The real story of artificial intelligence (AI) is the multi‑year progression of capital flows that will define which companies lead, when they lead and why.
AI is reshaping the opportunity set across markets, but from our perspective the real story is not the technology itself – it’s the multi‑year transformation unfolding in distinct waves, each with its own leaders and beneficiaries. We see capital flowing through the AI ecosystem in sequence, creating cascading opportunities as adoption deepens. Active managers who anticipate these rotations – and position ahead of them – are better placed to capture value as leadership shifts along the AI value chain.
Multi-year waves of AI investment
We focus on the following four waves to understand where capital builds next and which companies have the structural advantages to benefit across multiple phases (Figure 1). Today, we sit squarely in the first wave and are beginning to transition into the second. Early signs of the third wave are emerging in software, while a fourth wave of economy‑wide productivity gains sits on the horizon. Understanding where we are now and what comes next can help you align portfolios to the full scope of the opportunity.
Figure 1: The waves of AI investment growth
Phase 1: Infostructure and physical buildout
We remain early in the infrastructure phase. This foundation is the most visible and capital‑intensive stage. Hyperscalers alone are on track to invest more than $500 billion in 2026 to expand AI‑ready infrastructure1. This unprecedented scale reflects the intensity of model development and the physical demands required to train complex models.
We see this phase having another two to three years of strong momentum. The physical requirements of AI, the rise of model‑as‑a‑service offerings and the global race to build sovereign AI capacity drive this growth. Core exposure to infrastructure remains relevant, but leadership will evolve as the buildout matures.
Phase 2: Tools, data and cloud platforms for scaling
As infrastructure expands, companies move from early experimentation to scaled deployment. This transition fuels the second wave where tools, data architecture, cybersecurity and cloud‑based AI platforms play a defining role.
Snowflake and MongoDB stand out for their ability to help companies make data usable, structured and secure, and ready for AI. At the same time, cybersecurity becomes even more critical. Palo Alto Networks and CrowdStrike are positioned to protect data and AI models as enterprises expand their digital footprint.
Hyperscalers remain central at this stage because they lower the barriers to AI adoption. They provide pre‑trained models, managed services, guardrail tooling and integrated developer environments. We expect this wave to run three to five years as enterprises operationalise AI across business units.
Phase 3: Early enterprise AI integration
We now see early evidence of AI integration within enterprise software. Unlike infrastructure, which scales quickly once capacity is built, enterprise adoption is methodical. Companies test, validate and pilot before they roll out at scale. But the traction is real.
Microsoft is embedding Copilot across the productivity stack and cloud services. ServiceNow integrates AI into workflows that redefine service delivery, HR, IT operations and customer support. Datadog layers AI into observability to help companies troubleshoot increasingly complex systems.
- Walmart uses AI to assist with roughly 40% of new code development and elevate digital customer experiences3.
- Citigroup deploys AI to handle thousands of client and servicing questions.
- Pfizer accelerates clinical development with AI-driven data analysis.
We expect this phase to unfold over three to four years, accelerating as the first wave of adopters begin to demonstrate measurable ROI.
As with any major technological transition, moments of pullback or shifting expectations are to be expected. The path from AI investment to monetisation is measured in years, not quarters. Companies must gather data, build and test models, and determine deployment strategies, while their customers simultaneously vet those offerings for security and utility. In our view, these pauses serve to separate short-term experimentation from durable, enterprise-wide value creation, and we see recent pockets of caution as part of that healthy refinement process rather than a reversal in direction.
Phase 4: Productivity enabled value creation
The fourth wave is the most powerful and most difficult to quantify in the near term. Here, AI becomes an economic force rather than just a technology category. We estimate that AI currently contributes 10-20 basis points (bps) of productivity per year. We believe this could expand to 50-150bps annually as adoption scales, workflows adapt and companies fully integrate AI into core operations.
The differentiation here will be significant. Companies that move effectively from pilots to enterprise‑wide deployment will widen their competitive lead. These organisations leverage proprietary data, redesign processes and demonstrate real productivity gains. We expect the largest dispersion in performance across industries during this phase.
Why hyperscalers stand out across all waves
Most companies will benefit from one or two waves. However, we see the hyperscalers – Amazon, Microsoft and Google – as unique multi‑wave beneficiaries who participate in every stage:
- Wave 1: Building infrastructure and gaining internal efficiency.
- Wave 2: Powering AI development through tools, models and cloud platforms.
- Wave 3: Embedding AI into their own software and consumer experiences.
- Wave 4: Capturing long‑term cloud consumption as customers’ AI success compounds.
The breadth of their exposure, strong balance sheets and deep customer relationships position them as long‑term anchors in AI‑driven growth.
The bottom line
AI is a succession of waves rather than a single investment theme. Each wave creates new leadership and opportunities for active investors. At the same time, the path forward will include periods of adjustment as companies refine where AI delivers sustainable value. But early evidence of AI’s impact on revenue growth, margin expansion and competitive differentiation – across use cases from code development to clinical research – reinforces why adoption is becoming less optional than essential. We believe these represent healthy normalisation rather than a challenge to the broader multi-year opportunity. The key is to understand where capital flows next and identify companies with the structural advantages to benefit across multiple phases.
The transformation ahead will be uneven, iterative and multi‑year, but the opportunity set is broad and expanding. As these waves unfold, we believe disciplined research, selective positioning and early recognition of leadership transitions will be essential to capturing the full value of this AI‑driven cycle.
