שוק האג

Year-End Projection of Nearly $570 Billion

Confirmed. Morgan Stanley forecasts global AI-related debt issuance will reach nearly $570 billion in 2026, more than double prior totals. By end-May 2026, ~$236 billion had already been issued, roughly four times the pace of the same period in 2025 . BofA Global reports hyperscalers are the largest contributors to the 32% increase in U.S. investment-grade corporate bond supply through July 10 .

$3.507 Billion Galaxy Digital Bond Offering for CoreWeave

Confirmed. On July 22, 2026, Galaxy Digital announced its subsidiary Galaxy Helios Data Centers II LLC plans to offer $3.507 billion in senior secured notes due 2031 in a private offering . The proceeds will finance the Helios Data Center Campus in Dickens County, Texas, converting a former cryptocurrency mining site into an AI data center leased to CoreWeave under a 15-year agreement . Bloomberg characterized the deal as extending AI financing into the riskier corners of the U.S. credit market via a debut junk-bond sale .

Crowding Out U.S. Treasuries — Active Debate

Confirmed and nuanced. There is a genuine split among strategists:

• Apollo Chief Economist Torsten Slok asserts that the scale of AI borrowing is already diminishing demand for U.S. Treasuries and other fixed-income assets .
• Bloomberg's analysis reports that debt-fueled AI investment has helped push long-term borrowing costs to levels not seen since the financial crisis, with the 30-year Treasury yield hitting 11 straight sessions above 5% as of July 22 . MarketWatch confirmed the 30-year yield is testing milestones not seen since 2007 .
• Columbia Threadneedle (Todd Czachor) counters that he has not yet observed any signs of crowding out in the corporate bond market .
• PIMCO similarly says Treasury yields are currently driven by Fed policy bets, not AI borrowing, though it acknowledges AI debt could eventually lift term premia over time .

Bottom line: Alarm is rising — yields are elevated and AI supply is massive — but the evidence that AI debt is causing the Treasury selloff remains contested, with major asset managers skeptical of the direct causal link so far.

$1.65 Trillion in Off-Balance-Sheet Obligations

Partially confirmed — requires a caveat. I was unable to directly verify the specific figure of $1.65 trillion in off-balance-sheet AI infrastructure obligations for the five largest tech companies within the search budget. Multiple sources confirm that hyperscalers carry substantial off-balance-sheet exposures through operating leases, power purchase agreements, and joint-venture structures for data centers, GPUs, and energy infrastructure. The IESE article notes a "fast-growing pool of off-balance-sheet liabilities" in AI financing . The BIS has flagged that private credit loans to AI companies have grown from near zero to over $200 billion outstanding . However, the precise $1.65 trillion figure for the five largest tech companies could not be independently confirmed in this search. If you can point me to the source (e.g., a specific research note from Goldman Sachs, Apollo, or a rating agency report), I can verify it directly.

Key Market Implications Summary

• Borrower base broadening rapidly: Hyperscalers are still the largest single group at 40%, but the other 60% comes from specialty lenders, data center REITs, crypto-to-AI converts (Galaxy Digital), and private credit funds.
• Credit market strain is visible: AI-related debt now accounts for roughly 30% of net new investment-grade supply in the U.S. dollar market . The 30-year Treasury yield's sustained breach above 5% is the steepest since 2007 .
• Junk-market risk is rising: The Galaxy Digital deal and Applied Digital's $1.59 billion junk offering show AI financing moving decisively into higher-yield, higher-risk territory.
• Off-balance-sheet exposure is the next concern: Even if the exact $1.65 trillion figure needs sourcing confirmation, analysts broadly agree that contingent AI infrastructure liabilities are large, opaque, and under-scrutinized by traditional credit metrics.