The New Credit Crunch: AI vs Governments - Who Gets the Capital?
The global financial system has entered a strange and uncomfortable phase. For decades, governments were the undisputed giants of the bond market: they issued debt, investors bought it, and the system rolled forward with predictable rhythm. Corporations were background actors. Today, that hierarchy is breaking down. The world’s largest technology companies, those building the infrastructure for artificial intelligence, are issuing debt at a scale that rivals mid-sized nations and competing directly with governments for the same pool of global capital.
This is the new credit crunch. It is not driven by failing banks or collapsing collateral, but by demand. Two enormous borrowers, governments and AI hyperscalers, are tapping the same well at the same time. The well is deep, but not bottomless, and the effects are already visible in bond yields, credit spreads, refinancing costs and the behavior of institutional investors.
To put faces to the battle, imagine Kevin Warsh on one side of the table and Jensen Huang on the other. Warsh represents the sovereign world, governments struggling to secure affordable capital amid structural deficits and rising borrowing costs. Huang represents the AI world, hyperscalers consuming capital at unprecedented speed to build the infrastructure of the future.
Nor is this confined to the United States. The UK is issuing record levels of gilts to fund structural deficits, inflation indexed obligations, NHS pressures and defence commitments. Yet UK pension funds and insurers, historically reliable gilt buyers, are increasingly allocating to global corporate credit, particularly USD investment grade, drawn by higher yields, deeper liquidity and stronger credit quality. That diverts capital from gilts toward the very hyperscalers driving the AI debt super-cycle. Like the US, investors are not automatically choosing sovereigns.
The AI Build Out: A Physical Revolution
Artificial intelligence is not just an algorithmic revolution; it is a physical one. Training frontier models requires vast data centres, each consuming more electricity than a small town, plus land, fibre backbones, cooling systems and millions of GPUs. Analysts estimate that global AI related capex will exceed a trillion dollars annually within a few years. No company can fund that from cash flow alone, so debt markets are the only option.
The result is a surge in corporate borrowing unlike anything in modern history. Hyperscalers are now among the largest issuers of investment grade debt, bought by pension funds, insurers, sovereign wealth funds and asset managers drawn to their strong balance sheets, dominant market positions and enormous revenue potential. Investors lend at attractive rates even as spreads widen slightly to reflect the volume.
Governments Under Pressure
Governments face their own financing pressures. The United States, United Kingdom and much of Europe are running historically large deficits, driven by ageing populations, healthcare costs, defence spending and lingering inflation. Refinancing old debt at higher rates adds to the burden. Sovereign supply is rising sharply, and governments need lots of buyers.
This is where the collision occurs. Investors cannot buy everything, and increasingly they are choosing AI. The logic is simple: sovereign bonds offer safety but limited growth, while AI related corporate bonds offer strong credit quality, higher yields and exposure to one of the most transformative technologies of the century. For many institutions the decision is mathematical, not ideological. If a hyperscaler yields even slightly more than a government bond of similar duration, capital flows to the better risk adjusted return.
The Feedback Loop Tightening Global Markets
As more capital is absorbed by AI issuance, governments must offer higher yields. Higher yields mean higher borrowing costs, larger deficits and more issuance, which forces governments to compete even harder with hyperscalers. The loop is vicious, and it is already tightening financial conditions across developed markets.
The IMF has begun to warn about this. Recent stability reports list AI profitability risk and AI driven capital absorption alongside traditional systemic threats. The main concern is not that AI companies will default, but that their borrowing needs are so large they distort global credit markets. If governments struggle to fund themselves at sustainable rates, the effects ripple through mortgage pricing, corporate refinancing and sovereign credit ratings.
This is not a crisis in the traditional sense; it is a competition, but one that can destabilise when the stakes are high. Governments cannot stop issuing debt: they must fund public services, pensions and infrastructure. Hyperscalers cannot stop either: they must build the backbone of the next technological era or disappear. Both sides are trapped by necessity.
Who Gets Priority?
In the past, the answer was always governments. Today it is shifting. Investors increasingly treat hyperscalers as quasi sovereigns, with balance sheets stronger than many nations, global reach and cash flows that dwarf traditional industries. If this continues, governments may have to pay materially higher rates to secure the capital they need, reshaping fiscal policy, monetary policy and the long-term trajectory of sovereign debt sustainability.
How ETFs Reflect the Capital War
The ETF market has become a mirror for this macro narrative, and investors can express a view on the competition through the ETFs they hold. Sovereign bond ETFs are the traditional anchor of global portfolios: government debt, long duration gilts, Treasuries and inflation linked securities. When yields rise because governments must compete harder for capital, their prices fall and their duration risk increases.
Corporate bond ETFs, particularly investment grade, are increasingly dominated by hyperscalers, whose bonds occupy ever larger portions of the indices. Corporate bonds do not disclose use of proceeds in enough detail to isolate AI spending, but hyperscalers’ capex cycles make clear their issuance is overwhelmingly AI driven. LQD is therefore the ETF through which the AI debt supercycle expresses itself in fixed income. Buying broad corporate bond ETFs means, in effect, financing data centres, GPU clusters and fibre networks, so performance is tied to AI infrastructure spending as well as credit spreads.
Thematic AI equity ETFs sit outside the credit competition but capture the equity side of the story: the upside of the build out, while sovereign bond ETFs capture the strain it places on government finances.
In short, sovereign ETFs express confidence in governments, corporate ETFs with heavy AI exposure express confidence in hyperscalers’ ability to out compete them for capital, and AI equities express confidence in the long-term profitability of the infrastructure being financed.
A Crisis of Allocation, Not Scarcity
The new credit crunch is a crisis of allocation, not scarcity. There is enough capital to fund both governments and hyperscalers, but not at the prices both would prefer. Someone must pay more, accept tighter conditions or adjust.
The story is still unfolding. AI capex may accelerate or slow as models become more efficient. Governments may tighten fiscal policy or keep running large deficits. Investors may keep favouring AI corporate debt or revert to sovereigns if yields rise high enough. But the competition for capital between two enormous borrowers is here to stay, and it will shape everything from ETF flows to yield curves to the stability of the global financial system.
In the end, the new credit crunch is about power, not crisis. Governments once held unquestioned power in debt markets; hyperscalers are now challenging it. The outcome will determine not just the cost of capital but the architecture of the next technological era, and investors, through their choices in ETFs and other instruments, will play a quiet but decisive role in deciding who wins.
Until next time.
Allan Lane