Alphabet, Oracle and Amazon Shares Fall on Sept. 23 as 10-Year Treasury Yield Reaches Highest Level Since 2007
U.S. stocks closed lower Wednesday. Alphabet fell 3.8%, Oracle 3.1% and Amazon 2.2%, while the 10-year Treasury yield rose about 12 basis points to roughly 5.1%. Market reports named the heavy borrowing by AI companies as one pressure alongside strong economic data, a weak Treasury auction and oil-driven inflation worries.
The $220 Billion Question That Both Sides Answer Differently
On Wednesday, Sept. 23, 2026, the Dow Jones Industrial Average fell 352.10 points, a drop of 0.68%. The S&P 500 lost 0.75% and the Nasdaq Composite fell 1.13%[2]. Alphabet fell 3.8%, Oracle fell 3.1% and Amazon fell 2.2%[1].
Behind the selloff was a number that hadn't moved this high in almost two decades. The yield on the 10-year Treasury note rose about 12 basis points that day. A basis point is one-hundredth of a percentage point, so 12 basis points means the yield climbed 0.12 percentage points, to about 5.09%. That's the highest level since 2007[1][2]. The 5-year Treasury yield crossed 5% for the first time in the same stretch[3].
Yahoo Finance, reporting on the session as it happened, pointed to several causes at once: strong economic data, a weak government bond auction, and worries that high energy prices would push inflation higher[1]. But it also flagged something else sitting underneath all of that. It called the big AI "hyperscalers" — the handful of companies building giant data centers — "credit-sensitive," because they're borrowing money at record levels[1]. That single word is the seed of a much bigger argument.
Same Numbers, Opposite Conclusions
Here's what nobody disputes. Amazon, Alphabet, Meta, Oracle and Nvidia have sold about $220 billion worth of investment-grade bonds so far in 2026. S&P Global puts the number slightly higher, at $225 billion. Either way, it's already more than all of 2025 combined[5].
Some people look at that figure and see danger. Nomura estimates that roughly $200 billion of big-tech borrowing equals about 25% of all the new Treasury notes and bonds sold to private investors this year — five times the tech sector's share in 2025[6]. JPMorgan, using a different way of counting, puts AI companies' debt at 68% of new long-term Treasury borrowing this year[19]. Bloomberg summed up the concern with a quote from a portfolio manager: AI borrowing "just touches everything"[6].
Other people look at the exact same $220 billion and see nothing unusual. Microsoft carries a AAA credit rating. Alphabet is AA+. Meta and Amazon are AA-. Those are some of the highest ratings a company can get[5]. Vanguard, which buys a lot of these bonds for its funds, describes the group simply as investment-grade firms with strong balance sheets and low debt[17]. Sage Advisory found that after issuing all this new debt, these companies' leverage — their debt compared with a year of operating earnings — sits around 0.4 to 0.7 times. The average investment-grade company runs closer to 3 times[18]. By that math, the hyperscalers still have room.
Why a Company Building the Future Feels Today's Interest Rates Twice
To see why this argument matters so much to stock prices, it helps to understand how a stock is actually valued. A company's share price is roughly a bet on all the profit it will make in the future, converted into today's dollars. When Treasury yields rise, that conversion gets less generous — a dollar of profit ten years from now becomes worth less right now, because investors can get a safer, better return just from parking money in Treasuries instead[20].
That hits AI companies twice over, and here's why. First, like any stock whose value rests on future profits, their share prices take a hit when yields rise. Second, unlike a company that isn't borrowing, they've financed their buildout with debt — so when rates rise, their own bills get bigger too[20]. That double exposure is the mechanism the whole debate turns on.
It's also why the companies are borrowing so much in the first place. FactSet found that new debt covered about 9% of these companies' capital spending in fiscal 2024. By mid-2026, that had jumped to 32%[12]. The data centers cost more than the cash coming in from the businesses that already exist, so bonds fill the gap.
Oracle Is the Company Both Sides Point To
Not every hyperscaler carries the same risk, and this is where the argument sharpens. Oracle's capital spending has hit $55.7 billion, against operating cash flow of just $32 billion. One analysis projects the company will keep spending more cash than it brings in through 2029[14].
The market has noticed. Credit default swaps are essentially insurance contracts that pay out if a company can't pay its debts — their price rises when investors think default risk is going up. Oracle's CDS spreads have widened much faster than those of Meta, Amazon, Alphabet or Microsoft[15]. Oracle fell 4% on Sept. 1 during an earlier bond selloff, then another 3.1% on Sept. 23[13][1].
Oracle's defenders don't deny the spending gap. They argue the company has already signed AI contracts that will eventually turn this spending into revenue, and that its current credit worries are about timing, not whether the demand is real[15]. Its CDS spreads actually fell sharply back in February 2026, right after Oracle laid out a financing plan that reassured investors[16]. Bond buyers, meanwhile, have grown pickier across the board — even about debt from banks and industrial firms competing for the same pool of money, not just AI names[7].
One Pool of Money, Two Big Borrowers
Underneath the disagreement about individual companies sits a simpler structural fact. Pension funds, insurance companies and bond funds that buy long-term debt only have so much money to lend. This year, both the U.S. Treasury and the AI hyperscalers are drawing from that same pool[6][9].
Skeptics call this "reverse crowding out" — instead of government borrowing squeezing out private borrowers, as the textbook version goes, private tech borrowing is competing hard enough to put upward pressure on rates for everyone, including the government[9]. Investment-grade companies of all kinds have sold nearly $1.5 trillion in bonds this year, up 36% from a year earlier, and hyperscaler debt is a growing share of that total[6]. Estimates of just the AI-related slice for 2026 range from about $300 billion to $570 billion, depending on what gets counted; Goldman Sachs puts it at $489 billion[20].
Treasury Secretary Scott Bessent has taken the other side of that argument. In an interview with Fox Business's Larry Kudlow, he played down the rise in yields, treating it as a market story rather than a policy problem[6]. That's a predictable position for the official in charge of financing the government's own debt — every rise in yields adds to the interest costs taxpayers cover[6][9]. New AI bonds are already paying more to attract buyers than they did earlier this year: Alphabet's August 10-year bond paid 85 basis points more than comparable Treasuries, up from 63 basis points on a similar bond it sold in February[9][11].
What the Coverage Left Out — and What's Still Open
How this story got told depended a lot on who was telling it. The Daily Caller framed it as a "Big Tech's AI-Debt Binge," leaning on a Goldman Sachs projection that hyperscaler debt sales could hit $420 billion next year[8]. Fortune, a business magazine, called the same trend an "AI debt orgy" in one piece and "reverse crowding out" in another, emphasizing what it estimated as $1.65 trillion in hidden, off-balance-sheet borrowing[9]. Asset managers like Vanguard and Sage Advisory, who profit from buying these bonds, described the same companies as simply well-capitalized firms making ordinary investments[17][18].
The same-day market reports read as the least slanted of the bunch. Yahoo Finance, TheStreet and Benzinga all listed several causes for the Sept. 23 selloff — strong economic data, a weak Treasury auction, energy-driven inflation fears — with AI borrowing as one factor among others, not the whole story[1][3][4]. Coverage from outside the U.S. was thin; a Malaysian outlet, The Star, ran a headline tying the day's move to AI and the flattening yield curve, but only the headline could be confirmed, and it may simply be wire copy[4].
What isn't in dispute is that the data centers are getting built regardless of who wins this argument. The debt funding them will sit on corporate balance sheets for decades. Whether that debt looks, in hindsight, like ordinary investment or overreach depends on something nobody can check yet: whether the AI profits these companies are betting on actually show up, and whether they show up before the bill comes due[20].
Summary
U.S. stocks fell on Wednesday, Sept. 23, 2026. The Dow lost 352.10 points (0.68%), the S&P 500 fell 0.75% and the Nasdaq Composite fell 1.13%[2]. The yield on the 10-year Treasury note rose about 12 basis points to 5.09%, its highest level since 2007[1]. Among the biggest AI builders, Alphabet fell 3.8%, Oracle 3.1% and Amazon 2.2%[1].
Yahoo Finance's market report gave several reasons for the jump in yields: strong economic data, a weak Treasury auction and worries that high energy prices will push up inflation[1]. It described the AI 'hyperscalers' — the handful of firms building giant data centers — as 'credit-sensitive' because they are borrowing at record levels[1]. By one count, Amazon, Alphabet, Meta, Oracle and Nvidia have sold about $220 billion in investment-grade bonds so far in 2026. That is already more than all of 2025[5].
The core dispute is what this borrowing means. Skeptics, including some bond investors and financial writers, say it is so large that it now competes with the U.S. government for lenders' money. That, they say, pushes up interest rates for everyone and leaves the firms exposed if AI profits come late[6][9]. Defenders say most of these companies carry very little debt compared with their earnings. On this view, borrowing to build is normal for firms with top credit ratings[17][18]. Oracle is the exception both sides point to. It is spending far more than the cash its business brings in, and the market price of insuring its debt against default has risen faster than for its peers[13][15].
The dispute turns on one mechanism. When Treasury yields rise, borrowing costs more. Future profits also become worth less in today's money. So companies whose stock prices rest on big future AI profits, and that fund the buildout with debt, take the hit twice[20].
The Event
On Wednesday, Sept. 23, 2026, the Dow Jones Industrial Average fell 352.10 points (0.68%), the S&P 500 fell 0.75% and the Nasdaq Composite fell 1.13%[2]. The 10-year Treasury yield rose about 12 basis points (1 basis point equals 0.01 percentage point, so 12 basis points is 0.12 percentage points) to 5.09%, its highest level since 2007[1]. The 5-year Treasury yield passed 5% for the first time since 2007[3]. Alphabet shares fell 3.8%, Oracle 3.1% and Amazon 2.2%[1].
Undisputed Facts
- U.S. stocks closed lower on Sept. 23, 2026, with the Nasdaq Composite down 1.13%[2].
- The 10-year Treasury yield reached its highest level since 2007. Yahoo Finance reported 5.09%, and other reports put it near 5.1%[1][2][20].
- Alphabet fell 3.8%, Oracle 3.1% and Amazon 2.2% on the day[1].
- Market reports named several drivers of the yield jump: strong economic data, a weak Treasury auction and worries about energy-driven inflation[1].
- Amazon, Alphabet, Meta, Oracle and Nvidia have sold about $220 billion in investment-grade bonds so far in 2026, more than in all of 2025. S&P Global puts the figure at $225 billion[5].
- Microsoft is rated AAA, Alphabet AA+, and Meta and Amazon AA-, all near the top of the investment-grade scale[5].
- Alphabet's 10-year bond in August paid 85 basis points more than comparable Treasuries, up from 63 basis points on a similar February bond[9][11].
- Oracle's credit default swap spreads, the price of insuring its debt against default, have widened far more than those of Meta, Amazon, Alphabet and Microsoft[15].
The Pressure
Strip away the moralizing and blame. What structural realities persist regardless of which narrative wins?
- Capex outrunning cash flow
- The AI buildout costs more than these firms earn from operations, so they must borrow. FactSet found that new debt rose from 9% of capital spending in fiscal 2024 to 32% by mid-2026[12].
- One pool of long-term lenders
- Pension funds, insurers and bond funds that buy long-term bonds have limited money. AI firms and the Treasury now compete for it. Nomura puts big-tech borrowing at about 25% of the Treasury's net note and bond issuance to private investors[6].
- Why yields hit growth stocks
- A stock price is roughly today's value of expected future profits. When safe Treasuries pay more, profits far in the future are worth less today. Debt-funded builders also pay more to borrow. So AI stocks feel higher yields twice[20].
- Credit tiers differ
- The four AAA-to-AA- hyperscalers sit in a different risk tier from Oracle. CDS spreads show that split: Oracle's widened far more than its peers'[5][15].
Material realityThe 10-year Treasury yield is near 5.1%, the highest since 2007[1][2]. Investment-grade companies of all kinds have sold nearly $1.5 trillion of bonds this year, up 36% from a year earlier[6]. Hyperscaler borrowing is a large and growing share of that. Estimates for 2026 AI-related debt range from about $300 billion to $570 billion depending on what is counted, and Goldman Sachs puts it at $489 billion[20]. New AI bonds now must pay more extra yield to attract buyers[7][10]. The data centers are being built regardless, and the debt will stay on the books for decades, whichever story about it wins.
Narrative as a weaponTwo stories compete. The first is a 'debt binge' or 'bubble' story, pushed by bubble-warning commentators, parts of the business press and some right-leaning outlets. It asks readers to see AI borrowing as the main force behind rising yields and a sign of excess[8][9]. The second comes from the companies, their bond buyers and asset managers such as Vanguard. It asks readers to see well-capitalized firms making normal investments, with Oracle as a special case[17][18]. The Treasury has an interest in playing down yield worries[6]. Same-day market reports were the least slanted. They listed economic data, a weak auction and energy prices alongside AI debt, so framing that credits the Sept. 23 drop to AI borrowing alone goes beyond what those reports say[1][3][4].
How Each Side Sees It
Each major actor’s view — how it frames things, its underlying incentive, and how it’s materially affected. Tap a side to read it.
Frames it asTheir best argument is that the borrowing is small next to their earnings. Post-issuance leverage — debt compared with yearly operating earnings — often runs about 0.4 to 0.7 times, according to Sage Advisory, against an investment-grade average near 3 times[18]. Vanguard describes the group as investment-grade firms with strong balance sheets and low leverage[17]. On this view, a top-rated company borrowing to build data centers that customers are already renting is ordinary corporate finance, not a bubble. The ratings, AAA to AA-, back that up[5].
WhyTo win the race for AI computing capacity without slowing spending. Bonds let them build faster than their cash flow alone allows. FactSet found that yearly new debt rose from 9% of their capital spending in fiscal 2024 to 32% by mid-2026[12].
Impact on themTheir borrowing costs are rising. Alphabet's August 10-year bond paid 85 basis points over Treasuries, compared with 63 in February[9][11]. Amazon had to offer 18 to 21 basis points of extra yield on the longest part of a $25 billion sale after orders fell to 2.5 times the bonds on offer, down from 3.2 times in March[10]. Their stocks also fall when yields jump, as on Sept. 23[1].
Frames it asOracle's case is that its large AI contracts will turn today's spending into future cash flow. Its credit default swaps fell sharply in February 2026 after it laid out a financing plan[16]. Supporters describe its current credit worries as timing risk on contracts it has already signed, not a doubt about whether demand exists[15].
WhyTo finish its data-center buildout and turn its contract backlog into revenue before its funding costs climb much further[13].
Impact on themOracle is the most exposed. One analysis says its $55.7 billion in capital spending far exceeds its $32 billion in operating cash flow. It projects negative free cash flow, meaning more cash going out than coming in, through 2029[14]. Because its balance sheet carries more debt, rate moves hit its stock harder than its cloud rivals'[13]. It fell 4% on Sept. 1 in a bond selloff and 3.1% on Sept. 23[13][1].
Frames it asTheir strongest point is about scale, not creditworthiness. Nomura estimates that about $200 billion of borrowing by the biggest tech firms equals about 25% of the Treasury's net issuance of notes and bonds to private investors, five times the 2025 share[6]. JPMorgan puts AI firms' debt at 68% of new long-term Treasury borrowing this year, using a different measure[19]. Each dollar a bond fund lends to a data center is a dollar it does not lend elsewhere. So heavy borrowing can push up rates for everyone, which Fortune called 'reverse crowding out'[9]. Buyers are now choosier about AI bonds while still competing hard for bonds from banks and industrial firms[7].
WhyTo be paid enough extra yield for the risk of lending heavily to a single industry, and to stay within limits on how much of one sector a portfolio can hold[7].
Impact on themThey demand larger discounts on new AI bonds. That raises the builders' costs and could slow the buildout if it continues[7][10].
Frames it asTreasury Secretary Scott Bessent has played down the rise in U.S. yields in a Fox Business interview. Bessent also announced what Bloomberg called an unusual intervention in the bond market to restrain yields[6]. The government's view is that its own borrowing can be managed and yields reflect many forces, not just AI[6].
WhyTo keep federal borrowing costs down while financing large deficits[6][9].
Impact on themEvery rise in yields adds to interest costs on the national debt. Bloomberg reported in August that AI borrowing was helping drive Treasury yields to 25-year highs[6].
Frames it asTheir argument is that demand for AI chips is real and backed by customer orders. Bloomberg reported that Nvidia is working with Wall Street firms to raise another $500 billion for AI[6].
WhyTheir sales depend on hyperscalers continuing to spend. A tighter debt market threatens that flow of orders[6][12].
Impact on themTheir shares tend to move with AI-spending sentiment. My searches did not confirm exact Sept. 23 moves for Nvidia or Broadcom.
Like this article?
The Bias Ledger average rating 3.7
The same story, as framed by outlets across the spectrum, ordered least to most biased. The bias score (1 = straight, 10 = heavily spun) is an AI assessment of that framing — click an outlet to see its track record. The tell is the word choice or omission that reveals the angle.
| Outlet | Vantage | Bias | How they frame it | The tell |
|---|---|---|---|---|
| TheStreet | U.S. center | 2 | Nasdaq, Russell 2000 sink as 5-year Treasury hits 5% for first time since 2007 | It anchors the story in a checkable yield milestone rather than the AI-debt storyline. |
| Benzinga | U.S. center (retail-trader market news) | 2 | Nasdaq 100 Slips, 10-year Yields Hit 19-Year Highs | 'Slips' plays down the move where other outlets say 'tumble.' It frames the day as a rates story. |
| Yahoo Finance | U.S. center | 3 | Stock market today: Dow, S&P 500, Nasdaq tumble as 10-year Treasury yield surges to 2007 high | It lists several causes: data, a weak auction and energy-driven inflation. It calls the hyperscalers 'credit-sensitive' because of record debt. 'Tumble' is mildly stronger than a 0.75% S&P drop. |
| The Star | Malaysian (owned by a company controlled by the MCA, a Malaysian political party) | 3 | Financial stocks fall with AI and flattening yield curve in focus | It shifts the lens to banks and the yield curve instead of the tech names. Only the headline was reviewed, and it may be wire copy. |
| Bloomberg | U.S. center (financial data company) | 4 | AI Is Driving Up Treasury Yields: 'It Just Touches Everything' | The headline states AI causes higher yields as a flat claim, backed by Nomura's estimate. The body includes Bessent's pushback. |
| The Daily Caller | U.S. right | 6 | Big Tech's AI-Debt Binge Set To Hit New Eye-Popping Milestone | 'Binge' and 'Eye-Popping' treat the borrowing as excess. It leans on a Goldman projection for next year rather than current data. |
| Fortune | U.S. center (business press) | 6 | AI hyperscalers issuing a flood of bonds are 'reverse crowding out' the Treasury as US debt soars | 'Flood' and the crowding-out frame make AI borrowing the main driver of yields. A sister piece called it an 'AI debt orgy.' Other causes of higher yields get less weight. |
References
- Stock market today: Dow, S&P 500, Nasdaq tumble as 10-year Treasury yield surges to 2007 high — Yahoo Finance · U.S. center; mainstream financial news portal
- Dow falls 350 points as 10 year Treasury yields hit 5.1% — Invezz · Trading-oriented financial news site, ad and affiliate funded
- Stock Market Today (Sept. 23, 2026): Nasdaq, Russell 2000 sink as 5-year Treasury hits 5% for first time since 2007 — TheStreet · U.S. center; retail investor news
- Nasdaq 100 Slips, 10-year Yields Hit 19-Year Highs: Stock Market Today — Benzinga · Retail-trader market news, ad funded
- Hyperscaler bond issuance reaches about $220 billion as AI financing costs rise — MLQ News · AI-focused market news aggregator
- AI Is Driving Up Treasury Yields: 'It Just Touches Everything' — Bloomberg · U.S. center; financial data company serving institutional investors
- Corporate Bond Buyers Get Picky With Flood of AI Debt — U.S. News & World Report (wire copy) · U.S. center; carries financial wire reporting
- Big Tech's AI-Debt Binge Set To Hit New Eye-Popping Milestone — The Daily Caller · U.S. right; conservative digital outlet
- AI hyperscalers issuing a flood of bonds are 'reverse crowding out' the Treasury as US debt soars — Fortune · U.S. center; business magazine, privately owned
- The AI boom is increasingly built on debt, but investor demand is plunging just as hyperscalers ramp up their bond blitz — Fortune · U.S. center; business magazine, privately owned
- AI Borrowing Goes to Plaid — Intelligize · Legal and regulatory research service owned by LexisNexis
- Hyperscalers Tap External Financing as AI Capex Outruns Cash Flow — FactSet · Financial data vendor serving institutional investors
- Oracle Falls 4% as Bond Selloff Tests Its Debt-Funded AI Buildout, Nebius Slips — 24/7 Wall St. · Retail-investor financial site, ad funded
- The 5 Hyperscaler Stocks Most at Risk If AI Capex Spending Hits a Wall in 2026 — Yahoo Finance · U.S. center; mainstream financial news portal (syndicated investor content)
- Oracle leads record widening in hyperscaler CDS spreads — Seeking Alpha · Investor-focused platform; cites Offside Research via commentator Mike Zaccardi
- Oracle's credit default swaps are plummeting as financing plan boosts investor confidence — CNBC · U.S. center; business news network owned by Comcast/NBCUniversal
- The AI buildout comes to the bond market — Vanguard · Asset manager and large bond buyer; has a commercial interest in the corporate bond market
- AI Buildout Fuels Hyperscaler Bond Issuance and Credit Spread Volatility — Sage Advisory · Fixed-income investment manager; commercial interest in bond markets
- AI Companies' Debt Now Equals 68% of New Long-Term U.S. Treasury Borrowing This Year, JPMorgan Finds — Yahoo Finance · U.S. center; reports JPMorgan bank research
- Rising bond yields put tech valuations and AI financing under pressure — Financial World · Independent financial news site; cites Goldman Sachs estimates