Global Corporate Bond Issuance Hits Record $3.68 Trillion in First Half
Taylor Wilson
Global corporate bond issuance reached a record $3.68 trillion in H1 2026, up 10% year-on-year, driven by AI infrastructure financing; but credit spreads are already widening as investors begin repricing the risks of this borrowing boom.
What does $3.68 trillion actually mean?
It is the highest first-half total since LSEG — the London Stock Exchange Group's financial data arm — began tracking in 2000, roughly 50% above the most recent trough in 2022.
Yet the number of deals fell 12% year-on-year to 10,408. This means → each deal is getting much bigger — large companies are borrowing more per transaction.
The U.S. market accounted for over 30% of global deal count, crossing that threshold for the first time in six years.
Who is borrowing — and what for?
Amazon issued $36.8 billion in bonds in March; combined with a euro tranche, total proceeds reached $54 billion, earmarked for cloud computing and AI data centers.
Meta and Nvidia each raised roughly $25 billion; Nvidia's funds go toward advanced chip R&D.
Bank of America data shows cumulative AI-related corporate bond issuance has hit $220 billion since the start of 2026. In plain terms = this bond boom is, at its core, a race to build AI infrastructure with borrowed money.
Where is the AI financing wave spreading?
SoftBank issued ¥570 billion in dollar and euro bonds in April to cover costs tied to its OpenAI investment.
Panasonic Holdings completed a $500 million dollar-bond sale this month.
Beyond conventional bonds, convertible bonds — debt that can be converted into equity at a preset price — and other instruments raised a combined $470 billion in H1, the second-highest for the period since 2021.
Are investors pushing back?
After Amazon announced another $25 billion bond offering this month, its credit spread — the extra yield investors demand over risk-free government bonds — widened sharply.
Spreads for Meta and other hyperscale data-center operators jumped from roughly 1.2% to about 1.4%. This means → investors who previously absorbed massive supply without complaint are now demanding higher compensation for risk.
In plain terms = too many borrowers, too concentrated — bondbuyers are starting to push back on price.
Can the AI spending actually pay for itself?
In Q1 2026, combined capital expenditure at Amazon, Alphabet, Microsoft, and Meta exceeded each company's core operating profit for the first time since 2020.
Meta's capex plan for this year runs as high as $125–145 billion.
Oracle's June annual report explicitly flagged potential losses from AI investment, citing risks of data-center construction delays and major clients refusing to pay once facilities are built. This reflects a striking signal: even the tech giants themselves are warning the money may not come back on schedule.
What should investors watch in the second half?
S&P Global Ratings this month cut Oracle's long-term debt rating to BBB− — the lowest investment-grade notch — citing ballooning AI spending.
The Bank for International Settlements (BIS) warned in June: "Should the AI bubble burst, financial stability would be at risk — fixed-income markets are the most obvious vulnerability."
Goldman Sachs president John Waldron characterized AI financing as a "winner-take-all" race, noting that larger companies enjoy a clear edge in capital-market access. This means → the key variable for the second half is whether the widening in credit spreads is merely a brief digestion or the signal that AI financing is entering a higher-risk pricing regime.
Should the AI bubble burst, financial stability would be at risk — fixed-income markets are the most obvious vulnerability.
Bank for International Settlements (BIS)
June 2026 report
(Published June 2026)
Content is for reference only, not financial advice.