A Trillion-Dollar Bet: What the BIS Sees in AI Capital Spending

On September 10, 2026, at the Global Fintech Fest in Mumbai, Pablo Hernandez de Cos, the General Manager of the Bank for International Settlements (BIS), stated that the five largest technology firms are expected to invest more than $1 trillion in AI throughout 2025 and 2026. This projected spending is part of a broader trajectory where global AI investment is expected to grow from approximately $500 billion today to as much as $4 trillion by 2030. According to Reuters syndication, this massive influx of capital is creating new layers of financial risk that could impact global stability.
The Scale of the AI Investment Cycle
The current scale of AI-related capital expenditure represents a significant departure from historical tech investment patterns. While current global investments hover around the $500 billion mark, the projected jump to $4 trillion by 2030 suggests a period of intense, concentrated spending. This expansion is largely driven by a handful of dominant market participants. The five largest firms' spending is a large part of the picture the BIS describes.
This concentration of capital creates a specific type of systemic risk. When a small number of entities accounts for the vast majority of investment, the stability of the entire sector becomes tethered to the success of those specific firms. If the capital-intensive infrastructure being built does not yield the expected economic returns, the subsequent pullback could be more severe than a localized market correction. This risk was previously identified in the BIS Annual Economic Report from June 2026, which warned that supply bottlenecks and intense competition could lead to overinvestment, mirroring past boom-bust cycles.
| Investment Metric | Timeframe | Estimated Value |
|---|---|---|
| Current Global AI Investment | 2026 | ~$500 billion |
| Top 5 Tech Firm AI Spend | 2025-2026 | >$1 trillion |
| Projected Global AI Investment | 2030 | ~$4 trillion |
Mechanisms of Opaque Financing and Debt
A critical concern raised by the BIS involves how this massive capital expenditure is being funded. While equity plays a role, there is an increasing reliance on debt and private credit. This shift toward debt-based financing is often "opaque and interconnected," meaning the true level of exposure and the relationships between lenders and borrowers are difficult for regulators to monitor. This lack of transparency is a primary driver of potential financial instability.
The risk is compounded by a growing mismatch between capital expenditure and cash flow. The BIS flagged capital expenditure running ahead of cash flows as a pressure point. If the growth in revenue from AI services fails to keep pace with these debt obligations, the risk of defaults or credit contractions increases significantly.
"Should the returns to AI disappoint, a pullback in investment could turn today's capital expenditure boom into a bust."
Why It Matters: Pablo Hernandez de Cos
The Risks of Circular Financing
The BIS also pointed to circular financing among chipmakers, hyperscalers and AI companies, in which the same money can pass between firms in the chain. In my analysis, this can make demand harder to read: gross investment figures may overstate independent demand if part of the money circulates within a closed network. The summaries reviewed give no figure for how much financing is circular.
The Productivity-Capital Gap
The fundamental question facing the AI economy is whether the productivity gains realized by users will justify the massive capital outlays. Current studies cited by the BIS suggest that task-based gains can range from 10% to 65% in specific sectors. These gains are most notable in areas such as coding, consulting, and professional writing. However, a gap remains between these task-level improvements and the systemic economic productivity required to offset trillions in spending.
There are two primary scenarios regarding this gap:
- Scenario A: Productivity Alignment. AI technology achieves widespread integration, and the 10% to 65% task gains scale into massive cost savings and new revenue streams across all major industries, comfortably covering the costs of the infrastructure.
- Scenario B: The Capital Trap. While specific tasks show improvement, the cost of the hardware, energy, and cloud services required to perform those tasks remains so high that the net economic benefit is marginal. In this scenario, the capital expenditure is not an investment in future growth, but a sunk cost that fails to generate a return on investment.
This uncertainty is reflected in the elevated valuations currently seen in the market. Valuations may reflect optimism about Scenario A, but the sector's financial stability depends on which scenario actually unfolds.
Implications for Stakeholders
The current trajectory of AI investment requires different responses from various economic actors:
- For Executives: Decision-makers should prioritize the measurement of actual ROI on AI implementation rather than participating in the broader capital expenditure trend. The focus must shift from "capability acquisition" to "cost-efficiency gains" to ensure they are not caught in a deflationary loop.
- For Investors: Transparency is paramount. There is a critical need to look beyond gross investment figures and analyze the quality of the revenue. Investors should be wary of companies whose growth is heavily dependent on financing arrangements that resemble circular loops.
- For Policymakers and Regulators: The "opaque and interconnected" nature of private credit and debt used for AI infrastructure is a primary monitoring target. Regulators may need to look closer at the balance sheets of both the hyperscalers and the specialized hardware providers to ensure that the interconnectedness does not create a systemic "contagion" risk.
To determine whether the current boom is sustainable, several key indicators must be monitored over the next 24 months. First, the ratio of AI-related capital expenditure to operating cash flow must be tracked; a rising divergence is a warning sign. Second, the emergence of default rates in private credit specifically linked to AI infrastructure would be a leading indicator of a bust. Finally, the transition of AI from a "capital expenditure" item to a significant "revenue generator" in the earnings reports of the non-tech sectors will be the ultimate test of the technology's economic viability. The OECD AI Incidents Monitor also logged the BIS warning. Whether the trillion-dollar bet pays off will depend on whether adoption and returns keep pace with spending.
Sources: thestar.com.my, oecd.ai, bis.org.