Key points

• Equity markets experienced significant volatility as tariff developments have brought downward revisions to growth forecasts and upward pressure on inflation
• Fundamental AI progress continues at a fast pace, although near-term macro headwinds could bring challenges to the technology sector
• Investor sentiment and positioning are extremely negative, which could provide a tailwind to markets if tariff news flow improves

Market review

Global equity markets fell during March. The MSCI All Country World Net Total Return Index lost -6.4% and the S&P 500 Index declined -8.0%, while the DJ Euro Stoxx 600 fared better, down -2.3% (all returns in sterling terms).

Trump’s tariff announcements weighed heavily on US stocks, stoking fears of prolonged inflation, lower US growth and a potential global trade war.

Q1 was challenging for equity market returns: the S&P 500 posted its worst monthly return in two years in March, to end the quarter down -7.2%, while the Stoxx Europe 600 ended up +7.4% thanks to a meaningful fiscal regime shift in Europe in support of higher defence spending. The US dollar struggled, with the dollar index down -3.9% in Q1, while the euro was up +4.5% against the dollar. Notably, gold posted its best quarterly performance since 1986, returning +18.8%, US Treasuries returned +2.9% in Q1 as recession speculation rose, and the price of bitcoin fell -11.4%.

After a prolonged period of outperformance, the ‘Magnificent Seven¹’ fell 18.6% in Q1, while defensive sectors such as consumer staples outperformed. The FTSE and DAX indices hit all-time highs in a notable shift from ‘US exceptionalism’. European aerospace and defence stocks significantly outperformed as European countries agreed to spend more on defence, while Germany proposed reform of its constitutional debt brake to permit higher defence spending.

Previously announced 25% tariffs on Mexico and Canada, were eventually implemented on 4 March while China tariffs were raised from 10% to 20%. As a result, US one-year inflation swaps² moved up +72 basis points (bps³) in Q1 to 3.25%, the highest level in two years as investors weighed the inflationary impact of higher tariffs.

President Trump’s infamous ‘Liberation Day’ Executive Order on 2 April unfortunately unleashed further uncertainty and volatility. A baseline 10% tariff was set on imports from all countries from 5 April and much higher ‘reciprocal tariffs’ on around 60 ‘worst offenders’ under the International Emergency Economic Powers Act (IEEPA) from 9 April. Many individual country tariff rates were close to worst-case fears, especially for Asia.

The FTSE and DAX indices hit all-time highs in a notable shift from ‘US exceptionalism’.

If there were any positives, Canada and Mexico were not hit with additional tariffs and tariffs will not apply to steel, aluminium and autos where they have already been implemented, or to copper and lumber, which face ongoing Section 232 investigations. Sectors where Section 232 investigations are expected but have not yet begun (pharmaceuticals; semiconductors; critical minerals) will not face tariffs at this stage, and there are other exclusions. Another small positive was that these sectoral tariffs will not be ‘stacked’ on top of country-level tariffs (e.g. autos face a 25% tariff, but no additional tariffs based on where they come from).

An intensive round of negotiations is likely as countries and trading blocs seek to mitigate the impact of the reciprocal tariffs. This may bring further volatility as retaliatory measures will be included as part of these negotiations, but Canada’s success following a firm stance and the securing of the USMCA (US-Mexico-Canada Agreement)-compliant goods (perhaps c75% of goods) and energy products carve out from 25% tariffs, plus no incremental reciprocal tariffs so far, offers a potential model for others to follow. The 10% baseline tariff appears to be more permanent, but the fact that markets rallied on the (erroneous) headline that this constituted the entirety of the ‘Liberation Day’ announcement suggests investors are comfortable with a higher known tariff baseline.

Should all the tariffs remain in place, economists suggest this could present a significant headwind to US growth as well as core PCE inflation – although in practice some portion will likely be negotiated down. Federal Reserve (Fed) Chair Powell’s comments at the last FOMC (Federal Open Market Committee) meeting indicated a prevailing dovish stance and willingness to look through the ‘transitory’ impact of tariffs while policy is already ‘restrictive’, and Fed futures are currently pricing more than three 25bps rate cuts by the end of the year. Overall, the size and scope of the ‘Liberation Day’ announcement attests to the administration’s commitment to reordering global trade policy (and geopolitics), even if the details of implementation are still subject to President Trump’s will.

Post-month end update

Equity markets experienced significant volatility in early April as tariff headlines drove the tape. The VIX – a measure of market volatility – recently closed above 50 and the S&P 500 registered some of the largest intraday swings in history (8.5%, 7.2% and 10.8% on 7, 8 and 9 April) amid record trading volumes. On 9 April – the deadline for reciprocal tariffs to go into effect – following unsettling moves in the bond market and concerned commentary from JP Morgan Chase CEO Jamie Dimon, Trump paused the higher reciprocal tariff rates for 90 days on all countries excluding China (where the cumulative tariff was increased to 125%, effective immediately). The market rallied hard and the S&P 500 enjoyed its best day since October 2008, rallying +9.5%, while the NASDAQ 100 returned +12%, its biggest one-day gain since 2001.

Technology review

The technology sector underperformed during the broader market weakness in March, with the Dow Jones Global Technology Net Total Return Index (W1TECN) returning -11.2%, in sterling terms.

The selloff in March was relatively indiscriminate, the Philadelphia Semiconductor Index (SOX) declining -12.5%, while the NASDAQ Internet (QNET) and iShares Software (IGV) indices declined -10.7% and -10.9% respectively. The Russell 2000 Technology Index (small cap) fell -10.7%, while the Russell 1000 Technology Index (large cap) declined -12.9%, led by the Magnificent Seven (-12.4%). Small-caps struggled during Q1, with the Russell 2000 Technology Index falling -23.6%.

Beyond concerns about the possibility of a macro slowdown brought about by tariffs and DOGE (Department for Government Efficiency) disruption, the sector continued to be impacted by concerns around the sustainability of AI-related investments post-Deepseek. These were exacerbated by Microsoft’s retreat from several data centre projects during the month. However, this appears to be company-specific, relating to Microsoft’s decision to forgo additional training workloads from OpenAI. Recent meetings with companies in the NVIDIA ecosystem have been positive on the demand environment.

There were several notable off-quarter earnings reports in March. In the semiconductor sector, Broadcom reported encouraging results with revenue +25% y/y, above market forecasts, with strength in AI semiconductors and software more than offsetting weakness in non-AI semiconductors (which remain in a cyclical downturn).

Micron Technology announced strong top-line results and guidance, with data centre revenue +47% y/y and high bandwidth memory (HBM) revenue surpassing $1bn. However, gross margins were below expectations and while management did not raise full-year guidance for HBM revenue to be “multiple billions” of dollars, it continues to see demand accelerating.

Elsewhere in the semiconductor sector, Advanced Micro Devices was resilient during the month perhaps as a result of resetting investor expectations for AI growth in its GPU (graphics processing unit) business in 2025 and into 2026.

Software companies struggled in March as concerns about Federal spending and DOGE pressures added to ongoing macro pressures, particularly after Accenture*, a major systems integration partner for many software companies, reported that “many new [federal] procurement actions have slowed, which is negatively impacting our sales”. Channel checks elsewhere highlighted customer “uncertainty” but did not point to significant softening in pipeline generation or deal close rates. Oracle reported lacklustre results and next-quarter guidance, blaming server supply and power capacity constraints. However, AI training demand remains very strong, and management gave guidance for revenue to grow +15% in 2026 and +20% in 2027, above expectations, although investors remain sceptical given the recent execution.

Cybersecurity vendor Crowdstrike Holdings reported stronger than expected results across all key metrics. RPO growth accelerated to +42% y/y, driven by strong Falcon Flex adoption. Revenue growth was guided to decelerate to +22% y/y in FY26, reflecting the impact from the outage last year, but this was in line with expectations. Encouragingly the CEO expressed confidence in the company’s federal market position, despite concerns about the DOGE impacting spending.

Alphabet agreed to acquire cloud security startup Wiz for approximately $32bn in cash to support Alphabet’s security and cloud ambitions.

Elsewhere in the sector, Alphabet agreed to acquire cloud security startup Wiz for approximately $32bn in cash – 30x Wiz’s expected 2025 annual recurring revenue (ARR) – to support Alphabet’s security and cloud ambitions.

In the internet sector, Alibaba Group Holding (Alibaba) continued to benefit from the strength of the Chinese stock market in the wake of an improving macroeconomic outlook and more favourable view of China’s AI position post-DeepSeek. Alibaba Chairman Joe Tsai warned of a potential data centre bubble in the US, despite Alibaba announcing its own intention to spend $52bn on cloud and AI infrastructure over the next three years, more than it has spent over the past decade.

Outlook

White House tariff announcements and economic data released over the past month have added to downward revisions to growth forecasts and upward pressure on inflation expectations. Macroeconomic headwinds have shown up meaningfully in ‘soft’ data such as consumer confidence and business activity surveys but have so far had limited impact on ‘hard’ economic data such as employment and consumer spending. Economic data may now come in even softer than expected due to the ‘Liberation Day’ announcement and the market/tariff volatility impact driving incremental consumer and business uncertainty.

To coincide with the risk-off move in markets driven by tariffs and policy uncertainty, several recent datapoints have challenged the AI infrastructure investment theme and the sustainability of hyperscaler spending. The most important was Microsoft reportedly cancelling data centre leases and up to 2GW of capacity. This happened over the past six months, and the capacity was picked up immediately by Meta Platforms and Google. Microsoft “will continue to grow [data centre spending] strongly in all regions” but is “slowing or pausing some early-stage projects”. We believe the decision to “strategically pace or adjust our infrastructure in some areas” (as Microsoft puts it) is best seen as reflecting the diverging interests of Microsoft and OpenAI. Microsoft’s relationship with OpenAI has evolved as it moves away from training workloads to inference⁴ – where Microsoft believes the sustainable value will accrue over time.

In contrast, pursuing artificial general intelligence (AGI) is OpenAI’s stated mission which requires huge amounts of capital with highly uncertain returns. It likely made commercial and strategic sense for Microsoft to fund OpenAI’s training costs when they ‘only’ ran to billions of dollars (to which Microsoft’s $13bn AI revenue attests), but to deploy hundreds of billions of capital expenditure (capex) to fund (what looks like) a potential competitor in agentic⁵ AI does not – hence Microsoft’s decision not to take a leading role in OpenAI’s $500bn ‘Stargate’ project (although Microsoft retains a right of first refusal on OpenAI training work through 2030).

Prior to Liberation Day, the outlook for hyperscaler capex growth this year remained constructive given Meta Platforms, Amazon and Google have very recently given (above consensus) capex guidance, which has led to 2025 cloud capex growth expectations increasing to +32% from +20% coming into 2025 (per Morgan Stanley). These companies (plus Microsoft) are well able to afford further spending increases with c$240bn in combined free cashflow between them, even taking account of existing significant capex plans. Recent price action does, however, suggest the market is becoming sceptical that this spending is sustainable into 2026 and beyond. While the macroeconomic outlook has deteriorated, especially in the US, AI data points have remained strong.

On the longer-term question about whether hyperscalers will continue to spend, there is a lot of focus (understandably) on the return on investment (ROI) for this spending. Early data points are constructive (see below) and Morgan Stanley produced a comprehensive ‘GPU Economy’ report which looked at this ‘ROI question’ and found $45bn in total ‘AI revenue’ in 2024, growing to $1.1trn in 2028E (of this, Microsoft alone goes from $13bn in 2024 to $78bn in 2028E).

Near-term ‘AI revenue’ is not the only driver of AI investment. AI infrastructure investments are (perhaps to a large degree) defensive as hyperscalers see potentially existential threats to their businesses and offensive in terms of the pursuit of AGI. The AI story is so early and the improvement curve still so steep across three ‘scaling laws’ (pre-training; post-training; test-time) that it is impossible to evaluate how AI monetisation will actually look, any more than it was possible during early phases of electricity or the internet.

While the AI infrastructure story has become more complex post-DeepSeek and with the emergence of post-training scaling and test-time scaling vectors, cheaper, more powerful intelligence on demand will undoubtedly dramatically increase the number of activities to which AI can be economically applied, which will in turn drive more demand for compute and infrastructure.

LA-based movie studios suggested AI could reduce production costs by a third, with more limited need for set builds or onsite shoots.

Broader academic studies imply a 24% average productivity increase from using AI and company anecdotes imply gains of around 29%, according to Goldman Sachs. Recent data points have been very constructive regarding AI’s potential to lower costs across a diverse range of industries and use cases: LA-based movie studios suggested AI could reduce production costs by a third, with more limited need for set builds or onsite shoots (JP Morgan); Sky will reduce its customer call centre headcount by 2,000 as it adopts AI. The conclusion from Wells Fargo’s Bank And AI Summit Recap was unequivocal:

“Literally every presenter seemed to have more conviction versus a year ago given more advanced ecosystems, use cases, adoption, and line-of-sight on savings... presentations reinforced that AI is a game changer for efficiency, productivity, business models, and impact almost all businesses and employees”.

On the consumer side, OpenAI has reached 500 million weekly active users, up more than 100 million from February and adding more than a million users in a single hour. The company also confirmed a $40bn funding round at a $300bn valuation, led by Softbank. This is the largest private technology funding round in history and will “push the frontiers of AI research even further”. Sam Altman, OpenAI’s CEO, has referenced GPUs “melting” under overwhelming consumer demand for its new image generation capabilities (and broader demand) and called for anybody with 100,000 clusters to contact OpenAI if they have spare capacity.

It is important to remember that volatility is a normal feature of new technology cycles. There were seven >15% corrections between 1995-98 (before the 1999 ‘melt-up’), where the NASDAQ returned +354%. At the moment, ‘typical’ technology-driven volatility around the emergence of a new general purpose technology is being compounded by exogenous macro and political volatility. More importantly, despite greater complexity post-DeepSeek, the fundamental AI story has shown rapid progress this year: new models from OpenAI, Anthropic, Grok and Google are beating benchmarks; there has been great progress in reasoning models and test-time scaling (models perform better the longer they think about a problem); and the first examples of agentic AI (which can complete goals autonomously) have arrived.

It will take time for the implications of tariffs to be fully digested and reflected in downward revisions to earnings for the broader market (very few companies are likely to escape unscathed due to the indirect impact on consumer and CEO confidence and spending). The ‘Liberation Day’ tariff impact could delay AI investments in the very short term if buyers wait in expectation of lower tariff negotiations (or supply chain disruption occurs), although the US economy was on a solid trajectory entering the year and the worst of the tariff news is likely out.

Despite the near-term disruption, our base case remains that while the Trump administration is committed to its plan to bring production and jobs back to America, it does not wish to engineer a recession or inflationary scenario (not least since this would likely increase the deficit and hamper the large US Treasury issuance required this year). Investor sentiment and positioning are now extremely negative, which could provide a powerful tailwind for markets when we begin to see bilateral trade agreements and reduced tariffs (hopefully in the coming weeks). Ideally, we would like to have more clarity on US export controls and AI diffusion rules in this timeframe too (because the latter are due to take effect in mid-May).

The Trust makes use of NASDAQ puts⁶ to soften the downside beta⁷ of the portfolio in the event of a sharp market drawdown. These puts did not have a material impact in the first AI/networking-led move down but helped reduce the portfolio beta during the recent selloff, and made a meaningful contribution to relative performance after the month end. We have recently taken some profits in these NASDAQ put options, realising some gains and rolling into a new line further out of the money⁸to help soften the portfolio beta should market conditions/tariff news flow deteriorate further.

We also took advantage of recent market weakness to refocus the portfolio on our favoured names. While volatility – both market and fundamental – may persist, we remain as convinced as ever that AI represents the next general purpose technology (GPT) and one of the most important themes of our investment lives.

* not held

1. Apple, Microsoft, Alphabet, Amazon, NVIDIA, Meta Platforms and Tesla

2. An inflation swap lets investors exchange fixed interest payments for payments linked to inflation, helping protect against rising prices or speculate on inflation changes

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

4. The process of drawing an idea or conclusion from evidence and reasoning

5. 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

6. A put option grants the right to the owner to sell some/all of an underlying security at a specified price, on or before the option's expiration date

7. A measure of a stock's volatility compared to the market/ an index; the market/index has a beta of 1 with each stock rated at +/-1 in comparison

8. An out of the money put option on (e.g.) the NASDAQ 100 Index (NDX) allows a holder the right to sell the Index at a specified price before or on a certain expiration date