The AI economy: Interconnected forces, feedback loops and speeds of change
This report examines the AI economy through a systems lens, highlighting mismatched speeds between exponential technology and linear infrastructure. It details foundational requirements, including a US$290 billion private investment in 2025. Recommendations for stakeholders emphasise navigating bottlenecks and redesigning organisations to capture value amidst evolving global competition.
Please login or join for free to read more.
OVERVIEW
At a glance
The research suggests that the rapid advancement of artificial intelligence will leave few unaffected. While generative AI is already transforming how people think and work, its future trajectory remains deeply uncertain. To address complex questions regarding job displacement, economic growth, and societal upheaval, it is necessary to view AI as an interconnected system involving physical resources, institutions, markets, and people. Risks and bottlenecks emerge primarily because different parts of this system move at vastly different speeds. While AI capabilities advance exponentially, the development of physical infrastructure and organisational adaptation occurs more linearly, requiring leaders to learn how to manage at multiple speeds.
Introduction
AI is positioned as a highly consequential technology, yet its spread and effects are subject to intense debate. Workers are concerned about potential mass job losses, with some forecasts suggesting that AI could eliminate up to half of entry-level white-collar positions within five years. Investors are focused on whether massive infrastructure spending will yield returns or result in excess capacity. Meanwhile, policymakers in China, the European Union, and the United States are following distinct regulatory paths. To understand these forces, the report proposes a general framework that reveals relationships between factors often considered in isolation, such as the link between infrastructure readiness and technological adoption.
Mismatched timelines: Bottlenecks, risks, and opportunities
History shows that general-purpose technologies, like electrification and the internet, take decades to reach their full potential because they require complementary innovations and organisational changes. AI faces a similar challenge where model capabilities double every few months, but physical infrastructure builds out over years. By 2023, demand for AI chips had already outgrown manufacturing capacity, leading to server lead times of up to 11 months. By early 2026, electricity supply and grid connections became significant hurdles; the median new US power plant now takes approximately five years to move from an interconnection request to commercial operation. Community opposition further constrains growth, with at least US$130 billion in data centre projects blocked or delayed in the United States during the first quarter of 2026 alone.
AI as a system: How the pieces fit together
The AI system consists of foundations, transmission, and outcomes. Foundations—including capabilities, R&D, physical resources, and institutional stance—determine what AI can achieve. Transmission describes how people and organisations adopt and adapt these capabilities. Outcomes represent the resulting economic and broader societal effects. These elements are linked by feedback loops; for instance, high-performing models attract the investment needed to expand infrastructure, while public concern or weak governance can slow deployment. A systems view helps identify that optimising one piece in isolation, such as faster chips, will not improve overall outcomes if the real constraint lies in grid capacity or workforce skills.
The value of a systems view
For business leaders, lasting advantage requires looking beyond mere efficiency. They are encouraged to build new products and markets rather than just automating existing tasks. Leaders must also give employees and communities a reason to support change by delivering visible benefits. Investors should follow bottlenecks to find opportunities, noting that global investment in data centres is expected to reach US$7 trillion cumulatively between 2025 and 2030. However, they must also account for correlated risks, as a slowdown in AI demand could simultaneously impact chips, power, and software sectors. Policymakers should focus on relieving specific system constraints and promoting competition while ensuring regulations are responsive to evolving evidence.
Appendix: The AI system in detail
Detailed metrics highlight the system’s current state. In terms of foundations, leading AI models by 2026 can reliably complete tasks that would take a human expert more than 16 hours. Global private AI investment reached US$290 billion in 2025. On the transmission side, while 89 per cent of organisations used AI in at least one business function in 2026, only 46 per cent had progressed beyond pilot stages, and just one-quarter had redesigned workflows. Economic outcomes show that US non-farm labour productivity grew at an average annual rate of 2.5 per cent between late 2022 and mid-2026. Innovation is accelerating in fields like medicine, where AI has shortened drug-discovery timelines by as much as 80 per cent. However, risks persist; for example, trust is low, with only 47 per cent of survey respondents trusting companies to protect personal data when using AI.
Conclusion
Historical evidence suggests the economic impact of new technologies is arriving faster with each wave. While steam power took a century to reach its peak impact, the internet took approximately 25 years. AI may compress this timeline further, giving businesses and governments less time to adjust. The trajectory of AI will not be determined by the technology alone, but by the interaction of all parts of the system. Understanding these interactions and mismatches is the primary task for those seeking to influence what happens next.