The dispersion between AI winners and AI losers widens

The AI trade is often framed as a single bet on a handful of high-profile companies but our conviction is broader and more structural. AI is a general-purpose technology that is still early in its adoption cycle, with multiple years of extraordinary growth ahead, meaningful scope for positive surprises in capability and adoption, and substantial dispersion between the companies that harness it and those disrupted by it.

We are constructively positioned as AI data continues to be robust and growth remains extraordinary. For example, OpenAI is expecting to generate $20bn of revenue in 2025, only three years after the launch of ChatGPT, while Anthropic is on course for $10bn ARR (annual recurring revenue) in only four years. We expect 2026 to bring further significant progress in the underlying technologies, growing social and economic benefit with rising enterprise adoption and a continuation of the rapid pace of growth for AI-native applications.

Adoption of AI is expanding, unevenly

Adoption expands as companies clear educational, governance and change management hurdles; per a recent UBS cross-sector CIO survey, 17% of organisations reported using AI “in production at scale”, up from 14% in March 2025. However, we expect adoption and scaled implementation to remain heavily uneven, with a small number of companies positioned to deliver the corresponding benefits from AI use, while many more will lag and face disruptive threats. According to 22V Research, only 14% of small-cap companies reference AI usage, compared with 44% of mega-cap companies. This suggests that AI is not broadly diffused but rather it is a scale-dependent investment.

AI changes competitive dynamics to create both compounding winners and accelerating losers

We have long spoken of our expectation that tech-like dynamics will spread to non-tech sectors and this extends to last-generation winners being poor conduits for an AI-driven environment. Prior winners including Mag7, software stocks and many previously perceived ‘high quality’ capital-light companies across information services, online marketplaces and others are now struggling amid AI threats. We expect a growing section of the market to face challenges as models become more performant and unveil new capabilities. This asymmetry is central to our confidence – the opportunity is not simply that AI grows, but that AI changes competitive dynamics to create both compounding winners and accelerating losers, often within the same industry.

We expect further bouts of market volatility in the year ahead and believe an active investment approach is critical. Even aside from current geopolitical and macroeconomic tensions, volatility is often a function of periods of rapid technological change. For investors, this matters because progress will not always look linear or follow neat generational timelines. There will be unexpected breakthroughs in the underlying technologies, new techniques can reshape cost curves and the market could misread these shifts as demand destruction rather than demand expansion.

Models continue to improve

The continued advance of frontier models is fundamentally positive for the long-term development of the AI ecosystem. Better capabilities tend to expand rather than cap demand. As models become more powerful and more cost-effective, new categories of applications become viable and adoption spreads more widely across sectors and geographies. 2025 brought us deep research agents, coding agents, search functionality and image-editing language models, tools that are rapidly becoming daily use for many. This is reflected in commentary from leading platforms: Google’s head of AI infrastructure recently highlighted the need to double AI capacity roughly every six months simply to meet current demand expectations.

The first generation of models trained on NVIDIA’s Blackwell architecture should arrive in the first half of the year, delivering better performance, new capabilities and significantly lower cost per token. We are hopeful that AI capabilities will positively surprise which should embolden leading AI labs to invest (even more) aggressively in the next generation of models and put upward pressure on AI capex.

As models become more powerful and more cost-effective, new categories of applications become viable and adoption spreads more widely across sectors and geographies

The continued demand for AI infrastructure can also be underpinned by significant growth in agentic traffic. Agents will benefit from a maturing ecosystem and protocols that facilitate their ability to execute increasingly complex tasks. Measures of agent development and deployment have inflected orders of magnitude higher since the release of agentic coding capabilities. As agents begin to ‘do’ rather than merely ‘assist’, agentic AI can scale in a way untethered from human labour.

Taken together, these dynamics underpin our confidence in the AI trade. The top-down case for sustained investment remains firmly intact; the bottom-up reality is accelerating capability, uneven adoption and widening competitive dispersion. To assess opportunity and risk, we look at every company through an AI lens: Do they recognise the opportunity? How are they adapting? Do they recognise whether AI is a threat or are they complacent?

With over 1,000 company meetings conducted during 2025, the strength and depth of our now 12-strong investment team should stand us in good stead to navigate these exciting times. We believe the market still underestimates the cross-sector AI investment opportunity and the potential for widening dispersion between ‘AI winners’ and ‘AI losers’, making an active approach vital.