Positioning for AI progress amid tech-cycle volatility - June Trust Update

Key events

• US job creation continues to beat expectations, showing economic resilience despite a more cautious Federal Reserve
• Semiconductor and memory chip makers had a standout month
• Enterprise AI adoption remains at an early stage, with significant room for further growth as capabilities keep improving

Market review

Global stock markets increased modestly in June, with the MSCI All Country World Net Total Return Index up +0.8% and the US S&P 500 up +0.7%. Sentiment was dented by a more cautious tone from the Federal Reserve (Fed) following the first Federal Open Market Committee (FOMC) meeting chaired by its new head Kevin Warsh. This was partly offset by easing tensions in the Middle East and lower oil prices. European equities fared better, with the DJ Euro Stoxx 600 Index up +2.0% (all figures in sterling terms).

Economic data remained fairly supportive through the month. The US economy added 172,000 jobs in May, comfortably beating expectations of 88,000, following an upwardly revised gain of 179,000 in April. Revisions to the March and April job figures added a combined 93,000 jobs to earlier estimates, underlining how resilient the US economy remains.

The conflict in the Middle East pushed up inflationary pressures in the near term. The headline US Consumer Price Index (CPI) rose to 4.2% year-on-year (y/y) in May, largely due to higher energy prices - the highest reading since April 2023. Core inflation, which strips out food and energy costs, rose more moderately, to 2.9% y/y.

The price of Brent crude oil fell 18.1% in June to $73 a barrel after the US and Iran signed a Memorandum of Understanding, easing concerns over possible disruption to oil supplies through the Strait of Hormuz. While the agreement started a 60-day negotiation period covering Iran's nuclear programme and regional security, geopolitical uncertainty and market swings are likely to stay elevated.

The Fed left its main interest rate unchanged at 3.50-3.75% for a fourth meeting in a row in June, as expected. However, Warsh struck a more hawkish, cautious tone, reiterating the FOMC's commitment to bringing inflation back under control while removing forward guidance from its policy statement. Updated economic forecasts also pointed to higher inflation and weaker growth, with more policymakers expecting at least one further rate rise this year, reinforcing the view that interest rates are likely to stay higher for longer.

Technology review

The technology sector underperformed the broader market in June. The Dow Jones Global Technology Net Total Return Index (W1TECN) returned -0.8%, while the MSCI All Country World Net Total Return Index returned +0.8% (all figures are in sterling terms).

Large technology companies underperformed their smaller peers. The Russell 1000 Technology Index, which tracks large companies, fell -1.9%, while the Russell 2000 Technology Index, which tracks smaller companies, returned +2.9%. The Philadelphia Semiconductor Index (SOX), which tracks chip-making companies, returned +12.9%, while the NASDAQ Internet Index (QNET) and the iShares Software Exchange-Traded Fund (IGV) returned -6.0% and -9.4% respectively.

In short, we believe the level of performance and reliability required for AI to do 'real' work has been reached.

The market's attention was on SpaceX's stock market listing as well as the continued strong performance of semiconductor companies, particularly those linked to memory chips and central processing units (CPUs).

We took part in the SpaceX listing, the world's largest ever initial public offering (IPO), raising $75bn. We see it as a unique, vertically integrated business - meaning it controls most stages of its own supply chain - at the centre of several attractive, underpenetrated markets.

Turning to AI, demand continues to far outstrip supply, although several developments added to recent market swings ahead of the earnings season. On 2 June, President Trump signed an executive order creating a voluntary framework for up to 30 days of government review, before release, of the most advanced (frontier) AI models. Separately, on 12 June the US Commerce Department issued an export-control order barring foreign nationals from using Anthropic's newly launched Fable 5 and Mythos 5 AI models. These controls were lifted on 30 June after Anthropic made changes that addressed the government's concerns, restoring access worldwide. We remain alive to the risk that government intervention could slow the pace of AI development - an issue that, alongside growing local opposition to new data centre construction (a concern shared across the political spectrum, tied to rising electricity bills and fears about AI's impact on jobs), is likely to remain a hot political topic heading into the US midterm elections. However, we think it unlikely that the US administration will allow China or other nations to take the lead in the AI race.

June was a relatively quiet month for technology company earnings, though there were some key results from semiconductor and cybersecurity businesses.

Leading memory chip supplier Micron Technology (Micron) again beat market and investor expectations, growing revenue +346% y/y with gross margins - the proportion of sales left after production costs - now at 84.9%, driven by higher prices for its DRAM and NAND flash memory chips, in the low 60%s and mid 80%s respectively. Management said it has no visibility on when supply will catch up with demand and expects tight conditions to persist into 2028. Micron also announced 16 Strategic Customer Agreements - longer-term supply contracts, mostly for five years, with fixed prices or agreed price ranges - and expects to return all spare cash to shareholders towards the end of the year, once it passes the two-year anniversary of funding under the US CHIPS Act, having continued to generate significant free cashflow.

Broadcom's latest quarterly results were mixed. The company reported a strong quarter, with its AI semiconductor division growing +143% y/y and most measures beating expectations. However, despite guiding for c200% y/y growth in AI semiconductors next quarter, this fell short of some investors' hopes and others may have been disappointed the company did not raise its roughly $100bn AI semiconductor revenue target for its 2027 financial year (FY27). Gross margins for the coming quarter are also set to narrow as custom chip projects, known as ASICs, which are less profitable for the company overall, start to ramp up. That said, Broadcom announced its first chip developed with OpenAI, codenamed Jalapeno, and its position in computer networking remains extremely strong.

In cybersecurity, CrowdStrike Holdings saw net new annual recurring revenue (ARR) - a measure of additional yearly subscription revenue being added - grow +32% y/y, alongside strong profit margins. Guidance for next quarter was solid at 28%, and the company raised its full-year forecast for the same measure by 5% to 27.7% y/y at the midpoint, supported by a good pipeline of business, even though demand linked to the Mythos AI model is not yet fully reflected in the numbers. New AI products appear to be resonating well with businesses' AI security needs and the company has strong partnerships with Anthropic, OpenAI, Amazon Web Services and NVIDIA.

Cybersecurity peer Palo Alto Networks also posted better-than-expected results, with its Next-Generation Security ARR beating guidance by $20m to reach $8.1bn, up 60% y/y. Revenue grew 31% y/y to $3bn, ahead of expectations, with its network security business delivering its best quarter in years. Its integration of recently acquired companies CyberArk and Chronosphere appear to be running ahead of plan, with cost savings arriving 3-6 months early, and Chronosphere's recurring revenue up more than $100m quarter-on-quarter (q/q) thanks to a large customer migrating from a rival AI model provider.

Finally, Apple raised prices across its product range on 25 June, as outgoing chief executive Tim Cook cited rising memory chip costs as a "hundred-year flood". Mac prices rose 15-20% and iPad prices 15-25%. While iPhone prices were left unchanged for now, we expect to see increases when the iPhone 18 launches in September. Although consumer electronics make up only a small part of overall consumer spending, Apple's products are a highly visible sign of inflation and further price rises across the sector could follow if memory chip demand stays strong. This is expected to see both PC and smartphone unit sales fall 8-10% y/y in 2026, with a more marked slowdown in the second half of the year. The rising cost of components is also feeding through into higher prices for cloud computing which could ripple through many other parts of the economy that rely on it.

Outlook

A June 2026 report from research firm Exponential View found that AI application, model and infrastructure hosting revenues have reached a $175bn ARR. Measured from the start of 2023, AI is growing revenue three times faster than previous technology waves, including the internet (1995), mobile apps (2007) and cloud computing (2010). Each additional $1bn of industry revenue is arriving faster than the last and it now takes less than two days for an extra $1bn of revenue to be added, down from 180 days in 2023. Despite this extraordinary growth, AI revenue is still only equivalent to 0.4% of US gross domestic product (GDP, or economic output) and 0.8% of labour costs.

There have been concerns about the pace of revenue growth and the frenetic - some sceptics would say unsustainable - rate of adoption, driven by a 'token-maxing' approach where businesses use as much AI computing power (measured in tokens) as possible. Some high-profile adopters, such as Uber Technologies and Meta Platforms (Meta), have reportedly set limits on token use and pulled back from encouraging usage for its own sake. As adoption increases sharply, companies are managing rising AI spending more actively, using cheaper models, open-source models (including leading Chinese ones) and running models on their own computers rather than in the cloud, where possible.

The Trust remains well positioned to benefit from continued progress in AI and the potential upside from AI-related capital spending.

In our view, the returns generated by using frontier AI models remain very high. As industry analyst SemiAnalysis recently concluded: "Our work suggests that headlines are overblown, enterprises continue to spend and new use case/verticals for demand and token consumption are keeping the AI train moving forward at an aggressive pace." Data from Ramp, covering 70,000 US firms, also showed continued strength in demand: AI spending among the heaviest 1% of AI spenders still increased 14% month-on-month (m/m) in May - cost discipline might be being applied at the margin but total spending is still outpacing it.

AI adoption also remains at a very early stage: the top quarter of AI-adopting firms in Ramp's data allocate just 1% of their total spending to AI and median monthly spending across all firms is just $11 per employee (roughly the cost of a single enterprise ChatGPT or Claude subscription). The top 1% of firms most convinced by AI's potential (skewed towards technology companies) have seen spending rise sharply, to $7.45k per employee per month, so there is enormous room for the typical firm to grow as AI capabilities mature and extend beyond coding-related uses. SemiAnalysis believes coding-related uses make up more than 70% of combined recurring revenue for OpenAI and Anthropic, though these tools also empower amateur developers to build usable business applications - one reason we believe traditional software companies face a serious competitive threat.

The pick-up in enterprise AI adoption is clearly visible in the acceleration in revenue run-rate at Anthropic and other leading AI model developers so far this year, driven by advances in AI model capabilities, starting with Claude 4.5 and Gemini 3.0 in December 2025. Since then, lower error rates (from 'hallucinations' where AI produces incorrect information), exponential growth in how long AI can work independently, and agentic AI (AI that can carry out multi-step tasks on its own) have broken down the traditional relationship between users and individual queries, driving exponential growth in AI computing power used - albeit with a growing mix of Chinese and open-source models also in use.

In short, we believe the level of performance and reliability required for AI to do 'real' work has been reached.

We think it unlikely that the US administration will allow China or other nations to take the lead in the AI race.

The technology sector and the Trust have enjoyed a positive start to 2026, driven by strong earnings, fundamentals and rising AI-related capital spending, but more recently volatility has increased. This has seen AI-related shares caught on the wrong side of a broader shift in market momentum which picked up pace in early July. Recent concerns centred on the sustainability of current memory chip (and other component) pricing, made worse by news that Meta is considering offering AI computing power as a service to other companies. Initially, this was read negatively, suggesting Meta's own AI plans might be stalling and/or that it had built more computing capacity than it needed, potentially putting future capital spending plans at risk. However, these fears were partly eased by chief executive Mark Zuckerberg, who stated: "I don't know anyone in the industry who feels like they have too much compute...Meta is currently using all the computing power in its arsenal." Combined with the release of an improved AI model - Meta Muse Spark 1.1 - this led to a strong rebound in the company's share price and improved sentiment towards AI more broadly.

However, investors taking profits ahead of second-quarter earnings, combined with a more cautious mood ahead of the normally quieter summer months, led to a broader shift away from recent winners, some pain for retail investors in leveraged products and hedge funds reducing exposure - all adding to the rotation away from stocks that had performed well so far this year.

The Trust remains well positioned to benefit from continued progress in AI and the potential upside from AI-related capital spending. However, as we have cautioned in the past, we (and investors) should expect episodes of volatility that are endemic to new technology cycles, as we saw between 1995 and 1998 when a strong technology market was punctuated by seven selloffs of 15% or greater. While we are hopeful that Q2 earnings season should prove reassuring, we may see continued volatility over the summer, exacerbated by seasonally lower trading volumes and elevated macroeconomic and geopolitical risks.