The current landscape of artificial intelligence is profoundly shaped by the "returns to scale of new model architectures," a dynamic that is increasingly influencing market concentration and investment strategies. This observation was highlighted by Gabriel, who recently stated on X, formerly Twitter, that he "hadn't fully realised the extent to which current AI market dynamics are driven by the returns to scale of new model architectures." This realization stemmed from a "great essay, and a very interesting interview" discussing these trends.
The concept of returns to scale in AI refers to the economic principle where increasing inputs (like computational resources, data, and human expertise) leads to a disproportionately larger increase in output, such as improved model capabilities and performance. Early research in foundation models, as far back as 2013, demonstrated that performance improvements could outpace the proportional increase in computational resources within certain ranges, suggesting significant advantages for larger-scale operations.
This inherent scaling property is now driving a massive "AI capital expenditure supercycle," with estimates suggesting between $5 trillion and $8 trillion could be required over the next five years to fund AI technologies and their enabling infrastructure. This unprecedented investment, as reported by PwC's 2026 outlook, is primarily directed towards data centers, energy, chips, and technology development, diverting capital even from traditional M&A activities in the short term. Goldman Sachs also notes that assumptions around silicon useful life, data center costs, and the build-out's composition and timing are critical determinants of this capital deployment.
The substantial investment required to achieve and leverage these returns to scale is contributing to "winner-takes-all dynamics" in the AI sector. The OECD highlights that while AI can lower entry barriers for some firms by automating tasks and reducing labor intensity, the high upfront investment for data infrastructure, computing resources, and specialized personnel can also entrench incumbent players. PwC's analysis of the M&A market reveals a "K-shaped" recovery, where large, technology-led deals are concentrated among well-capitalized buyers, with AI influencing approximately one-third of the largest corporate M&A transactions in 2025.
Companies are increasingly prioritizing strategic capital allocation towards AI, with a focus on acquiring capabilities necessary to deploy AI at scale. This includes investments in cybersecurity, data analytics, and specialized infrastructure. The rapid pace of AI development and the significant financial outlay required mean that a company's "AI readiness" is becoming a key driver of valuation, underscoring the strategic imperative to engage with these scaling dynamics.
