Tech Jumbo Bonds Recast the Investment-Grade Credit Market

USD investment-grade bond sales have already topped $1.5 trillion in 2026, lifted by 20 jumbo deals that raised $384 billion. Technology companies have taken about 60% of jumbo supply since 2025, and the new tech paper carries a weighted average maturity of 14.2 years.

That mix turns a record year for corporate borrowing into a longer bet on a handful of AI builders. Buyers who track the investment-grade index pick up more technology and more duration than they did when these firms funded data centers from cash.

Twenty Jumbo Deals and a $1.5 Trillion Tape

Jumbo, on this tape, means $10 billion or more of proceeds at announcement. Behnood Noei, CFA, director of fixed income at WisdomTree, counted 20 of those prints in 2026 and said dollar investment-grade sales have already cleared $1.5 trillion, a pace he said can beat the pandemic-era record.

The broader corporate market is running even hotter than that IG cut. SIFMA’s tally of U.S. corporate issuance through August stands at $1,899.8 billion, up 29.8% from a year earlier, with $11.7 trillion of bonds outstanding as of the first quarter. Goldman Sachs has raised its 2026 dollar investment-grade forecast to $2.3 trillion.

THE 2026 SUPPLY TAPE

• All corporates: SIFMA logged $1,899.8 billion of U.S. sales through August, with average daily trading of $66.9 billion.
• Full-year call: Goldman Sachs now pencils in $2.3 trillion of dollar investment-grade supply for 2026 and $2.4 trillion for 2027.
• Summer run: TwentyFour Asset Management put August near $150 billion, above the $136 billion August 2020 mark, after June and July also set monthly highs.
• Latest print: Dell raised $5 billion in an investment-grade sale on September 10.

Early September was quieter after that summer sprint, which is the usual calendar, not a sign the AI borrowers have left. The heavy names still have 2026 and 2027 projects to fund, and the market has already shown it will take $10 billion-plus packages in a single morning.

Tech Companies Now Dominate the Mega-Deal Calendar

Noei’s cut is the one that should bother anyone who owns the index on autopilot. Technology has accounted for roughly 60% of jumbo dollar investment-grade sales across full-year 2025 and 2026 so far, and the sector is more than 20% of all dollar investment-grade new issues, jumbo and smaller together.

Lawrence Gillum, LPL’s chief fixed income strategist, wrote that the five large cloud builders (Amazon, Alphabet, Meta, Microsoft and Oracle) sold $121 billion in U.S. corporate bonds in 2025, more than four times their $28 billion average from 2020 through 2024. Barclays put 2025 debt sales for U.S. technology as a whole near $200 billion.

The 2026 book is already past that 2025 total for the inner circle. Johnathan Owen of TwentyFour Asset Management wrote on September 3 that Alphabet, Amazon, Meta, Microsoft, Oracle and SpaceX had issued over $180 billion of dollar bonds so far in 2026, with street estimates near $250 billion by year-end and about $400 billion if the wider technology and AI-related complex is included. Those six names are about 4.5% of the ICE BofA U.S. Corporate Index, up from a rounding error when they still paid for servers with cash.

WHERE THE NEW PAPER IS COMING FROM

JPMorgan’s weekly market report dated September 9 split that $266 billion into $182 billion from hyperscalers, $42 billion from data-center operators and $42 billion from semiconductor firms. The same note put the 2025 AI-related high-grade total at $139 billion and the 2024 total at $29 billion, more than a ninefold jump in two years, and it sketched $2.1 trillion of AI capital-spending funding demand in high-grade markets over the next five years.

The Bank of Korea’s monetary policy report put the turn in cash-versus-debt in one ratio. Bond sales by Amazon, Microsoft, Alphabet, Meta and Oracle reached $169.15 billion in the first half of 2026, equal to 51.3% of their combined capital spending of $329.93 billion. That share was 3.3% in the first half of 2024, 11.7% in the first half of 2025, and 37.2% in the second half of 2025. Debt now funds more than half the build, which is why the bond market, not the equity tape, is where the next constraint shows up.

Why Tech Bonds Now Last 14.2 Years

Technology bonds sold in 2026 have a weighted average maturity of 14.2 years, about 3.5 years longer than the broader dollar investment-grade market at 10.7 years. Among the five hyperscalers that have supplied much of the new paper, that average stretches to 16.5 years, which is how a rate move hits credit portfolios harder than the old tech-bond mix did.

Owen’s maturity split is sharper still. Across 2025 and 2026 supply, the group placed 41% of its issuance beyond 15 years, against 16% for the ICE BofA U.S. Corporate Index, and 31.2% in the over-26-year bucket against 13.0% for that index, more than twice the market weight. Data centers last decades, so 30- to 50-year debt matches the asset. It also leaves anyone who owns the corporate index holding a longer rate bet than the coupon screen suggests.

HOW FAR THE NEW TECH PAPER RUNS

• Amazon: Dollar tenors reach 2076, with other lines in the 2060s, which now anchors the ultra-long corporate bid in dollars.
• Alphabet: A £1 billion sterling century bond due 2126, the first from a technology issuer since Motorola in 1997, drew £9.5 billion of orders at 120 basis points over gilts.
• Meta and Oracle: Longest paper runs to 2066, sitting in the same pension-and-insurer bucket as 30-year Treasuries.
• Microsoft: The short outlier. It has not yet termed out 2026-27 capital spending in the bond market, so more long-end supply is still queued.

Owen put 2026 AI capital spending above $750 billion and 2027 above $1 trillion. He also noted a 20-year average return on invested capital of 17.75% against funding costs of 5.5% to 7%, which is why a 30-basis-point spread move has not slowed the printers. They are not price-sensitive issuers. Index funds that have to buy the new long bonds are.

Oracle Sits One Notch Above Junk

The first credit that has actually broken is Oracle. S&P Global Ratings cut Oracle’s long-term rating to BBB- from BBB on July 9, 2026, and cut the short-term rating to A-3 from A-2, with a stable outlook. That is the last rung of investment grade. One more step is junk, which would change who can hold the bonds and what they cost to roll.

Oracle Corp.’s rapidly expanding AI infrastructure business is increasing its overall credit risk, reflecting our more cautious view of the AI infrastructure industry, including rising capital expenditure (capex) requirements, an uncertain path to profitability, rapidly evolving industry and competitive landscape, and high customer concentration.

S&P Global Ratings, July 9, 2026 rating action

The specimen in the cash market is the 5.95% notes due September 26, 2055, a $3.5 billion line issued on September 26, 2025. On September 11, 2026 those notes sat at 79.13 cents on the dollar for a yield of 7.781%. A 30-year investment-grade print that trades in the 70s is the price of stretching a software balance sheet over an infrastructure cycle.

ORACLE’S CREDIT FILE AFTER THE CUT

• Cash burn: Fiscal 2026, which ended May 31, showed capital spending of $55.7 billion and free cash flow of about -$23.7 billion.
• Next year’s gap: S&P forecasts a free-operating-cash-flow deficit near $42 billion in fiscal 2027, against Oracle’s $90 billion to $95 billion capex guide and about $70 billion of net cash outlay.
• Backlog: Remaining performance obligations were $638 billion at the end of fiscal 2026. S&P estimates OpenAI is roughly half of that book.
• Off the balance sheet: The rating firm adds $260 billion of extra lease commitments due to start between fiscal 2027 and 2029, plus $13 billion of purchase obligations as of May 31, 2026.
• Leverage path: S&P-adjusted leverage stays above 4 times over the next two years and in the mid-4 times area in fiscal 2027. A move above 4.5 times, or a view that cash flow will not turn positive by fiscal 2029, could bring another cut.

S&P also counts a $5 billion mandatory convertible preferred issue from February 2026 and a planned $20 billion equity raise in calendar 2026 as reasons the company can stay investment grade. Cloud infrastructure was 27% of fiscal 2026 revenue and is projected at almost 60% by fiscal 2028, which is the mix shift the rating firm calls riskier than the old software stack. Alphabet, Amazon, Meta and Microsoft still sit in a different ratings neighborhood. Oracle is the name that shows what happens when the capex line outruns the cash-flow line and the debt still has to clear as investment grade.

Takeovers No Longer Flag the Next Jumbo

Only 11% of the jumbo deals have been tied to mergers and acquisitions. The old pattern gave credit desks a filing, a merger proxy, and weeks to resize a line. AI infrastructure paper does not arrive with that map. Treasurers print when the window is open, often in $10 billion-plus packages that land in the index the same week.

That is a portfolio problem as much as a trading problem. A passive investment-grade fund cannot decline the new Oracle or Amazon long bond if the rules say it has to look like the benchmark. The concession on the day of issue is the only price that fund will get. Owen’s coverage snapshot shows the bid is still there, but it is no longer automatic at the long end: the 30-year book was 4.5 times for Oracle, 3.6 times for Meta and 4.6 times for Alphabet, against a 2.98 times year-to-date average for U.S. investment-grade deals. Amazon’s July 30-year, which caught the market off guard, cleared at 2.5 times, the weakest of that set.

When the book shrinks to two and a half times on a name of Amazon’s quality, the constraint is no longer whether the deal gets done. It is how much extra spread the next $25 billion print has to pay, and how much of that spread the index then locks in for 30 years.

A Fight for the Same 30-Year Buyers

The second-order hit sits outside the corporate index. When Amazon or Alphabet prints $5 billion to $6 billion of 30-year paper, it bids for the same pension, insurer and liability-matching money that takes 30-year Treasuries. Some of those buyers then hedge the rate risk by selling the matching Treasury, which is a second round of supply at the same point on the curve. Owen put the 30-year Treasury yield at 5.25%, about 40 basis points above the start of 2026, and treated hyperscaler supply as extra weight on a long end that already has large deficits to digest.

WHERE EXPERTS DISAGREE

• The crowding view: Owen argues that long tech paper competes directly with 30-year Treasuries for a finite buyer pool, and that the AI cycle through 2028 keeps that pressure in place.
• The global-premium view: Long yields have also risen in Japan and Germany, which points to a worldwide term-premium move. On that reading, tech supply is extra weight, not the engine.

Both can be true at once. A global repricing of long money still leaves U.S. corporate desks competing with the Treasury for the last 30-year bid, and the corporate side can pay up in a way a government auction cannot. TwentyFour’s own arithmetic is that hyperscaler supply is a marginal negative technical, not the source of the move. The practical result for a credit portfolio is the same: more long tech paper at wider concessions, sitting next to a Treasury curve that no longer offers a free hedge.

How QIG Is Cutting Technology Risk

The WisdomTree U.S. Corporate Bond Fund holds a Technology and Electronics weight of 5.64%, and its average maturity is 10.30 years. That is a deliberate underweight versus the benchmark, including less of the weaker tech credits such as Oracle, against a market where technology is more than 20% of new dollar investment-grade supply and about 60% of the jumbo calendar.

As of June 5, 2026 the fund’s largest sector was Financials at 33.29%, with an effective duration of 6.33 and a 0.18% expense ratio. Gillum has put technology near 10% of the Bloomberg Corporate Bond Index, up from 9% in 2024, so a 5.64% tech sleeve is a real skip relative to that benchmark, not a rounding difference. The skip does not shrink the $384 billion of jumbo paper. It only changes who is forced to own it.

Index investors still take the new 16.5-year hyperscaler mix. Active books that can say no, QIG among them, are treating Oracle-style BBB- paper and 40-year tenors as optional. Microsoft still has not termed out its 2026-27 capital plan in the bond market, and street estimates still point to about $250 billion of dollar supply from the inner circle by year-end. The next jumbo will not come with a merger filing. It will come with a 20- or 30-year tranche, and the index will have to find a home for it.

Disclaimer: This article is news reporting and analysis of corporate-bond supply, credit ratings and fund positioning, and it is for information only. It is not investment advice, a recommendation to buy or sell any bond, fund or stock, or a rating opinion. Readers should consult a licensed financial adviser or credit specialist who can review their own holdings, tax status and risk limits before making any portfolio change. Figures and ratings reflect the cited statements and market quotes as of the dates named in the piece and can move as new deals price, as agencies revise outlooks, and as 2026 issuance continues.