AI adoption broadens beyond early winners - August Trust Update

Key points

• Global stock markets rose in August as confidence grew around lower inflation and supportive US monetary policy
• Strong results in Q2 earnings highlighted ongoing momentum in AI and data centre demand
• Robust AI investment continues, with companies expanding capacity to meet powerful, long-term growth opportunities
  

Market review

Global equities continued to rally in August, the MSCI All Country World Net Total Return Index gaining +0.3%. Regionally, the S&P 500 Index decreased -0.1% and the DJ Euro Stoxx 600 Index returned +1.2% (all returns in sterling terms).

Equity markets initially sold off due to softer than expected employment data in the US, before rallying as the market factored in a more dovish¹ stance from the Federal Reserve (Fed). Data indicated that the US economy is slowing but remains relatively resilient. The Consumer Price Index increased +0.2% month-on-month (m/m) in July, decelerating from +0.3% in June, matching forecasts and resulting in a relatively benign +2.7% year-on-year (y/y) increase in inflation. Having temporarily closed above 5% in May and July, the 30-year US Treasury bond yield declined to 4.9% in August, while the trade-weighted dollar index fell -2.2%.

At the end of July, the Fed left the federal funds rate unchanged at 4.25-4.50%, for a fifth consecutive meeting, in line with expectations, although two governors dissented in favour of a cut – the first such dual dissent since 1993. Policymakers observed that recent indicators point to a moderation in economic activity in the first half of the year, contrasting with earlier assessments that growth was proceeding “at a solid pace.” However, the Fed maintained a wait-and-see approach amid rising concerns that the ongoing trade war could undermine progress toward its 2% inflation target.

Later in August, equity markets rallied after a speech from Fed Chair Jerome Powell at Jackson Hole, who said “the shifting balance of risks may warrant adjusting our policy stance”, reflecting labour market concerns. Powell spoke with the Fed under intense pressure from the White House, with Trump critical and urging him to cut interest rates. Concerns about Fed independence have driven longer rates and term premiums higher, despite the prospect of lower yields near term.

Technology review

The technology sector slightly underperformed the broader market in August.

Large-cap technology stocks significantly underperformed their small and mid-cap peers; the Russell 1000 Technology Index (large cap) and Russell 2000 Technology Index (small cap) returned +0.8% and +5.3% respectively.

The NASDAQ Internet Index (+2.7%) outperformed the Philadelphia Semiconductor Index (+1.1%) while iShares Software fell -3.2% as growing fears of AI disruption saw the software sector face significant selling pressure.

Solid returns belied a bout of profit-taking and a mid-month pullback in momentum and higher volatility technology stocks. This weakness was driven by uncertainty surrounding the Jackson Hole symposium, alongside scepticism on AI’s returns on investment following a study by MIT and cautionary, perhaps misread, remarks from OpenAI CEO Sam Altman on AI valuations. Despite these fears, August’s tech results season once again highlighted broad-based strength across AI-related businesses.

August’s tech results season once again highlighted broad-based strength across AI-related businesses

AI bellwether NVIDIA posted another strong quarter against elevated analyst expectations, growing 56% y/y with gross margins at 72.7%. NVIDIA’s financial Q3 guide of $54bn is $7bn higher than in Q2 – one of the largest quarterly additions ever – noting the possibility of an extra $2-5bn in revenue from China should geopolitical issues be resolved.

Advanced Micro Devices (AMD) also delivered a strong quarter, driven by demand for graphics processing units (GPUs), servers and continued strong PC momentum. As with NVIDIA, the company did not reference any expected China revenue in Q3. Pure Storage saw strong product sales and stabilising subscription growth. The company started pilot shipments to Meta Platforms (Meta) during the quarter while strong outsourcing growth was another positive.

Amazon’s results were mixed, with strong retail growth of +12% y/y offset by underwhelming Amazon Web Services (AWS) growth at +17.5% y/y. Corporate operating profit was much better than expected. The company saw capital expenditure (capex) of $32bn in Q2 with similar numbers expected in Q3 and Q4, implying $115-120bn for the year versus previous expectations of $105bn. Amazon is still capacity constrained versus the AI demand it is seeing which explains the significant additional capacity in 2H25.

In software, Snowflake delivered solid numbers with quarter-on-quarter (q/q) net adds nearly double their seasonal norms and revenue growth accelerating to +32% y/y. Key performance indicators improved across the board, with guidance only modestly raised. Consumption trends remain healthy and strength was broad-based across product categories, with AI workloads primarily supporting greater core data warehouse usage.

In networking/optical, Arista Networks posted another very strong quarter with revenues +30% y/y, gross margins 65.6% and deferred revenue +92% y/y. Management raised its full year growth guidance to 25% and reiterated confidence in achieving >$1.5bn of AI networking revenues this year. The company also now expects to reach $10bn in revenue in 2026 versus $8.75bn implied this year.

In semiconductor capital equipment, KLA saw a strong quarter with revenue and earnings per share well ahead of consensus. Management guided positively into 2026, highlighting broadening demand from the likes of TSMC, Intel, Samsung and Rapidus.

Cloudflare saw revenue growth accelerating to +28% y/y, net revenue retention (NRR) rebounding to 114% and the strongest large customer growth in two years. The company is seeing early traction with its new AI Audit product which management sees as a potential foundation for agentic² AI transactions rails. Crowdstrike Holdings’ quarter looked softer at first glance but management guided to at least 40% net new annual recurring revenue (NNARR) growth in H2, ahead of expectations.

Internet platforms produced a series of strong results. Reddit’s revenues grew +78% y/y with earnings and margins well ahead of expectations, driven by broad-based advertising strength. Q3 guidance was materially above analyst expectations and implies further acceleration. Advertiser counts and average revenue per user both rose and traffic trends have stabilised after earlier Google-related headwinds. ROBLOX also beat elevated expectations with bookings +51% y/y and viral titles such as Grow-a-Garden driving record engagement. While guidance was conservative, the company’s ability to surface and scale new experiences appears to have improved significantly.

Tech research specialist Gartner expects traditional search volume to fall 25% by next year as generative AI chatbots and agents continue to gain share

Tencent’s Q2 results were ahead across most major segments, with broad-based gains across games, ads and FinTech, with gross margin at a record 57% despite elevated AI-related spend, reinforcing its position as an increasingly AI-levered platform. Shopify reported orders and revenue growth of +31% y/y, well ahead of expectations, with strength in Europe and newer verticals such as B2B. Guidance was raised, although margins could be pressured by increased marketing spend in H2. Management emphasises targeted, return on investment-driven investments and highlighted growing traction in new agentic shopping products. The agentic web is still in its infancy, but Tech research specialist Gartner expects traditional search volume to fall 25% by next year as generative AI³ chatbots and agents continue to gain share.

US meal delivery firm DoorDash increased gross orders +23% y/y, ahead of the previous quarter, with order frequency at all-time highs. International momentum via Wolt+ also helped drive upside and guidance for Q3 was comfortably above consensus.

Axon Enterprise also delivered another robust quarter with revenues +33% y/y and bookings +50% y/y, underpinned by demand for its AI Era plan and large multi-product deals. The company continues to demonstrate durable growth and significant upsell opportunity.

Outlook

The Magnificent Seven (Mag7) were generally perceived to be good early conduits for AI given their significant distribution and data advantages but, as with previous technology cycles, we now see a new more pernicious AI era unfolding as generative AI transitions from complement to substitute. The track record for once-dominant incumbents ‘crossing the chasm’ is not strong which suggests the favourable tailwind for active managers (in place since 1Q25) may sustain for some time.

We have perhaps already seen a harbinger of the complement-to-substitute dynamic in application software. Early in August there was a material negative inflection in software sentiment as mildly disappointing earnings reports were punished heavily while questions about the terminal value of SaaS (Software as a Service) businesses in an AI-first world gained traction. We remain sceptical that incumbent SaaS companies will prove good conduits for AI adoption and as such have reduced the Trust’s software exposure to the lowest levels in many years.

Recent AI capex announcements and upbeat longer-term projections suggest confidence in further AI progress. NVIDIA CEO Jensen Huang spoke after recent results of “$3trn to $4trn of AI infrastructure spend by the end of the decade” significantly higher than the near-term $1trn referred to at NVIDIA’s analyst day. The difference is primarily due to higher expectations for the build-out of ‘AI factories’ as AI permeates through the real economy.

Despite initially mixed reviews, the recent launch of GPT-5 has put the most advanced AI in the hands of ChatGPT’s 700 million weekly active users. The most powerful reasoning models have historically only been available to paying users, but that is now changing. Rather than just performance, cost and ease of use were two notable areas of improvement. An intelligent routing system now automatically selects the most suitable model (e.g. ‘fast’, ‘thinking’, ‘mini’) to handle each user request. To give a sense of the cost/efficiency improvements underlying this democratisation, Wharton professor Ethan Mollick has noted that GPT-4 initially cost $50 to process one million tokens (words) whereas today the cost is $0.14 using GPT-5 – which is also a far more capable model. The proliferation of reasoning models should spark greater consumer and commercial use as users gain awareness of the extraordinary, and growing, power of leading-edge models.

We believe there is plenty of evidence of tangible benefit from AI pilots and deployments across companies in multiple sectors

Altman’s comments about an AI bubble also pressured AI stocks mid-month. We believe his messaging was directed more at AI researchers who might be tempted away from OpenAI by large sign-on bonuses at, in his view, inferior competitors as well as to potential investors in OpenAI competitors – rival Anthropic is reportedly in the process of raising $10bn. A recent MIT study added to investor fears with the claim that 95% of attempts to incorporate generative AI into business so far are failing. While the study raised some interesting points, we believe there is plenty of evidence of tangible benefit from AI pilots and deployments across companies in multiple sectors.

Our experience from previous technology cycles, as well as the MIT study’s own detail, suggests many companies will still be in the exploratory/experimentation phase ahead of broader deployment. A recent Citi study used the results of many other corporate AI studies to identify $274bn of potential AI efficiencies in S&P 500 companies alone across call centre ($90bn), coding ($64bn) and knowledge summarisation ($120bn). Morgan Stanley recently suggested $920bn in long-term AI adoption benefit for S&P 500 companies.

The enormous size and wide dispersion of these estimates reflect excitement about the growing scope of activities AI models can address at human-performance level (or better) but also indicate how early we are in AI adoption. Indeed, the adoption of 'shadow AI' (i.e. workers using chatbots and other AI tools without the knowledge or formal blessing of their companies) is "already transforming work, just not through official channels", according to the MIT study. A Stamford paper published in August found that early-career workers (aged 22–25) in the most AI-exposed occupations have faced a 13% relative decline in employment, even after controlling for firm-level shocks.

We continue to believe that AI represents a general-purpose technology (GPT) that has the potential to drastically alter societies and economies. GPTs are a distinct category of technology which are widely adopted, spur abundant knock-on innovations and show continual improvement. Historical examples include the steam engine, electricity and information technology (encompassing PCs and the internet).

This should create a broad opportunity for the Trust, spanning companies with infrastructure and capex-related exposure as well as those implementing AI to deliver differentiated growth relative to their peers. AI investment will add up to $3trn in global data centre spending, not including related power investments, through 2028, according to Morgan Stanley. Yet demand is still outpacing supply, including and beyond data centres. This introduces secondary investment opportunities in power generation, grid expansion, cooling technologies and efficiency enablers which are all focus areas for us.

Of course, risks to our constructive stance remain. These include weak seasonality, geopolitical uncertainty (e.g. Taiwan), macro and market volatility as well as potentially new tariffs or export controls that could complicate the AI spending backdrop. Valuations are also not cheap, although in our view remain far from ‘bubble’ territory; during the dot.com period of 1999-2000, the technology sector traded at above 2x the market multiple compared to 1.3x today. In addition, we expect further bouts to disrupt the AI narrative, as per the 1995-98 period and expected features of new technology cycles and steep innovation curves.

Despite these risks, we remain constructively positioned and believe the AI cycle, as it enters a more pernicious phase for incumbents, is likely to require a more active and dynamic investment approach. The team is exceptionally busy, travelling extensively in the US and Asia following Q2 earnings season. Our meetings with many of the leading AI companies (including Nvidia, AMD, TSMC and OpenAI) and the AI supply chain have been reassuring both in terms of demand and our longer-term ‘AI maximalist’ view.

1. A stance that favours a central bank stimulating growth and lower unemployment through measures such as reducing interest rates and increasing money supply

2. 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

3. GenAI: a type of AI that can create new content and ideas including conversations, stories, images, videos and music