The calculus around artificial intelligence investment has entered a new phase. After years of exuberant spending pledges—Goldman Sachs estimates Microsoft, Amazon, Google and Meta alone will approach a combined US$1 trillion in annual AI-related outlays by 2027—Wall Street has pivoted sharply from applauding capacity expansion to interrogating how those sums will convert into earnings. The shift is measurable: private equity dealmakers in London report that investment committees are now running tougher screens on how protected wealth management firms are from AI disruption, with some processes paused after a technology stock sell-off in February. The question is no longer whether AI can grow, but which business models will survive its diffusion and where the profits will ultimately pool.

Goldman’s answer, circulated in a recent note, is that the hyperscalers will recoup their outlay not by selling more chatbots but by capturing a share of corporate spending currently directed at human labour. The bank identifies a profit pool of over US$60 billion in the global recruitment and staffing industry that could migrate toward AI infrastructure and software as companies automate processes. This logic is already visible in dealmaking patterns. Singapore’s Temasek, which holds stakes in Anthropic and OpenAI, aims to lift AI exposure to 15% of its portfolio while simultaneously increasing allocations to hard assets and private credit—asset classes it views as less vulnerable to AI-driven disruption. Asian fund managers speaking in Singapore described a strategy of investing in the “picks and shovels” of the AI buildout, such as liquid cooling and data centres, rather than betting on which application will dominate.

Regulatory divergence is now amplifying these investment choices. In Dubai, the Virtual Assets Regulatory Authority has positioned the emirate as a hub for digital asset firms seeking transparent, innovation-friendly oversight, a model that contrasts with the European Union’s harmonised Markets in Crypto-Assets framework. Meanwhile, Brazil’s effort to incentivise data centre investment through a special tax regime remains in legislative limbo after the enabling provisional measure lapsed in February, injecting caution into M&A negotiations even as a major data centre acquisition was announced in April. Across Africa, a different dynamic is unfolding: cryptocurrency adoption is being driven by utility rather than speculation. Stablecoins are used for cross-border payments and working capital, a pattern that Binance’s regional head describes as a blueprint for how digital assets create economic value.

Moody’s Ratings has flagged a further structural shift: the combination of autonomous AI agents and digital money—stablecoins, tokenised deposits—could transform these agents from advisory tools into economic actors capable of initiating payments and managing liquidity. The agency notes that this expands risk across financial, cyber and operational domains, requiring robust guardrails. For now, the immediate milestone is legislative: the fate of Brazil’s data centre incentive bill in the Senate, and the practical implementation of Europe’s MiCA regulation, will signal whether regulatory clarity can keep pace with the speed of capital reallocation.