funded cloud provider. Leadership within the AI trade is therefore likely to shift towards companies with strong balance sheets and proven abilities to monetise their offerings.

The market is also becoming more selective on valuation – with the initial landgrab over, investors are no longer paying for AI exposure for exposure’s sake, instead wanting to see returns on the capital being spent. Anthropic’s debut suggests investors will still pay for growth, but not at any price.

The last big AI scare is a useful comparison. In January 2025, after the Chinese start-up DeepSeek released a model trained at a fraction of US costs, Nvidia fell 17% and lost $589bn of market value in a day, a record for any company. The Nasdaq fell 3.1%, and the fear was that cheaper AI would need fewer chips. Within about three weeks Nvidia had recovered around 90% of the loss, because the hyperscalers raised their capex plans rather than cut them, and by late June 2025 it was at a record high, despite the tariff sell-off in between.

In that case, the question was about demand, and it was quickly answered in the spending figures. This time the pressures are the cost of capital and political consent. Those are unlikely to be resolved as quickly. Investors should expect more volatility, especially in power infrastructure and second-tier AI names, until there is clarity on data centre deals and on the Fed’s next move.

How exposed are PCT, ATT and MNL?

All three trusts hold semiconductor leaders Nvidia, TSMC and Broadcom near the top of their portfolios. How much each has been impacted depends on concentration, benchmark discipline and hedging.

 PCTATTMNL
Market cap£6.2bn£2.0bn£318m
Concentration91 stocks; top 10: 43.7%53 stocks; top 10: 57.9%Compute, networking and memory accounts for 74.1% of NAV
Downside tools6.7% cash and Nasdaq puts1.0% cashShorts and derivatives
Three-month NAV total return (at 19 November)(15.9%)(11.4%)(21.5%)
Three-month share price total return (at 19 November)(13.2%)(10.5%)(12.1%)
Discount to NAV (at 19 November)4.1%4.3%13.9%

Source: QuotedData website

MNL is the most direct bet on AI hardware. Manager Mark Sheppard has diversified away from the two-stock portfolio that once had more than 60% in Nvidia and Microsoft. Even so, around three-quarters of NAV sits in compute, networking and memory, with Nvidia, TSMC and Broadcom together making up more than a third. That makes MNL the most sensitive of the three to the AI capex cycle in both directions. Its NAV fell furthest, and it should gain most if Nvidia’s results and any data centre deals rebuild confidence. It is a high-conviction thematic vehicle with a big dividend (a 4.8% yield), not a diversified technology fund.

PCT’s and ATT’s underweight exposure to Nvidia cushioned the shock from Huang’s departure. ATT’s October NAV still lagged the Dow Jones World Technology index by 340bps. In commentary dated 11 November (written after the 5 November rout but before Huang left) ATT manager Mike Seidenberg said that the weakness “appears driven more by shifting investor sentiment than a deterioration in underlying fundamental factors”. His response was to adjust the portfolio rather than retreat, buying SpaceX, Lumentum, Dell and Tesla in October and selling STMicroelectronics. Seidenberg also pointed to opportunities broadening into software, cybersecurity and IT services. Before the sell-off he described hyperscaler debt as both an opportunity and an “existential threat”. That focus on balance sheets and valuation discipline looks well suited to a more discerning market.

PCT went into the sell-off with cash and Nasdaq puts, which were there for a macro or oil shock. As Rogoff acknowledges, the Nasdaq puts do not help when money rotates out of AI within a flat market. PCT’s overweights in AMD, Lam Research, Seagate, Sandisk and Lumentum, built in a year of strong outperformance, were among the areas hit hardest. PCT’s technology team has since cut second-tier AI and power-infrastructure names and added to mega-caps with strong balance sheets. At the end of October, PCT was still 7.0% ahead of its benchmark for 2026, and its discount narrower than its 12-month average.

PCT’s manager sees the sell-off as a “clearing event”, flushing forced sellers out of crowded positions and resetting valuations. That leaves the strongest AI beneficiaries priced for much lower growth than they are likely to deliver, he adds, provided demand holds and data centre deals emerge in the coming months.

The AI trade is not over, with demand for computing still growing faster than supply. However, with an elevated cost of capital and data centres needing political consent, companies that can fund the build-out from cash flow and can show a return on it are likely to be rewarded by the market.

That favours active stock pickers, like PCT. With AI’s long-term structural growth drivers still intact, we think that, for long-term investors who can tolerate volatility, the sell-off offers an opportunity and is more of a reset in expectations rather than a reason to abandon the theme altogether.

Richard Williams
Written By Richard Williams

Senior Analyst

Leave a Reply

Your email address will not be published. Required fields are marked *