Key events

  • Global equities rose in February despite political and geopolitical uncertainty
  • AI spending expectations rose sharply after recent company results, with the large cloud providers now forecast to invest $667bn in 2026 – c60% higher than last year – as the AI infrastructure buildout continues at pace
  • AI revenues are scaling rapidly, with Anthropic’s annual recurring revenue (ARR) run rate increasing from $9bn at the start of the year to $19bn by early March


Market review

Global equities extended gains in February, with the MSCI All Country World Net Total Return Index rising +3.4%, the S&P 500 returning +1.1% and the DJ Euro Stoxx 600 Index advancing +5.3%, supported by its greater exposure to asset-heavy sectors such as energy and industrials. The Nikkei 225 Index rose +11.3% following a snap general election in which Prime Minister Sanae Takaichi's Liberal Democratic Party secured a large majority and a supermajority for the ruling coalition. All returns are stated in sterling terms.

February was another eventful month. Trade policy uncertainty increased after the US Supreme Court struck down President Trump's tariff authority under the International Emergency Economic Powers Act. In response, Trump announced a new global tariff regime under Section 122 of the Trade Act, initially setting a 10% tariff on all countries, and signalling his intention to raise this to 15%, although this would require Congressional approval after 150 days.

Geopolitical uncertainty also increased, with US military buildup in the Middle East during the month culminating in co-ordinated US and Israeli strikes on Iran, raising concerns about broader regional conflict and potential disruptions in the Strait of Hormuz, a chokepoint for roughly 20% of global oil shipments. This heightened safe-haven demand prompted inflation concerns as the price of Brent crude, the international benchmark for oil prices, rose to a multi-month high.

In its January meeting, the Federal Reserve (Fed), the US central bank, left the federal funds rate – the interest rate at which banks lend to each other overnight – unchanged at the 3.5-3.75% target range, in line with expectations, after three consecutive rate cuts last year. The minutes of the Federal Open Market Committee meeting showed that Fed officials are divided over the future path of interest rates, reflecting a tension between the need to contain inflation and the desire to support the labour market.

Early March has, unfortunately, started on a more volatile note, with the price of Brent crude spiking to nearly $120 per barrel, its highest level since 2022, amid concerns over an extended conflict in the Middle East and partial closure of the Strait of Hormuz.

Technology review

Technology investors digested a flurry of AI-driven catalysts that continued to hurt sentiment across software and services. February saw further AI model and tool progress: OpenAI launched its GPT-5.3-Codex product, a new frontier coding model demonstrating strong benchmark performance and notable efficiency gains. Although a relatively minor model release, GPT-5.3-Codex is one of the first models trained on NVIDIA Blackwell Graphics Processing Units (GPU), underscoring sustained AI model progress – something we expect to continue with upcoming major model releases.

OpenAI confirmed it has 900 million weekly active users and 50 million paying consumer subscribers. The market reaction was mixed, with the market still concerned about the durability of spend on AI into 2027.

Meanwhile, peer Anthropic rolled out multiple updates to its Claude platform, including the Opus 4.6 and Sonnet 4.6 releases, as well as an expanded suite of Claude Cowork plugins and enterprise connectors, tools that allow AI to integrate directly with business software systems. These releases renewed fears over AI's role in replacing traditional software-as-a-service workflows and contributed to sharp selloffs in software and professional services stocks as investors reassessed long-term value risks in seat/user-based software and data-heavy research tools.

February also saw a record-breaking c$110bn funding round for OpenAI backed by Amazon, NVIDIA and SoftBank Group as well as continued large technology company infrastructure commitments including Meta Platforms (Meta)'s large-scale six gigawatt (GW) Advanced Micro Devices chip deal. OpenAI confirmed it has 900 million weekly active users and 50 million paying consumer subscribers. The market reaction was mixed, with the market still concerned about the durability of spend on AI into 2027. As long as model improvements continue, we expect further upside given the size of the opportunity that can be addressed by AI.

Q4 earnings season was characterised by strong fundamental results in companies exposed to AI infrastructure investment.

In the semiconductor sector, NVIDIA posted very strong results, growing +73% quarter on quarter (q/q), the third consecutive quarter of year-on-year (y/y) growth acceleration. The company guided to $78bn next quarter, which would represent +99% y/y growth and continued acceleration. NVIDIA expects to see q/q revenue growth through every quarter in 2026, which would exceed the $500bn Blackwell and Rubin revenue number shared with investors last year. Purchase commitments stepped up materially in preparation for the Rubin GPU ramp.

Competitor Advanced Micro Devices (AMD) also posted strong results, with data centre revenues growing +24% q/q from Instinct GPUs and extremely strong demand for EPYC server central processing units (CPU) and are seeing renewed interest for inferencing, the process by which an AI model applies what it has learned to make predictions or decisions. As such, AMD expects server CPUs to be up by double digits in Q1 versus typical seasonality of being down quarter on quarter.

Alphabet's results demonstrated that Search momentum of +17% y/y is intact as more AI functionality is brought into the product offering. Google Cloud Platform, the AI and cloud infrastructure business, grew +48% y/y, well ahead of consensus, and now has a $240bn backlog, which grew +55% q/q and more than doubled year on year. Google guided full-year 2026 capex to $180bn at the mid-point, considerably above the $130bn expected as the company seeks to meet extraordinary AI demand and suggested that $180bn may still leave them capacity-constrained. Google is also exploring a joint venture with a large investment firm to lease its own semiconductor chips to external customers. Investors grappled with downward revisions to the levels of free cashflow estimates (the cash a company generates after accounting for capital expenditure), the role of search advertising in an agentic1 commerce world and a markedly different capital intensity profile.

Amazon also guided to a major increase in capex: $200bn this year across AI, semiconductor chips, robotics and low earth orbit satellites, compared to expectations of c$170bn before earnings, a growth rate of +60% y/y, up from the +30% expectation only a few days before the results were announced. The quarter itself was mixed: Amazon Web Services grew +24% y/y, which marked an acceleration but looked unimpressive against Google's +48%. North American retail growth was only in line with expectations and although operating income margin was 50 basis points (bps2) ahead of expectations, international retail missed by a considerable amount.

In networking, Lumentum Holdings reported solid results and forecast +85% y/y growth. Two-thirds of this growth is driven by components and optical circuit switches where the backlog has now extended to $400m from three customers. In addition, the company stands to benefit from the upcoming co-packaged optics transition, a technology that integrates optical components directly onto semiconductor chips to improve speed and reduce energy consumption, potentially shipping in the second half of 2027. Peer Coherent also delivered strong numbers as data centre and communications growth accelerated to +11% q/q.

NVIDIA announced non-exclusive strategic partnerships with both shortly after month-end. These agreements, which include $2bn investments in each firm, aim to accelerate the development of silicon photonics, optical interconnects and a dedicated laser manufacturing capacity.

Software has been under significant pressure due to AI disruption fears, and our exposure is highly selective.

Arista Networks grew +29% y/y with gross margins of 63.4%. Importantly, Q1 was guided to +30% y/y and the full-year revenue outlook was taken up to +25% y/y, with the AI revenue target also raised from $2.75bn to $3.25bn. Arista now has two customers each representing more than 10% of revenues and expects to have a third or even a fourth in 2026, all for large-scale AI deployments. MACOM Technology Solutions Holdings (MACOM) saw broad-based strength across data centre, telecom and industrial and defence markets. Looking ahead, MACOM raised its full-year 2026 data centre growth outlook to 35-40% y/y, up from 20%.

In data centre infrastructure and power, AI demand supported strong results and a robust outlook. Siemens Energy reported a strong order intake and reiterated +12% full-year growth at the mid-point, while anticipating current favourable demand trends to continue. Bloom Energy cleared elevated buyside expectations, guiding to $3.2bn in revenue, +58% y/y. Its backlog grew +135%, underlining the breadth of demand. Longer-term margin commentary was also constructive and the company expects traction to continue with the very large cloud computing companies (hyperscalers). Rolls-Royce Holdings' results were also above the high end of expectations and the company upgraded its mid-term guidance, expecting operating profit of c£5bn versus £3.75bn previously. The company's power generation sales grew +30% in 2025 as power demand continues to structurally increase globally for AI data centres.

Software has been under significant pressure due to AI disruption fears, and our exposure is highly selective. Snowflake reported a solid quarter with product growth of +30% while Snowflake Intelligence customers doubled q/q to 2,500. Palantir Technologies also reported strong earnings. US commercial revenue grew +137% y/y with management guiding to at least +115% y/y growth in 2026. Government business remained robust with growth accelerating to +66% y/y in Q4, now representing 41% of total sales. In cybersecurity, Cloudflare saw revenue growth accelerate to 33.6%, total new annual contract value up +50% q/q, the fastest growth since 2021, and the company signed its largest ever annual contract value deal for $42m per annum. Guidance for Q1 and the full year was above expectations and remaining performance obligations were +48% y/y. Margins were under pressure, however, given a higher allocation to network costs which in the near term hampers operative leverage.

Shopify reported a strong quarter with above-consensus sales and guidance of mid- to high 20%. Q1 guidance was also above expectations with revenue set to grow at a low 30% y/y rate. Meanwhile, Reddit grew +70% y/y above buyside expectations and daily active users landed a touch ahead at 121.4 million, +19% y/y. Q1 guidance was strong with c53% growth expected and the company announced a $1bn share buyback. Finally, Spotify reported net additions of nine million, around one million ahead of consensus, and monthly active users rose by a record 38 million to 751 million. Gross margin expanded to 33.1% as price hikes outpaced net content costs and higher margin advertising stack yields and marketplace contributions helped as well.

Outlook

2026 has seen a meaningful rotation away from companies or subsectors perceived to have AI risk in favour of the secular AI infrastructure winners, those companies providing the physical building blocks of AI such as semiconductor chips, data centres and power infrastructure. Against this backdrop, investors have moved away from long-duration growth and asset-light business models towards value and cyclical stocks, heavier asset or real economy exposures. The AI defensibility premium has widened for businesses perceived to have high physical execution content, regulatory moats or hard to disintermediate asset bases. Strikingly, after the recent weakness, the technology sector now trades on 23x forward price-to-earnings — a measure of how much investors are paying for each pound of expected future earnings — and 1.1x the S&P 500 multiple, much like consumer staples.

The acute focus on AI disruption risk has coincided with growing awareness of an inflection in coding capabilities from Anthropic's Claude Code, its agentic software offering Cowork and industry plugins, as well as the release of Google's Genie 3 model. The Trust's application software exposure is negligible due to our belief that agentic AI combined with outcome-oriented pricing is the future of software, as opposed to headcount-based software-as-a-service solutions. The Trust has limited exposure to cybersecurity stocks, which will also be impacted, but still holds a limited number of infrastructure software companies we believe are set to benefit from accelerating enterprise AI adoption. AI should drive more code, more applications and more data, which should result in the need to store and analyse a great deal more data including more telemetry (the automated collection and transmission of data) for ongoing training and improvement purposes.

Meanwhile, AI capex expectations have increased significantly, breadth has improved and single-stock dispersion is high which should provide fertile ground for stock-picking. The rotation has also seen investors move away from the US at the margin and we remain significantly underweight the hyperscalers and the Magnificent Seven group of the largest US technology companies for now. Consensus estimates imply hyperscaler capex of $667bn in 2026, $127bn higher than forecast at the start of the Q4 earnings season and representing c62% y/y growth. This is expected to consume around 90% of hyperscaler operating cashflows, which brings incremental risk as the market debates the potential return on and motivation for these aggressive investments. Only those able to clearly demonstrate a return on this elevated spending will be rewarded.

In our view, 2026 will be the year AI model performance and capabilities improve so much that the ability to supply will be the differentiator

MIT economist Eric Brynjolsson has argued that the material downward revisions to 2025 US payroll growth in the context of strong real GDP growth suggested that US productivity growth reached c2.7% in 2025, nearly double the 1.4% annual rate for the past decade. This suggests "we are transitioning from an era of AI experimentation to one of structural utility" as the economy works through the productivity J-curve, the idea that productivity initially dips before rising sharply as new technology is adopted. The macroeconomic bull case here rests on AI resulting in greater productivity gains than disruption in the near term, leading to a disinflationary growth reacceleration accompanied by lower interest rates. Encouragingly, industries with higher AI adoption rates are showing slightly larger productivity growth over the past year in official US data, according to Goldman Sachs.

Enterprise monetisation is scaling at an unprecedented rate and has accelerated materially in recent months. Rapid advances in model capability, coding automation and capital investment are the driving force behind this. Claude 4.5 and Gemini 3.0, both released late last year, demonstrated a step-change improvement in AI model capabilities. The revenue acceleration at Anthropic since then is extraordinary, from $1bn annualised recurring revenue in December 2025 to $14bn in early February and $19bn by the beginning of March 2026. OpenAI’s annualised recurring revenue growth is equally staggering, with an increase from $10bn to $20bn in less than 12 months to the end of 2025 and already hitting $25bn today.

In our view, 2026 will be the year AI model performance and capabilities improve so much that the ability to supply will be the differentiator, favouring the larger vendors who have secured key components at scale years ahead. OpenAI's recent $110bn financing round has alleviated funding concerns that weighed on its supply chain until recently. Investors now await the next generation of frontier models from OpenAI, xAI and Meta, which will be the first to be trained on larger clusters of Blackwell hardware.

Our team of 12 dedicated technology and AI portfolio managers and analysts have been travelling extensively recently, with Nick Evans, Fred Holt, Lina Ghayor, Alastair Unwin and Nick Dumas-Williams in the US and Xuesong Zhao in Asia in the past two weeks alone. Encouragingly, the tone regarding AI demand and supply chain resilience remains robust. As soon as macroeconomic fears subside, we expect strengthening AI demand will return as a core factor in 2026 performance, both directly for technology and AI growth prospects and more broadly as AI progress becomes impossible for investors to ignore. Even without further AI model advancements, the significant progress made recently is driving accelerating enterprise adoption and strong growth in AI inference. However, we would also remind investors that volatility is a feature of new technology cycles and therefore should be anticipated.

We are obviously hopeful that the conflict in the Middle East will prove short-lived and containable, although for now it also remains unpredictable. Our position is that the AI race, as evidenced by capex, model progress and enterprise adoption, is intensifying. The Trust has no direct holdings in the Middle East and as such exposure is indirect, likely to manifest in potential demand weakness associated with higher oil prices, travel restrictions and a loss of consumer confidence, or potential supply chain challenges.


1. Agentic AI refers to an advanced AI system that autonomously takes actions, adapts in real-time and solves multi-step problems based on context and objectives

2. A basis point is a common unit of measure for interest rates and other percentages in finance; one basis point equals 0.01%