Militarised AI, Private Credit, and Iran War – Developing Economics

By Farwa Sial and C.P. Chandrasekhar 

Private credit markets are showing real signs of stress, with multiple major funds restricting withdrawals as investors struggle to exit illiquid holdings. The fears of investors in these funds, which explain the withdrawals, is driven by the success of AI, which, while driven by enormous capital spending financed in part by private credit, is perceived as disrupting the pre-existing software landscape, many of the creators of which had been financed with credit from these funds. These two dynamics are increasingly tied to military demand, with the US government encouraging private capital to build defence-linked AI infrastructure. The war on Iran is amplifying these trends by squeezing energy costs, tightening liquidity, and accelerating a shift wherein AI investment becomes less market-driven and more concentrated around state-backed priorities.

Around the 22 March 2026, two of the largest players in private credit, Apollo Global Management and Ares Management, dropped redemption gates on flagship retail credit vehicles, temporarily limiting and/or restricting investors from withdrawing their money. While investors had requested withdrawals of 11.2% and 11.6% respectively, both funds capped redemptions at 5%, leaving roughly half of requested capital locked in place.

The gating at Apollo and Ares is just one visible manifestation of broader strains across the roughly $1.8 trillion private credit market. On April 2, it was reported that Blue Owl capital had received redemption requests of upto $5.4 billion over the first quarter of 2026, with those requests amounting to 22% of its private credit fund and a much higher 41% of another of its funds target at software and technology firms. In response, Blue Owl announced a cap on redemptions of 5% of shareholder funds. Earlier BlackRock restricted withdrawals on its HPS Lending Fund, which stands at approximately $26 billion. Blackstone faced roughly $3.8 billion in redemption requests from its flagship private credit fund and stepped in with its own capital to help meet those withdrawals. Morgan Stanley saw around 11% repurchase requests in its North Haven Private Income Fund and Cliffwater honoured only about 7% of roughly 14% redemption requests.  

Private credit involves lending by funds that are ‘financial innovations’ that have proved to be popular and rapidly expanding vehicles in recent years. They have mobilised private investor capital to finance borrowers who find it difficult to access funds from regulated creditors, because they are seen as not adequately creditworthy or are even burdened by pre-existing loans from regulated creditors that they are finding difficult to service. Riding on easy access to liquidity, private credit funds step in to ‘fill the gap’ and lend to these risky borrowers at rates that promise their investors high returns.

One set of such borrowers consists of software firms built in the business landscape that preceded the surge in the development and deployment of Artificial Intelligence (AI). The AI boom, while driven by hype, has progressed enough to threaten the software domains inhabited and the business practices supported by these firms. The results are losses, falling valuations, and declining investor interest that are making it difficult for these firms to service their liabilities, including credit advance by private credit vehicles. That in turn affects the liquidity position and viability of these vehicles.

In this environment, investors are recognising that the growth of AI is the source of credit risk in the portfolios of private credit vehicles. The Apollo/Ares gating actions reflect liquidity strains arising because investors, perceiving that a worrisome proportion of the clients of these private credit vehicles are burdened with illiquid loans and are likely to experience defaults that would wipe out their capital, are demanding early redemptions.

Though, as of now, default rates on these loans are only around 2%, that figure is seen as understating the degree of underlying stress. Around 40% of private credit borrowers are now cash flow negative. Stress scenarios and the pattern of redemptions, indicate that default rates will likely rise significantly to the high single digits and even beyond in more severe cases.

In fact, a growing share of private credit clients are avoiding default through payment-in-kind arrangements, where interest is capitalised (or added to the original loan size) rather than paid in cash. This has triggered “rescue Payment In Kind” arrangements, or the practice in which the terms of loans are amended midlife to avoid default. These now account for roughly 6.1% of deals, up from 2.6% in 2021, according to Lincoln International data. Thus, the full extent of underlying stress is significantly higher than reported defaults suggest, and is increasingly being deferred rather than being actually resolved.  

In response, investors are seeking to exit these vehicles holding illiquid loans through early redemptions, creating a whole new level of illiquidity since redemption demands are rendering unsustainable the financial strategy of private credit vehicles that have built high-yield promising structures involving relatively illiquid, long-duration loans. 

The fragility these developments reflect is being addressed by finance capital and the American state through new manoeuvres in parallel areas that are the current hunting ground for yield-seeking finance capital. Within days of the Apollo-Ares gating decision, the US Army initiated exclusive negotiations with Carlyle Group and KKR to build large-scale AI data centres. The first proposed facility, a roughly $2 billion project at Fort Bliss, Texas, is on track to begin operations by 2027. The second, also valued at $2 billion, will be located at Dugway Proving Ground, Utah, and will be developed through CyrusOne (a platform backed by KKR and BlackRock) with completion expected around 2029.

These developments point to a deeper structural reality, that the US national growth model is increasingly organised around a triangular system between private credit financing, AI infrastructure, and military demand.  The financial and real economic fragility of which the private credit liquidity crisis is a symptom, is intersecting and converging with AI in structurally significant ways. While disrupting the software services universe, AI is absorbing large sums of capital for investments in chip production, data centres and related infrastructure (including in requirements such as land, energy sources, cooling systems, and computing capacity). A significant share of that investment is being funded with credit from private credit vehicles. It is now emerging that the hype around AI, driven by the surging demand for high-end chips produced by ‘market-favourite’ Nvidia and newer entrants like AMD, is now being questioned by the lack-lustre returns from deployment of AI by downstream businesses. While some firms do record revenue growth from AI use or deployment, the investment needed is proving to be difficult to justify for many others. So, the demand and revenue growth projections implicit in the rising valuations of AI firms themselves increasingly appear to be driven by hype rather than reasoned, evidence-backed judgements.   

But given the long-term nexus between finance capital and the US state, the AI boom is being propped up by military spending justified by the potential for AI deployment for defence purposes. The Fort Bliss and Dugway Proving Ground investments are illustrative of this. Given the enormous capital needed to set up AI infrastructure such as data centres, private credit, which has become a favoured site for extraction of financial profits, is emerging as a financing conduit for that infrastructure. The Fort Bliss and Dugway projects demonstrate a shift in US procurement strategy, whereby private capital is leveraged and employed to build dual-use infrastructure, with guaranteed access secured through long-term arrangements.

The integration of military demand into AI development creates an implicit stabiliser. Defence demand is less cyclical than that of private markets, being dependent on long term “forecasts” of geopolitical risk. This justifies long-term procurement arrangements, unlike in the case of demand driven by market sentiment. Those forecasts and arrangements in turn, are being justified by actual military action as in Iran, based on claims of imminent threats to the United States—an assessment not shared even by all its allies. The perceived geopolitical tension that accelerates investment also precipitates irrational and costly conflict.

This nexus between finance, the State and war, can turn uncomfortable even for the AI business on which it currently rides. This is signalled by the spat between Anthropic and the Pentagon, which had signed a deal for use of the former’s technology by the latter. After entering into the arrangement, the company valued at $380 billion refused, possibly for shrewd business reasons, the defence departments demand for full control over how it uses the technology. Anthropic has argued that while it is willing to service national security needs with special versions of its model, it had some “red lines” on use of its product for “mass surveillance” of US citizens or for development of fully autonomous weapons systems. Few are ready to hand over the nuclear trigger to an AI agent. Even OpenAI, which reached a more conciliatory agreement with the defence department, has been forced by widespread protest to step back. The military has threatened to bar Anthropic from access to military procurement and Anthropic has sued the US government. It wants access to the military market, since financial returns are intrinsically linked to infrastructure of strategic military value, but on its own rather than the Pentagon’s terms.

This dilemma only illustrates that there are no certainties under capitalism, even when State action is deployed to ensure them. War, that is a corollary, is itself a source of uncertainty. Thus, the State-Finance-AI-War entanglement is now being tested in West Asia. Iran’s defensive Ramadan response to the US-Israeli military aggression against it has disrupted shipping through the Strait of Hormuz, contributing to sharp volatility in oil prices, which spiked above $120 per barrel in early March. The conflict has direct implications for AI infrastructure. Data centres are energy-intensive, and particularly reliant on natural gas and stable electricity supply. Rising energy costs increase operating expenses, while supply chain disruptions, affecting everything from fuel to semiconductor inputs, delay expansion. Simultaneously, the war is feeding back into financial markets, creating what might be seen as a reflexive loop, where war escalation leads to energy shocks, which prompt investor withdrawals, tighten credit conditions, reduce AI capital expenditure, weakens growth expectations, and feedback to drive new rounds of withdrawals.

Military demand may stabilise the system, but not without changing it. As private capital becomes more cautious, AI investment is likely to become more concentrated among large, government-linked actors, more selective with higher thresholds for new projects, and more strategically directed rather than purely market-driven. At the limit, AI infrastructure itself becomes a strategic asset and therefore equally emerges as a potential target in war.

The war against Iran itself is unlikely to trigger a sudden collapse in the AI sector. Unlike prior speculative cycles, AI is already embedded in both economic and military systems. However, the conditions that enabled its rapid expansion, including cheap energy, abundant liquidity, and relative geopolitical stability, are increasingly deteriorating.

What makes the war on Iran catalytic is that it activates three mechanisms simultaneously.

First, private credit, prior to the conflict, was already under considerable strain, and the war by tightening financial conditions precipitated a liquidity squeeze, precisely when capital is required to sustain existing positions. Secondly, AI infrastructure depends on stable, affordable power, and disruption in the Strait of Hormuz undermines the input-cost assumptions upon which expansion plans were based. Thirdly, the reflexive loop discussed above appears stable below a certain threshold, where the triangular system of finance, AI and military spending feed each other. Yet beyond that threshold, the same interconnections that provide resilience generate stress and threaten to down the system.

The result, however, is unlikely to be a sudden collapse. In fact, the outcome would be selective re-pricing and consolidation. Projects misaligned with state priorities will stall, capital will become more conditional, concentrate around strategically backed infrastructure and returns will depend increasingly on alignment with state priorities rather than purely commercial viability. The deeper observable shift is structural where private finance, technological infrastructure, and military strategy are no longer separable domains. This is already evident in Iran’s attacks on American data centres and the more recent threat to target a comprehensive list of large US and Israeli tech companies. A sector born of private finance is progressively reorganising around imperatives of War, under US imperialism.

Farwa Sial is a Research Associate at the Department of Economics, School of Oriental and African Studies (SOAS).

C.P. Chandrasekhar is the Global Head of Research and Policy at the International Development Economics Associates (IDEAs)

This article was first published in Economic and Political Weekly.