ABSTRACT

Payment fintechs, acting as lenders, possess a potential solution to weak debt enforcement because of their ability to deduct a part of a merchant’s digital sales towards loan repayment. Analyzing payments processed by an Indian fintech company offering sales-linked loans, we find that some borrowers discontinuously reduce sales flowing through the company immediately after the loan disbursal to circumvent repayment and strategically default. Using credit bureau scores sourced independently and the spatial and temporal heterogeneity in cash availability generated by a cash-crunch episode, we find that competition from other lenders and cash limits the effectiveness of this enforcement technology.

ABSTRACT

Data security is often viewed as a defensive cost. But in today’s data-driven economy, can it also create value? We propose a novel measure of “data innovation complementarity”, capturing the extent to which firms have structurally integrated data-security expertise into their broader innovation processes. We show that such integration unlocks cross-domain knowledge spillovers, fueling firm-wide innovation and growth. Exploiting U.S. Data Breach Notification Laws as a quasi-exogenous increase in the perceived importance of data security, we find that only high-complementarity firms realize these gains. A calibrated real-options model shows that value creation arises only above a complementarity threshold, turning defense into opportunity.

ABSTRACT

The technology revolution is transforming firm and industry dynamics, yet the roots of firm dominance in the modern economy remain unclear. Is industry dynamism driven by compute capabilities (AI), access to data, or the interaction between them? We develop a dynamic model in which firms gain knowledge from raw data using AI, but face “informational entropy”: without sufficient AI, more raw data leads to information overload and has negative returns. The model has two key predictions: (1) improvements in AI (compute) disproportionately benefit data-rich firms; and (2) access to processed data substitutes for compute, increasing industry dynamism and reducing market concentration. We confirm these predictions using novel data from 2000–2023 and two exogenous shocks: the 2006 launch of Amazon Web Services (AWS) and the 2017 introduction of transformer-based architectures. Our findings suggest that regulating data usability, not just AI models, is essential to preserving competition in the modern economy.

ABSTRACT

Open Banking initiatives aim to make payment data interoperable across institutions. I examine the value of this interoperability in lending using data that link borrowers’ payment histories with non-bank small-business loans in India. Payment data improve lenders’ ability to screen and monitor loans beyond traditional sources, including credit bureaus. These information sources are complementary because they specialize in different determinants of repayment. Subsample analysis shows that payment histories primarily reflect real-time repayment ability, while a natural experiment removing borrower discretion confirms that credit bureaus primarily measure willingness to repay. Shifting to payment-based screening benefits most borrowers but disadvantages those with poor credit scores and thin payment records, highlighting distributional trade-offs in Open Banking implementation.

ABSTRACT

Does AI make firms vulnerable or resilient to cyberrisk? To answer this, we develop a measure of AI intensity for US public firms using patents and 10-K business descriptions. Cyberrisk suppresses innovation for non-AI firms, reducing patenting by 25–30%, with the sharpest declines in data-management technologies. Frontier AI firms neutralise this effect: their patenting holds steady, and their valuations rise when cyberrisk is high. But not all AI capabilities provide protection. Firms that adopt external AI tools without building internal innovation capacity remain as exposed as non-AI firms. Cyberrisk acts as a tax on data-intensive innovation. Only firms with deep, internally developed AI capabilities can absorb it. Rising cyberrisk therefore widens the gap between AI frontier firms and the rest.

KEY POINTS

— Between 2021 and 2025, 28.7% of artificial intelligence (AI) firms’ investment deals (by deal value) involved a target company that was also an AI firm, while 55.2% of incoming investments in AI firms came from other AI firms.

— Of all the AI-to-AI investment deals between 2021 and 2025, 16.1% (by deal count) and 46.4% (by deal value) also involved commercial supply chain relationships between the investor and target firms.

— Circular investment relationships reflect key economic features such as the need to secure critical inputs and the presence of information asymmetries, yet they entail macroeconomic risks and increase opacity.

KEY POINTS

— Globally, the largest artificial intelligence (AI) firms are based in the United States, China, Chinese Taipei, Korea and the Netherlands.

— These global AI giants account for an increasing share of total market capitalisation, capital expenditure and revenues in their respective jurisdictions.

— For many AI giants – and particularly those from the United States and China – scope and scale go together as they expand the breadth of their activities into multiple layers in the AI supply chain.

ABSTRACT

Additional Tier 1 (AT1) instruments are designed to operate as going-concern capital. Nonetheless, in practice, the effectiveness of AT1 Instruments in fulfilling their primary role as going-concern capital is undermined by low trigger thresholds, the discretionary nature of activation and insufficient incentives for recapitalisation. Restoring this function rests on three elements: (i) sufficiently dilutive, variable and market-linked conversion as a principal loss-absorption mechanism; (ii) elimination of writedown and of discretionary going-concern triggers; and (iii) Common Equity Tier 1 (CET1)-linked triggers set high enough to support recovery. Regulators have to weigh such reforms against potential adverse implications for banks' funding costs, buffer usability and resolution funding.

KEY POINTS

— We develop a procedure based on large language models to help supervisors review prospectuses against capital rules. This procedure ranks potential divergences for their consideration. We apply it to Additional Tier 1 (AT1) prospectuses from European global systemically important banks.

— We validate the procedure on anonymised documents from Credit Suisse and Yes Bank, recovering rule-prospectus differences that later mattered in disputes over AT1 loss absorption.

— This procedure could support other supervisory comparisons, including reviews of national rules against the international standards they implement.

ABSTRACT

In this paper, we trace the geography and economic characteristics of firms that produce artificial intelligence (AI) products and services. Many economies around the world are evaluating their strategic priorities in AI, yet relatively little is documented about the global distribution of AI production. We construct a new database that identifies 1,246 AI-producing firms across 32 economies. We map these firms in each economy into the five layers of the AI supply chain: compute, cloud and related infrastructure, data tools, AI models and AI applications. The biggest markets for AI production are China and the US. Most economies specialise only in a few supply chain layers and many focus largely on compute. AI firms in all economies exhibit strong home bias in investment activity, with a focus on downstream applications. Finally, we find that venture capital inflows are strongly correlated with the presence and density of AI firms in a given economy.

ABSTRACT

The resolution framework for global systemically important banks has been over a decade in the making. The failure of Credit Suisse (CS) in March 2023 was its first major test. Authorities had a resolution plan in place but chose a different path amid financial stability concerns. They facilitated a takeover of CS by UBS, backed by public guarantees. Additional Tier 1 (AT1) bonds were written down in full; bail-in creditors, who would bear losses next under resolution, were left whole. We study how this episode reshaped bail-in credibility across Europe. Using bond-level data from 94 banks in 22 countries, we trace the repricing of AT1, bail-in, and senior debt over the subsequent year. AT1 spreads moved in line with jurisdiction-specific regulatory signals, while bail-in spreads and credit default swap subordination premia narrowed across the board, consistent with markets assigning a lower probability to bail-in. Lower-rated banks saw larger spread declines, and investor responsiveness to firm-specific disclosures fell, pointing to reduced market discipline. This evidence suggests the CS episode weakened bail-in credibility.

ABSTRACT

AI compute suppliers increasingly put capital at risk in developers through circular investments. Yet they require commitments from other developers, payable even if their projects do not scale. I show how both arrangements can coexist endogenously when privately known project ambition determines the project’s risk-return profile and its developer’s outside opportunity. Their interaction can make the maximum acceptable success payment under full information hump-shaped in ambition. Intermediate developers then have incentives to imitate both predictable and frontier developers. Commitments deter them from imitating predictable developers but make the contract intended for frontier developers more attractive to them. The supplier counters that incentive by combining circular investment with a higher success payment on the frontier contract. This preserves revenue from intermediate developers and can justify investment even when developers are not liquidity constrained.