AI moves deeper into the economy - April Trust Update

Key events

• Global equity markets rallied +6.9% (all returns in sterling terms) in April as Middle East tensions eased
• Semiconductor stocks surged +34.7%, driven by robust AI-related capital spending and strong company earnings across the technology sector
• AI model capabilities advanced dramatically, with 12 new model releases in 37 days, while hyperscaler capital expenditure estimates for 2026 reached $751bn
  

Market review

Global equity markets rallied in April, as geopolitical tensions in the Middle East showed tentative signs of easing. Global markets¹ increased 6.9% while the US² rose 7.6%, both reaching new highs. European markets³ underperformed on a relative basis but nonetheless recovered from oversold levels, gaining 4.4% (all returns in sterling terms).

Strong April returns came despite the ongoing restrictions on passage through the Strait of Hormuz, through which a significant portion of the world's oil supply passes. The blockade marks the largest energy supply disruption since the 1970s. The impact has been mitigated so far by the release of strategic oil reserves, increased production in other regions, and supply chain adjustments. Sentiment recovered from late March after US and Iranian officials signalled a willingness to negotiate and bring an end to the conflict. For now, a fragile ceasefire remains in place while both sides continue to negotiate terms for a lasting deal.

Macroeconomic indicators have remained relatively supportive so far. Job growth has been volatile year-to-date (YTD) but is still resilient. The US economy added 178,000 jobs in March, rebounding from a downwardly revised decline of 133,000 in February, when a strike in the healthcare sector weighed on employment.

The war in the Middle East has had a near-term impact on inflation, however. The Consumer Price Index – the standard measure of inflation – indicated that the annual inflation rate in the US increased 3.3% in March, primarily driven by higher energy costs. This marks the highest level since May 2024, and a sharp increase from 2.4% in both February and January. However, core inflation, which excludes food and energy, increased more moderately to 2.6%.

Central banks typically look through energy price shocks and, at the end of April, the Federal Reserve (Fed) kept the federal funds rate, the key interest rate it uses to manage the economy, unchanged at the 3.5-3.75% target range for a third consecutive meeting. The decision was not unanimous, however, as three Federal Open Market Committee members (the group of officials who vote on US interest rate decisions) objected to the language in the statement that suggested the central bank would eventually resume cutting rates, while Governor Miran voted to lower interest rates by 25 basis points⁴.

At his Senate Banking Committee hearing, Fed Chair-designate Kevin Warsh indicated a preference for a more restrained Fed communication strategy, advocating the elimination of forward guidance (the practice of signalling future policy intentions to markets) fewer routine press conferences and reduced public commentary to enhance policy clarity. He also called for a stricter inflation framework centred on price stability and expressed a preference for alternative inflation measures. He highlighted the disinflationary potential of AI-driven productivity gains, which he suggested could support lower interest rates alongside a smaller Fed balance sheet.

Technology review

The technology sector significantly outperformed the broader market in April, rebounding strongly after the pullback in March, as investors began to look through the near-term macro and geopolitical uncertainty and refocus on the notable strengthening in AI fundamentals over recent months.

AI model progress continued apace with a remarkable 12 new model releases in 37 days, reflecting the ongoing intensity of the AI race

AI model progress continued apace with a remarkable 12 new model releases in 37 days, reflecting the ongoing intensity of the AI race. This included Meta Platforms' Muse Spark, Anthropic's Claude Opus 4.7, DeepSeek v4 and OpenAI's GPT 5.5, beating previous benchmark scores at a variety of tasks. Crucially, these models demonstrated continued returns to compute – meaning that investing in more processing power continues to produce meaningfully better model performance – as the scaling laws remain intact.

What the largest technology companies are spending

Big tech earnings season proved supportive to AI capital expenditure (capex) once again, with an incremental c$33bn in spending added to 2026 budgets.

Q1 earnings season proved positive for the sector.

Alphabet (Google’s parent company) delivered robust numbers at 19% growth on a constant currency basis⁵, with very impressive 63% year-on-year (y/y) Google Cloud Platform growth. Search remained solid and the company also raised its full-year 2026 capex guidance to $185bn at the midpoint, up from $180bn. AI adoption continues to accelerate as Google disclosed an acceleration in Gemini application programming interface token growth to 60% quarter-on-quarter (q/q), up from 40% in Q4. Google also offered its first view on 2027 spending, noting that 2027 capex will "significantly increase" versus 2026, suggesting a continued constructive backdrop for AI infrastructure investments through next year.

Amazon delivered Amazon Web Services growth of 28.5%, a little light of elevated expectations, but an acceleration of Amazon Web Services backlog⁶ — to 49% q/q, excluding the $100bn Anthropic deal, which is indicative of strong future growth. Retail beat expectations, and profits were boosted by strong Amazon Web Services margins of 38%, ahead of expectations. The AI infrastructure business remains supply-constrained, with chief executive Andy Jassy noting revenue would have been higher if they could have met the additional demand. Meanwhile, Amazon's Trainium custom chip business now has more than $225bn in revenue commitments, a new disclosure this quarter.

Meta Platforms posted strong earnings, growing Q1 revenue 29% y/y in constant currency due to solid pricing and impression growth of 12% and 19% respectively. Cost discipline drove a strong operating income beat. Meta Platforms guided for 25% growth at the midpoint, which was slightly below investor expectations, and its operating expense guidance was not reduced despite the recent news of a 10% layoff. Encouragingly, the recent release of Meta Platforms' Muse Spark AI model showed progress out of the Meta Super Intelligence lab and capex guidance was raised by $10bn at the midpoint for AI-related spending.

Microsoft Azure and Microsoft 365 both met expectations and were guided to accelerate next quarter, although to a lesser degree than Amazon Web Services or Google Cloud. A new OpenAI agreement removes uncertainty associated with its previous clause around artificial general intelligence while Microsoft's revenue share payments to OpenAI end, as do Microsoft's and OpenAI's exclusivity agreements. Microsoft Copilot adoption is picking up and could accelerate further as the company incorporates newer higher-performing AI models. Capex guidance was raised well ahead of expectations as consensus estimates for the financial year 2027 rose from $170bn to c$230bn, due to very strong customer AI demand and the need for Microsoft to invest more aggressively in its first-party AI model and product offerings.

Semiconductors

Bellwether Taiwan Semiconductor Manufacturing Company (TSMC) grew 40.6% y/y with gross margins at 66.2%, well ahead of expectations. The company raised its full-year revenue guidance from 'close to 30%' to 'above 30%', while guiding capex to the high end of the previously announced range of $52-$56bn. TSMC will add additional 3 nanometre (nm)⁷ capacity to a node that is at full ramp, an unusual move and a reflection of exceptionally strong demand.

Foundry peer Intel posted one of its best quarters in many years. Most importantly, the company is benefiting from strong demand for its x86 server central processing units (CPUs) which are playing an important role in the orchestration layer of agentic AI, meaning AI systems that can plan and execute multi-step tasks autonomously. As such, pricing is moving higher just as previous supply constraints are easing. Intel's 18A manufacturing process – its most advanced chip-making technology – yields in the foundry business are ahead of internal projections.

In memory, Samsung Electronics (Samsung) and SK Hynix both posted another set of strong results as the industry benefits from better pricing and a continued customer focus on the strategic importance of memory. SK Hynix revenues almost doubled y/y and delivered another record quarter of operating profit. DRAM pricing, the most common type of high-speed memory used in computers and servers, was up mid-60% and NAND (memory used for longer-term data storage) up mid-70%, reflecting the memory shortage. In addition, the company commented that High Bandwidth Memory demand – a specialised, high-speed memory stack used in AI chips – will exceed supply for the next three years.

Western Digital also continued to benefit from increased AI data creation, driving demand for large-capacity hard disk drives and posted very solid results

Both SK Hynix and Samsung are seeing multiple customer requests for long-term agreements as customers now see memory pricing and supply uncertainties as key business risks. Samsung flagged a shortage of NAND supply and expects overall memory supply shortages to worsen due to strengthening demand. It is becoming more apparent, as we have expected, that this memory cycle is likely to be longer and with a larger amplitude than previous cycles.

SanDisk also reported another record quarter with earnings per share up 60% above analysts’ expectations and gross margins guided to 80% next quarter. Similar to Samsung and SK Hynix, the focus of the earnings was on new long-term NAND memory supply agreements.

Seagate Technology Holdings (Seagate) and Western Digital also continued to benefit from increased AI data creation, driving demand for large-capacity hard disk drives and posted very solid results. Seagate saw particular strength in gross margins, up more than 10% y/y, with incremental margins well ahead of expectations. This has been driven by the combined tailwinds of pricing discipline, mix shift to higher-capacity nearline drives⁸ and improving new product cost leverage from its heat-assisted magnetic recording technology, which allows significantly more data to be stored on each drive. Management raised the long-term annual growth target from low-to-mid teens to a minimum of 20%.

In the semiconductor capital equipment sector KLA reported a strong quarter, albeit the guidance was short of some investor expectations. Demand remains very strong and the issue is more around space constraints in fabrication plants, which will ease into the second half of 2026 and 2027. KLA cited a record advanced packaging process-control business that has now been raised to $1bn in calendar year 2026. ASML Holding's comments regarding 2027 cutting-edge extreme ultraviolet lithography machine – the most advanced chip-printing equipment in the world, essential for producing the smallest and most powerful chips – capacity of c80 disappointed the market.

Liquid cooling provider Vertiv Holdings is seeing a robust ramp in Q2 while seeing its order pipeline accelerate. 800V data centres become more important in 2027, offering the potential for higher content. Commentary on pipeline and orders set the company up well for 2027.

Power and infrastructure

GE Vernova also delivered strong results, pulling in its $200bn backlog to 2027 from 2028, and saw pricing move up 10-20% q/q. Of note, $2.4bn of data centre-specific electrification orders were booked in Q1 2026, more than all of financial year 2025. Meanwhile fuel cell maker Bloom Energy grew revenues 130% and doubled its revenue guidance. The company reiterated its expectation of double-digit cost reductions annually, creating a path to a cost-effective alternative data centre power solution.

In the cables and connectors space, Corning posted a solid quarter, delaying upgraded guidance to its investor conference in early May, which also saw NVIDIA make a strategic investment in Corning alongside a multi-billion dollar prepayment to secure future supply of fibre and photonics⁹. Cables and connector company Amphenol saw record orders of $9.4bn, up 78% y/y, accelerating from last quarter. The strength was organic growth across most end markets but exceptional in the information technology datacom market for products used in AI applications.

Finally, retail trading platform Robinhood Markets delivered a softer-than-expected Q1 earnings report amid slowing retail trading activity and subdued crypto volumes and take rates, although prediction markets and new products continue to show rapid growth. Investments in the newly introduced Trump accounts, which Robinhood Markets provides the technology for, are strategically important for long-term growth but modestly dilutive to margins in the near term.

Outlook

AI model capabilities have advanced significantly since late 2025, with the launch of Google's Gemini 3, Anthropic's Claude Opus 4.5, 4.6 and 4.7, and OpenAI's GPT-5.4 and 5.5. The step change in the trajectory of model performance and significant improvements in agentic capabilities – the ability of AI to work independently on complex, multi-step tasks without human intervention – mean the length of time a model can work autonomously before running into trouble is increasing rapidly. The first GPT-3 model could work autonomously for less than one minute; the latest Claude models can now do so for more than 12 hours. With that comes the complexity and economic value of tasks that can be addressed.

Moreover, accelerating model progress and hyperscaler capex growth have helped address various AI concerns from last year. OpenAI has produced strong models and raised $122bn of equity to help support a highly ambitious infrastructure build. Spot pricing of NVIDIA GPUs (graphics processing units) has increased as denser models can produce more tokens for the same hour of accelerated compute, at the same time as those tokens produced by more sophisticated models are more valuable. Hallucination rates have fallen such that models can address complex, mission-critical tasks.

Most importantly, frontier large language models have continued to get better, with Anthropic's Claude Mythos Preview, recently released in limited capacity, sufficiently capable at identifying zero-day code vulnerabilities that Treasury Secretary Scott Bessent and Fed Chair Jerome Powell convened an emergency meeting with the heads of the largest US banks to discuss systemic cyber risk. OpenAI's GPT-5.5 reaches a similar level of cyber performance and can also solve multi-step cyberattack simulations exceeding human capabilities.

As model providers have seen an explosion in demand, this has been reflected in significant upward revisions to hyperscaler AI capex. Analyst estimates for 2026 AI hyperscaler capex have reached $751bn, $80bn higher than estimates at the start of Q1 earnings season, and growing 83% y/y, according to Goldman Sachs. Increased capex guidance has been reflected in rising estimates for AI infrastructure companies as demand for infrastructure is accelerating beyond the industry's ability to supply it.

Strong AI investment reflects insatiable customer demand. Alphabet noted "unprecedented internal and external demand for AI compute resources." This is reflected across cloud hyperscaler backlog growth, which accelerated materially in Q1, growing 112% y/y to reach $1.26trn, up from $904bn in Q4 and $747bn in Q3, according to JP Morgan. Looking forward, while strong AI capex revisions have driven AI infrastructure fundamentals and stock performance, consensus is now modelling 31% AI capex growth in 2027 to c$1trn, with room for further upward revisions if AI progress and adoption continue to trend higher.

We have taken profits in selected high-growth stocks and subsectors, particularly optical-related stocks, but with AI fundamentals remaining robust we have reinvested much of the proceeds into adjacent areas such as memory, where considerable supply and demand imbalances remain

As AI fundamentals continue to strengthen, the market has refocused on the AI story, which has led markets higher as AI beneficiaries enjoyed their best month on record, according to Morgan Stanley. Estimates for AI infrastructure-exposed S&P 500 stocks have increased by 20% for 2026 and 27% for 2027 so far this year, driving upward revisions to overall S&P 500 estimates, likely accounting for more than 40% of earnings per share growth this year, according to Goldman Sachs. This is driving Q1 S&P 500 earnings per share to be on track for the strongest growth rate in five years, even adjusting for one-off factors, but also one of the narrowest rebounds on record, with only 25% of MSCI US stocks outperforming.

In our view, we are only in year one of the enterprise AI adoption cycle. We are shifting from early chat-based AI to a new period where long-horizon autonomous agents can do real work, unlocking meaningful economic value. As model capability has advanced and task length has increased, AI demand in terms of token consumption has scaled non-linearly and AI revenue has inflected with it. The most striking proof point of this is the unprecedented growth of Anthropic's annualised revenue, rising from $9bn at the start of 2026 to $30bn by early April, and reportedly as much as $44bn by early May, according to SemiAnalysis.

The challenge for investors is to balance considerable macro and geopolitical risks against a rare moment of discontinuity tied to the arrival of agentic AI. If a US/Iran deal leads to the reopening of the Strait of Hormuz, accelerating AI adoption and capex growth will likely continue to dominate and support markets, with a potential deflationary growth tailwind from AI productivity gains. However, a prolonged closure or further escalation of hostilities remains a genuine risk and could cause a more material economic and market setback or a shorter period of elevated volatility.

After a strong period of YTD performance, we have taken profits in selected high-growth stocks and subsectors, particularly optical-related stocks, but with AI fundamentals remaining robust we have reinvested much of the proceeds into adjacent areas such as memory, where considerable supply and demand imbalances remain. Positioning remains constructive and strongly pro-AI, although the Trust continues to own out-of-the-money Nasdaq put options to soften the excess market sensitivity that comes with our growth-centric approach should a more meaningful setback occur. It should be noted these are not intended to protect absolute returns and would have limited benefit if a short-term risk-off period for markets leads to more widespread AI profit taking.

1. As per the MSCI All Country World Net Total Return Index

2. As per the S&P 500 Index

3. As per the DJ Euro Stoxx 600 Index

4. One basis point equals 0.01%

5. Adjusted to remove the effect of exchange rate movements

6. The value of contracts signed but not yet delivered

7. A measure of chip circuit size where smaller means more powerful

8. High-capacity drives designed for data centre storage

9. The technology that transmits data using light rather than electrical signals