Funding & Finance

Bond Market 2026: Everyone Is Borrowing at Once

SpaceX went to the bond market and raised $25 billion while sitting on $100.8 billion in cash. The company

By Conor Healy

August 13, 2026

Bond Market 2026: Everyone Is Borrowing at Once

SpaceX went to the bond market and raised $25 billion while sitting on $100.8 billion in cash.

The company did not need the money in any ordinary sense. It had just completed a record initial public offering, and its bank balance was larger than the GDP of most countries. It borrowed anyway, across five tranches stretching to 2056, at rates between 5.35% and 6.65%. Investors put in nearly $90 billion of orders, more than three times what was on offer.

That transaction is a small window onto something much larger. Corporations and governments are borrowing at a pace not seen since the pandemic, for entirely different reasons, at the same time, and into a market that has lost its most reliable buyer.

The Corporate Boom Is an AI Boom

The companies borrowing hardest are the ones building artificial intelligence, and the sums have become difficult to comprehend.

American investment-grade companies sold roughly $175 billion of high-grade bonds in June alone, according to data compiled by Bloomberg. That is about 60% more than the whole of June 2025, and it beat the previous monthly record set during the emergency of 2020. Wall Street dealers had forecast something closer to $130 billion.

Two deals accounted for nearly 29% of the month’s issuance by themselves. Nvidia and SpaceX each sold $25 billion of bonds, meaning two companies in a single month took almost a third of the entire American investment-grade market.

Amazon had already raised about $54 billion earlier in the year, Alphabet about $31.5 billion, and Oracle $25 billion. Total American investment-grade issuance has reached roughly $1.15 trillion so far in 2026, matching the pace of 2020, when the full-year figure hit a record $1.75 trillion.

Nearly all of it funds the same shopping list of chips, data centres, power infrastructure and networking equipment. This is the debt-funded half of the $725 billion the hyperscalers are spending on AI infrastructure, and the bond market is where that spending is actually financed.

SpaceX makes the logic unusually explicit. During its IPO roadshow, president Gwynne Shotwell and chief financial officer Bret Johnsen told prospective shareholders that future capital would come through borrowing rather than issuing more stock, effectively treating the IPO as a one-off equity sale. Debt is the funding strategy, and the bond market is the tap.

The Government Boom Is a Defence Boom

While companies borrow for compute, governments are borrowing for something older.

Sovereign issuers sold a record $504 billion of syndicated bonds in the first half of 2026, according to Bloomberg data. The previous record was $472 billion in the first half of 2020, when governments were funding emergency pandemic support. Peace-time borrowing has now overtaken the crisis.

Italy alone raised nearly €70 billion, around $81 billion, and has been the largest syndicated sovereign borrower in eight of the past ten years. Germany, having rewritten the fiscal rules that once defined its national character, is borrowing to fund defence and infrastructure. The United Kingdom, Belgium and Serbia each sold their largest deals ever.

Zoom out and the numbers get heavier still. The OECD reports that central governments issued $17 trillion of bonds in 2025, with $18 trillion projected for 2026, and that both sovereign issuance and total outstanding debt hit record highs.

The drivers are a stack of simultaneous pressures: military spending, infrastructure, the energy transition, ageing populations, and price shocks from the war involving Iran. Underneath all of it sits an unglamorous mechanical fact. The bonds issued during the pandemic are now maturing and have to be refinanced. Analysis by Natixis found that refinancing deals by euro-area sovereigns jumped 26% in 2026, well ahead of the 11% rise in total syndicated issuance, which suggests the record is driven more by rolling over old debt than by rushing to beat rate rises.

Why Are Bond Yields Rising?

Bond yields are rising because supply is at record highs while the largest and least price-sensitive buyer has walked away.

For over a decade, central banks bought government bonds in enormous quantities through quantitative easing. They were not chasing a return and did not care what they paid, so their presence suppressed yields and absorbed supply automatically.

That support has been withdrawn. After three years of quantitative tightening, central banks are stepping back, and the OECD notes that a growing share of bonds is now held by price-sensitive investors: pension funds, foreign buyers, hedge funds, people who will only buy at a price that suits them. Record issuance now has to be absorbed by investors who negotiate.

The evidence is showing up at auction. A 30-year United States bond auction in May drew a yield above 5% for the first time since 2007. The United Kingdom’s £15 billion offering in April attracted record orders, but only because it carried the highest ten-year yield since 2008. Buyers are still turning up, and they are charging more to do it.

There is a structural shift in demand too. The long-running move from defined-benefit to defined-contribution pension schemes has reduced the natural appetite for very long-dated bonds, exactly as governments need to sell more of them.

The Rush Is Its Own Cause

Sitting underneath the record numbers is a psychology that makes the whole thing self-reinforcing.

Companies are accelerating their borrowing plans partly out of concern that the Federal Reserve could raise rates again as it focuses on inflation. If money is going to get more expensive, the rational move is to borrow now and lock the rate in. Every issuer reasons this way at the same time, which floods the market with supply, which pushes yields up, which makes borrowing more expensive, which sharpens the incentive to hurry.

This is what explains a company with $100 billion in cash selling $25 billion of bonds. The bank account was full. SpaceX was converting a bridge loan into long-term debt and securing a rate before the window narrowed.

Governments are doing a version of the same thing. Syndications from Belgium, Spain, Austria and Portugal came earlier than strategists had anticipated. Everyone is trying to be first through a door that gets smaller the more people push through it.

The Case That This Is Fine

The bearish reading is easy to write, so it is worth taking the other side seriously, because the other side is currently winning.

Demand is ferocious. SpaceX’s $25 billion drew nearly $90 billion of orders. Greece tapped the market for €3 billion and received more than €36 billion of bids, twelve times what it wanted. That is not a market losing confidence in debt.

The OECD’s own assessment is that sovereign bond markets have continued to function effectively, supported by improved liquidity, and have smoothly absorbed the record volumes of supply. Nothing has broken.

The composition is reassuring as well. The corporate borrowers driving the surge are the most profitable companies in the world, funding assets they believe will generate returns. The sovereign surge is substantially about refinancing existing obligations rather than piling on new ones. Neither is the profile of a market in distress.

And higher yields, for the buyer, are simply a better deal. After fifteen years of being paid almost nothing to hold government debt, investors are finally being compensated. That is less a crisis than a normalisation, and quite a welcome one if you are a pension fund.

Where the Strain Sits

The risk is less dramatic than a collapse, and considerably more boring: the bill arrives later.

Faced with expensive long-term borrowing, many governments are issuing shorter-dated debt instead. The OECD found that in 2025 the ratio of bonds with maturities of thirty years or more to those maturing within one to five years was the lowest since at least 2008. Treasury bills have overtaken fixed-rate bonds by issuance volume and now make up 15% of the debt stock.

That is cost-efficient today and dangerous tomorrow. Short-dated debt has to be refinanced sooner, and if rates are higher when it comes due, the problem compounds. A government that borrows short to save money is betting that rates will fall before the bill lands.

For the corporate side, the concentration is the issue. One analyst put it bluntly when discussing SpaceX: holding both the equity and the bonds gives an investor the same execution risk twice, dressed up as two different asset classes. Someone who owns technology stocks, plus AI-adjacent corporate bonds, plus a pension fund exposed to both, has made one bet several times over while believing they have spread it.

That risk is systemic rather than personal. When two companies supply almost a third of a month’s investment-grade issuance, and both depend on the same AI thesis, the bond market has become a leveraged wager on that thesis paying off.

What the Bond Market Is Betting On

The bond market in 2026 is doing something it has rarely had to do, which is finance two unrelated historic build-outs simultaneously.

One is corporate and speculative: the largest infrastructure gamble in business history, made by profitable companies on the assumption that artificial intelligence becomes essential fast enough to justify the spending. The other is sovereign and defensive: rearmament, energy security, and the slow work of refinancing a pandemic that was itself paid for with borrowed money.

They have nothing to do with each other, and they are competing for the same finite pool of savings, at the same moment the buyer of last resort has left the room. That is the entire explanation for why yields keep climbing while order books stay oversubscribed.

None of it has broken, and it may never break. But the market has already delivered its verdict on the terms. A company with $100 billion in the bank borrowed anyway, because the money was there and it might not be later. Everybody else looked at that and reached the same conclusion.