Global bond yields reached their highest level since 2008, and Stephanie Aliaga said JPMorgan estimates the largest technology companies could put another $1.5 trillion of debt on their balance sheets before their borrowing matched the average of the broader investment-grade market.
The reading in the market is that a flood of technology borrowing is pushing yields up. Aliaga's reading is that the borrowers are unusually creditworthy, the demand for the paper is there, and the issuance is one contributor among several rather than the cause.
"Now, there's going to be some choppiness, and you see that in spread widening during these periods of significant issuance. But we think that the market is very capable of absorbing this new issuance."
Aliaga is a global market strategist at JPMorgan Asset Management, and the estimates she cited come from her own firm's investment bank rather than from a third party.
I listened to the full segment so you can skip it.
Here are the 6 takeaways that matter.
👤 Guest: Stephanie Aliaga, global market strategist at JPMorgan Asset Management, citing her own firm's investment-bank estimates for the AI build-out
🎙️ Host: Ed Ludlow, who anchors Bloomberg Tech from San Francisco
📰 Published: 1 September 2026 on YouTube (Bloomberg Tech)
🔴 YouTube | 🟣 Apple Podcasts | ⏱️ 8 min
Key Takeaways
Technology borrowing is a small share of the bond market but a fast-growing share of new issuance Ludlow put six hyperscalers at about 5% of the US high-grade index, double their share globally
The hyperscalers could borrow another $1.5T before their leverage matched the investment-grade average
The AI build-out needs $5.5T by 2030 and company cash flow covers only a fraction of it Private debt, SPV structures and government capital fill the rest
Signed customer commitments are now growing faster than capital spending That backlog growth is what Aliaga said improves visibility on the return
Supply, not demand, is the binding constraint for the foreseeable future
The data-center backlash shifts who pays rather than stopping the build-out The benefits of AI are spread across the country and the costs land in one county
1. Tech's Share of the Rise
Ludlow opened on the bond market rather than on a stock, and gave the show's own figures for how much of the high-grade market technology now is.
The scale of the move is what put the segment on air. Ludlow said global bond yields were climbing to their highest level since 2008, and that borrowing by the biggest US technology firms to fund AI has crowded out demand for government bonds
The show's own research put a number on the crowding. "At the same time, just six hyperscalers make up about 5% of the U.S. high-grade index, which is double their share globally."
Aliaga separated the stock of debt from the flow of new debt. "Well, just like what you cited, as a share of the overall market, it's still quite low. But in terms of new issuance, it is becoming a growing share of all of this debt hitting the markets. And we do think that this is set to just continue."
The credit quality of the issuers is her first answer to the worry. "Now, this debt is coming from very high-quality issuers. I mean, relative to the current investment-grade bond index, the leverage ratios for the hyperscalers are significantly lower." She said JPMorgan estimates they could add $1.5 trillion more debt before reaching the average lease-adjusted leverage ratio of the broader market
She was careful not to make issuance the whole story. "But to your point, I mean, this is just one factor among many that is contributing to higher yields on the long end. And it's something investors want to be very nimble around."
2. The Market Is Absorbing It
Ludlow said he is not a bond market expert and asked what the daily headlines of hyperscaler debt sales have actually done to prices.
Aliaga read heavy issuance that clears as evidence of demand, not of stress. "Well, I think it shows that there's just ample demand in the market for this, too."
The visible cost is wider spreads during the heaviest weeks, which she described as choppiness rather than damage. "Now, there's going to be some choppiness, and you see that in spread widening during these periods of significant issuance. But we think that the market is very capable of absorbing this new issuance." A spread is the extra yield a company pays over a government bond of the same maturity; it widens when buyers demand more compensation
She argued borrowing makes the build-out more durable rather than less, because the assets have long lives. "And debt is not inherently bad by any means, but it can be a really attractive form of financing for some of these hyperscalers building data centers that they intend to use for five, 10, or more years."
3. $5.5T by 2030
Asked whether the issuers are finished or will keep going, Aliaga said keep going, and gave the arithmetic behind it.
JPMorgan's investment bank estimates the total AI build-out amounts to $5.5 trillion by 2030, and Aliaga said internal funds do not reach it. "And hyperscaler cash flows can only get us a fraction of the way there."
The gap gets filled from several places, not just the bond market. She named private debt alongside public debt, special-purpose vehicle structures, and government capital A special-purpose vehicle is a separate company set up to hold an asset and its financing, which keeps the borrowing off the parent's own balance sheet
4. Cash Flow Meets Capex
Ludlow noted that these companies were not borrowers at all until recently and asked how the debt story and the equity story fit together.
Aliaga said cash generation has grown alongside the borrowing. "I think what's really important to recognize is in the midst of growing tech issuance, you also have significant growth in the operating cash flows of these companies." "In fact, right now, operating cash flows essentially neck and neck with capex expenditures this year."
The earnings season gave her the evidence she wanted on returns. Backlog growth at the top three hyperscalers has outpaced growth in capital spending, which she said means signed customer commitments for cloud capacity years into the future are rising faster than the money going into building it. "So there is visibility on the ROI equation."
She expects the spending to slow from here, and margins to benefit. "And as we get further along here, we do expect that capex will decelerate and that should allow some relief for margins and free cash flows."
5. Nvidia's Supply-Led Guidance
Ludlow offered Nvidia as the test case and read the company's own numbers at her.
The host put the guidance on the table. Ludlow said Nvidia had told the market it will have top-line growth of 70% in fiscal 2028, and that it would have been higher, around 100%, were it not for supply constraints He noted the guidance landed in the same five-day period as Jackson Hole, and that Nvidia does not normally give it
Aliaga took the willingness to sign long contracts as the signal. "I think the fact that you have AI demand now, these customers are scrambling to get a hold of compute, that they're willing to sign medium to longer term commitments."
The constraint is supply and she does not expect it to lift soon. "I think the reality is we're going to be operating in this supply-constrained environment for the foreseeable future." She said "tech is very good at clearing bottlenecks", memory among them, and framed the open question as one of timing and identity: "The key question for investors is when and who can actually succeed in doing so."
6. Data Centers' Local Cost
Ludlow flagged the later segment on voter opposition to data centers and asked whether JPMorgan treats it as a headwind.
Aliaga called it a risk to the economics rather than to the build-out. "It is a risk, but not one that we think will derail the AI build-out, but rather shift the economics of it."
The direction of the shift is toward the companies paying more of the bill. "And I think ultimately what is coming out of these debates, conversations, negotiations with state and local legislators is that the hyperscalers are shouldering more of the cost burden versus ordinary taxpayers." She said the detail varies by state and by local area
The structural problem is that the gains and the costs land in different places. "I think the challenges are that the benefits of AI are very diffuse, right, across the country." "But the cost of it can be very local."
Her conclusion is that policy has to close that gap, and that the public argument is running ahead of the facts. "There is this consumer surplus in AI services." "I think there's just a lot of misinformation around the benefits and costs and who's ultimately taking what."
Bonus Insights
Ludlow introduced the segment by saying technology has a part in a story that is usually told as a government-bond story, which is the framing the whole conversation ran on
Aliaga said the policy response she expects is action to ameliorate local costs rather than to slow construction
Aliaga's bottom line is that the AI build-out is going to be financed with debt because nothing else is large enough, that the credit market has shown it can take the paper, and that the argument worth watching is the local one about who pays for the power and the land.
Products, Companies & Tools Mentioned
JPMorgan Asset Management (Aliaga's firm; she cited its investment bank's $5.5 trillion estimate for the AI build-out by 2030)
Nvidia (Ludlow's case study: guidance of 70% top-line growth for fiscal 2028, held back by supply rather than demand)
If this was worth your time, send it to someone who has to have a view on this.
Get the latest market chatter as it happens: