Research/Fixed Income / Artificial Intelligence
The AI Buildout is Now a Credit Cycle.
Jefrin Jabez Christian
Hong Kong
12 Nov 2026
Thursday's 5.5% drop in the Nasdaq was the equity market catching up to something the bond market had been pricing for a month. The question is no longer whether AI pays off. It is who lent the money, against what collateral, and who gets paid first when Texas says no.
On Thursday the Nasdaq Composite fell 5.52% and the S&P 500 fell 4.36%, the worst day for the index since April 2025. The semiconductor ETF had its worst session since the first week of Covid. Roughly $3 trillion of S&P 500 market value disappeared in six and a half hours, two days after the Democrats took the House, the Senate and, most importantly for this story, the governor's mansion in Texas.
That is the headline. It is also, I think, the wrong place to look. The equity market was the last to find out.
A month ago, CoreWeave's 9.625% bond due 2032 was yielding 14.8%. On Thursday it went through 16%, and the cost of insuring the company's debt for five years spiked above 1,289 basis points. Oracle, an investment-grade borrower, has a 4.95% bond due 2031 yielding about 7.3%, a wide premium for its rating. SoftBank's 2032 dollar bond had its biggest one-day fall in ten weeks on Friday, and by FourWeekMBA's count of FINRA TRACE prints, 13 of the 14 SoftBank dollar lines that traded on 4-5 November last printed below par. The average spread on Asian investment-grade dollar bonds is at its widest since July 2025.
The AI trade stopped being an equity story some time in the third quarter, when the combined free cash flow of the five largest hyperscalers turned negative. Everything built since then has been built on someone else's money. This piece is about whose.
From Cash to Credit
For most of the past decade the hyperscalers were the one corner of the market that did not need lenders. They funded capex out of operating cash flow and bought back stock with the rest. That has reversed quickly:
- Debt funded roughly 9% of hyperscaler capex in fiscal 2024, by FactSet's count. Goldman puts the figure at 27% for 2025 and about a third for 2026.
- The five big hyperscalers issued $108 billion of bonds globally in 2025. Year to date in 2026 they have issued $296 billion, which already exceeds Goldman's full-year forecast of $250 billion.
- Five AI-related bond deals totalling $67 billion have priced since October 29 alone, including Oracle's $10 billion on Monday. Meta is reportedly looking at a further $20 billion or more across several currencies.
- Hyperscaler paper is now around 9% of all investment-grade corporate supply. Five companies in one industry writing a tenth of the world's high-grade bonds is not a treasury decision any more. It is a macro supply factor.
The forward numbers are larger. Goldman expects about $440 billion of direct hyperscaler issuance in 2027 plus roughly $300 billion of project-level debt on top. Bank of America's estimate for all AI-related credit issuance, public and private, is $685 billion in 2027, up 31%, before it falls 23% to $531 billion in 2028.
That last number is the one I would frame. Everyone wants to know when AI capex peaks. Bank of America's answer is that AI borrowing peaks in 2027. Peak issuance is not the end of a credit cycle. It is the point at which the refinancing cycle begins, and the maturities already on the calendar cluster uncomfortably: Oracle's stressed bond matures in 2031, CoreWeave's in 2032, and Meta's Hyperion lease does not even start paying rent until June 2029.
The Three-Layer Stack
The useful way to think about AI debt is not by issuer but by seniority, because that is how losses get allocated. There are three layers, and each is held by a different kind of investor.
Layer one: hyperscaler senior unsecured
This is the paper that shows up in every investment-grade index fund, every insurer's general account and every 401(k) bond allocation. It is also, for now, the safest layer: Nvidia's five-year CDS is around 139 basis points, Oracle's around 298, against CoreWeave's 1,155. A 76% jump in Nvidia's CDS over a month, as happened, is a large move off a small base, not a solvency signal.
Oracle is the interesting case because it has run out of balance-sheet room before the others. In fiscal 2026 it raised $43 billion of debt and $5 billion of equity. In February it announced a $50 billion debt-and-equity programme and placed $30 billion of it within days. Last week, at its investor day, it said it would cap any additional capital-markets funding at $20 billion for the entire fiscal 2027 to 2030 investment cycle, then on Monday raised the whole $20 billion at once: $10 billion of common stock at a 2.5% dilution and $10 billion of senior notes across five tranches from three to twenty years. The stock fell 8.8% on the day. It is down 33% this year and 50% over twelve months.
A cap sounds like discipline. I read it differently. The cap applies to bonds and equity. It does not apply to the other pocket Oracle has been using: special-purpose vehicles that build the data centre and lease it back. On that side the running total is about $13 billion for the OpenAI site in Abilene (Blue Owl and JPMorgan, roughly $10 billion of it debt), a $38 billion package for sites in Texas and Wisconsin, and an $18 billion loan for New Mexico. Call it $70 billion of project-level money that never appears in the "capital markets" number. The cap is a statement about which pocket, not about how much.
Layer two: leases and SPVs
This is the layer that does not show up as debt at all, and it is now bigger than layer one.
Across the five hyperscalers, roughly $1.0 to $1.1 trillion of signed data-centre leases had not yet commenced as of the mid-2026 filings, up from about $200 billion five years ago. Meta's second-quarter 10-Q lists $278.99 billion of uncommenced leases running out to 2036, then adds another $68 billion signed in July alone with 18-to-20-year terms. That is about $347 billion of future rent against $84 billion of balance-sheet debt at the end of 2025. Alphabet discloses $85.2 billion of the same thing. Moody's has put the sector-wide off-balance-sheet figure at $662 billion and has said, politely, that the disclosures may not show the full picture.
The template for how this works is Hyperion, Meta's 5-gigawatt campus in Richland Parish, Louisiana. Meta contributed the site to a joint venture that is 80% owned by Blue Owl funds and 20% by Meta. The vehicle, Beignet Investors, issued $27.3 billion of bonds, arranged by Morgan Stanley, priced at about 225 basis points over Treasuries and rated A+, a single notch below Meta's own AA-. Because Meta owns only a fifth of the vehicle, the debt is not consolidated, even though Meta is the only tenant, signed a lease of twenty-plus years and gave a residual value guarantee of about $28 billion. Creditors have no direct recourse to the data centre itself.
One notch. That is the market's entire price for moving $27 billion off the balance sheet of the world's most-watched borrower. By October those bonds were reported at about 91 cents on the dollar. Industry trackers count more than $120 billion moved off hyperscaler balance sheets through structures like this in roughly eighteen months, and the other four are expected to copy it as their own balance sheets tighten.
Layer three: project finance, private credit and the neoclouds
Underneath the SPV bonds sits the money that actually pours the concrete and buys the chips before any lease has commenced. Bloomberg estimates more than $500 billion of AI data-centre debt outstanding, with about $200 billion held by private credit funds. Morgan Stanley expects private credit to supply more than half of the $1.5 trillion the build-out needs through 2028; the Bank of England's figure is about $700 billion of private-credit-financed data-centre capex over 2026 to 2028.
The visible face of this layer is CoreWeave, which carries about $51.6 billion of debt, including roughly $12 billion of loans from Blackstone's credit arm since 2023. Its secured lenders are, by JPMorgan's account, well protected by the GPUs and contracts they lend against. Its unsecured bondholders are the ones at 16%. Goldman now calls CoreWeave's bonds the bellwether for the entire AI trade, which is a strange thing to say about a company with a market share in the single digits, but it is correct: it is the marginal borrower, and credit cycles end at the margin, not at the median.
The Lenders Are Leaving the Table
Here is the part that I think deserves more attention than it has had. The banks that originated this exposure are now selling it.
- Barclays is seeking default protection from insurers on more than $10 billion of AI-infrastructure loans, and wants cover for over half the credit risk on some facilities.
- HSBC has been working on a significant-risk-transfer trade tied to more than $1 billion of data-centre loans and has sounded out insurers on a portfolio made up entirely of data centres.
- Deutsche Bank has been assembling an SRT on about $2 billion of project-finance debt that includes data-centre loans.
- On Friday ING put a $500 million portfolio of twelve data-centre loans up for auction, including a $25 million slice of the Stargate site in New Mexico and exposure to Vantage and CyrusOne.
The intended buyers in every case are the same: private credit funds, asset managers and insurers. Meanwhile a Willis Towers Watson survey in May found several credit-risk insurers "approaching sector limits or are already full" on data-centre exposure. That survey is six months old. The capacity has not grown since.
So the risk is moving from the institutions that underwrote it to the institutions that are least able to see it. US insurers held about $1.21 trillion of private credit at the end of 2025, roughly 13% of invested assets; life insurers alone hold $807 billion of private or illiquid bonds, a fifth of their fixed-income books. The new statutory disclosures that would let a regulator see how much of that is data-centre debt will appear for the first time in year-end 2026 filings, submitted early next year. Below the insurers sit the retail channels: BDCs, interval funds and the "semi-liquid" private credit products that asset managers have spent two years selling to individual investors.
If this pattern sounds familiar, it should. Three weeks ago I wrote about pension funds selling venture stakes at 65 to 78 cents on the dollar to new retail vehicles with $25,000 minimums. This is the same trade in a different asset class. The people with the most information are sellers. The people with the least are being offered the product.
Why Texas Matters More to Bondholders Than to Shareholders
Now to the election, and to the point I most want to make.
Texas accounts for about 27% of planned US data-centre capacity by Wells Fargo's count, and roughly a quarter of the new power capacity due through 2030. Its governor-elect, Gina Hinojosa, campaigned on a moratorium on all new data-centre approvals until the legislature passes protective laws, along with an end to tax breaks and tighter rules on water and power. Ohio and Georgia elected governors who have called for pauses of their own. Goldman estimates that from January, states run by Democrats who support extending or expanding construction moratoria will account for 45% of America's planned data-centre capacity. UBS reckons about 30% of expected 2028 capacity additions are now at risk of delay, with Texas alone representing 18 gigawatts of that. Wells Fargo's rule of thumb is that every 10% reduction in planned Texas capacity cuts Big Tech AI infrastructure spending by about 1.6%.
The equity market sold the hyperscalers on this news. I think it sold the wrong layer.
Look at who holds what in a project like Hyperion or Abilene. The hyperscaler has signed a lease that commences when the building is delivered. Until that day it pays nothing. The construction lender, by contrast, is accruing interest from the day the first draw is made. A moratorium, or even a twelve-month permitting delay, does not change the hyperscaler's obligation at all. It pushes the start date back. What it does change is the economics of the vehicle in between: a lease-backed project bond quietly becomes a loan against land, concrete and a half-installed power substation, with no tenant yet paying.
The arithmetic is not subtle. Take a $10 billion project financed 70% with debt at the roughly 7.5% that a 225-basis-point spread implies with the long bond at 5.26%. That is about $525 million a year of interest the SPV must carry. A one-year delay means half a billion dollars of carry funded by the vehicle's equity or its reserves, while the tenant's commitment sits untouched and the residual value guarantee, which only bites at the end of the lease, is of no help whatsoever. The hyperscaler holds an option on the political outcome. The SPV's lenders wrote it, and were paid one rating notch for the privilege.
Walk down the stack and the exposure gets worse at each step. The bank is secured and, increasingly, hedged through an SRT. The SPV bondholder is at 91 cents and falling. The private credit fund holding the construction loan has no public mark at all. The neocloud bondholder is at 16%. And the common-stock investor who sold Nvidia on Thursday is, in a delay scenario, the party whose cash obligations change the least.
I should be fair about the limits of this. Governors cannot halt every project, a national moratorium would face a presidential veto, and Texas already had a pause of sorts under Governor Abbott that Hinojosa derided as an election stunt. Demand for compute has not gone anywhere. But credit does not need a cancellation to get hurt. It needs a delay, and delays are the one thing a state government can deliver reliably.
The Rate Problem Underneath
All of this is happening into a tightening cycle, which the equity market has largely ignored because it has been busy with other things.
The Fed raised its target range to 3.75% to 4.00% in September and held in October. After Friday's payrolls came in at 165,000 against expectations around 65,000, futures put the odds of a December hike at roughly 70% to 80%, with another move priced for 2027. Cleveland's Beth Hammack said on Thursday that policy will need to become restrictive to get inflation back to 2%. Richmond's Thomas Barkin called inflation "the most important risk right now" and, notably, said AI is adding to price pressure today. The 10-year closed the week at 4.86% and the 30-year at 5.26%. Brent crude, which spent the autumn above $100 because of the Iran war, fell 13% this week on talk of peace, and the Fed still thinks it needs to hike.
For a data-centre bond this is the whole ballgame. These are 15-to-30-year structures priced off the long end. Every 50 basis points on the 30-year flows straight into the rent the SPV must charge, which flows into the hyperscaler's cost of compute, which flows into the price it must charge Anthropic and OpenAI, who are themselves funding commitments in the hundreds of billions. And the building is the only part of the asset that lasts thirty years. The chips inside it are depreciated over five or six, and their rental price falls every time a new generation ships. Marsh's credit team put the tension plainly earlier this year: lenders want asset lives that comfortably exceed loan tenors, and GPUs don't oblige. The industry's own phrase for the result is the "GPU debt treadmill".
This is why a 7.3% yield on Oracle, or 16% on CoreWeave, is not just a number. It is a hurdle rate. If the marginal dollar of AI infrastructure now costs more than the marginal dollar of AI revenue returns over the life of the chip, the build-out does not stop because anyone decides to stop it. It stops because the next SPV cannot be placed. Jefferies' Christopher Wood put it in a note on Friday: the likeliest end to the capex arms race is that somebody takes the credit card away. We are not there. But the card's limit is being reviewed.
The View From This Side of the Pacific
Two things stand out from the Hong Kong desk.
The first is where the Asian damage is concentrated. It is almost entirely in SoftBank, which raised $11.1 billion in one of the largest junk-bond offerings on record in September to fund its AI commitments, saw its shares fall as much as 16% on Friday, and is now the worst-performing name in Asian dollar credit with its CDS at a four-year high. SoftBank is not a builder. It is a financier of builders, the equity behind Stargate, and it is being repriced as such. That is exactly what you would expect if the stress is in the capital structure rather than in demand for compute.
The second is how differently the Chinese hyperscalers are funding the same race. Tencent raised about $4.66 billion in June and is marketing roughly $5 billion more in dollars and offshore yuan, with tenors of ten to thirty years, and told investors the money is mostly for chips. Its capex is running at a touch over 200 billion yuan a year, it has about $22 billion of offshore notes outstanding, nothing due this year, and no plans to issue equity or convertibles. Z.AI raised $9.6 billion through its IPO, placements and converts. Set that beside Oracle's $70 billion of debt and equity in a single fiscal year, before the SPVs, and the Chinese side looks almost frugal. Part of that is choice. A larger part is that export controls on advanced chips act as an involuntary governor on capex: you cannot over-borrow to buy what you cannot buy. There is a certain irony in the possibility that Washington's chip policy ends up protecting Chinese balance sheets from the thing it is doing to American ones.
What to Watch
Five things, in order:
- Hyperion's bonds. If an A+ rated, Meta-guaranteed structure cannot get back toward par, no SPV will price.
- The actual Texas statute. A moratorium on approvals is very different from a moratorium on construction. The first delays projects not yet permitted; the second strands ones already drawn on. The legislature convenes in January.
- CPI on Tuesday and the FOMC on December 9. A hike into a credit selloff is the scenario nobody's project model contemplated.
- Year-end insurer statutory filings. The first look at how much of this paper sits in annuity general accounts.
- The ING auction clearing price. It is small, but it will be the first public print on a mixed book of data-centre loans in this environment. If it clears in the 90s, the banks are tidying up. If it clears in the 80s, they are getting out.
Conclusions
For three years the debate about AI infrastructure has been framed as a question about returns: will the models earn enough to justify the spend? That is a fair question and I do not know the answer. But it is an equity question, and the asset class that now funds the build-out is not equity. Somewhere in 2026 the marginal dollar of AI capex stopped coming from a hyperscaler's cash flow and started coming from a bond fund, a life insurer's general account, a private credit vehicle, or an SPV whose only asset is a lease that has not started.
Credit does not ask whether AI will pay off eventually. It asks whether a specific borrower makes a specific payment on a specific date. A moratorium in Austin does not change the first question at all. It changes the second one a great deal, and the people it changes it for are not the ones who sold stock on Thursday. They are the ones who bought a 20-year bond at 225 over Treasuries because it had a famous tenant and an A+ rating, and who, in most cases, have never read the lease.
Sources and References
Market data and this week's trading
- CNBC. "Dow drops 1,800 points; Nasdaq tumbles into correction territory on AI tech wreck." 5 November 2026. https://www.cnbc.com/2026/11/04/stock-market-today-live-updates.html
- CNBC. "Dow soars 1,500 points for best day since April 2025 on clear-cut Democratic win." 4 November 2026. https://www.cnbc.com/2026/11/03/stock-market-today-live-updates.html
- CNBC. "Nasdaq starts November with 1% loss as tech melts down again." 2 November 2026. https://www.cnbc.com/2026/11/01/stock-market-today-live-updates.html
- Investrade. "Market Review: November 06, 2026." https://investrade.com/market-review-november-06-2026/
- The Closing Print. "Post-Market & Weekly Summary – Nov 6, 2026." https://www.theclosingprint.net/2026/11/post-market-and-weekly-summary-2026-11-06/
- Bloomberg. "Chip Stocks Climb After Rout, But AI Doubts Linger." 6 November 2026. https://www.bloomberg.com/news/articles/2026-11-05/stock-market-today-dow-s-p-live-updates
AI credit stress and issuance
- CNBC. "The AI 'credit card' is getting costlier as CoreWeave bond yield tops 16%." 6 November 2026. https://www.cnbc.com/2026/11/06/coreweave-bond-yield-tops-16percent-as-ai-credit-stress-rises.html
- Quartz. "CoreWeave's bond yield tops 16% as AI credit stress spreads wider." 6 November 2026. https://qz.com/coreweave-bond-yield-16-percent-ai-credit-stress-110626
- Goldman Sachs Exchanges. "How AI Debt Is Reshaping Credit Markets." August 2026. https://www.goldmansachs.com/insights/goldman-sachs-exchanges/how-ai-debt-is-reshaping-the-credit-market
- 404K Research. "The AI Buildout Enters the Debt Era: Goldman Sachs Estimates Hyperscalers Have Issued $296B in Debt YTD." 6 November 2026. https://404kresearch.substack.com/p/the-ai-buildout-enters-the-debt-era
- MLQ.ai. "Hyperscaler bond issuance reaches about $220 billion as AI financing costs rise." August 2026 (FactSet, S&P Global Ratings, BNP Paribas via Reuters). https://mlq.ai/news/hyperscaler-bond-issuance-reaches-about-220-billion-as-ai-financing-costs-rise/
- Bloom Energy. Form 10-Q for the quarter ended 30 September 2026, "Financing Availability for Data Centers." https://www.sec.gov/Archives/edgar/data/0001664703/000162828026066931/be-20260930.htm
- Forbes (Chastity Murphy). "The Banks Behind AI Are Protected. Your Retirement Is Not." 23 October 2026. https://www.forbes.com/sites/chastity-murphy/2026/10/23/the-banks-behind-ai-are-protected-your-retirement-is-not/
Oracle
- Oracle Corporation. Form 8-K, 2 November 2026 (common stock and notes offerings). https://www.sec.gov/Archives/edgar/data/0001341439/000119312526447507/d131790d8k.htm
- Oracle Corporation. Form 8-K exhibit 99.1, fiscal Q4 2026 results, 10 June 2026. https://www.sec.gov/Archives/edgar/data/0001341439/000119312526265848/orcl-ex99_1.htm
- Oracle Corporation. Form 8-K exhibit 99.1, fiscal Q3 2026 results (February 2026 capital raise, RPO). https://www.sec.gov/Archives/edgar/data/1341439/000119312526100148/orcl-ex99_1.htm
- CNBC. "Oracle files to sell $20 billion in new stock, bonds to finance AI buildout." 2 November 2026. https://www.cnbc.com/2026/11/02/oracle-files-to-sell-20-billion-in-new-stock-bonds-to-finance-ai-buildout.html
- Bloomberg. "Oracle Seeks $20 Billion for AI Growth as Meta Weighs New Debt." 2 November 2026. https://www.bloomberg.com/news/articles/2026-11-02/oracle-is-offering-10-billion-of-high-grade-dollar-bonds
- Quartz. "Oracle is raising $20 billion in bonds and stock for AI buildout." 2 November 2026. https://qz.com/oracle-10-billion-bond-sale-debt-cap-110226
Leases, SPVs and Hyperion
- Meta Platforms. Form 10-Q for the quarter ended 30 June 2026, "Leases and Contractual Commitments." https://www.sec.gov/Archives/edgar/data/0001326801/000162828026050705/meta-20260630.htm
- Meta Platforms. Form 10-Q for the quarter ended 31 March 2026. https://www.sec.gov/Archives/edgar/data/0001326801/000162828026028526/meta-20260331.htm
- Global Data Center Hub. "Hyperion's Number Is Not $50 Billion (It Is 80)." July 2026. https://www.globaldatacenterhub.com/p/the-hyperion-financing-is-the-signal
- Global Data Center Hub. "Meta Takes Hyperion To 5GW (And Off Its Balance Sheet)." July 2026. https://www.globaldatacenterhub.com/p/meta-scales-hyperion-to-5gw-and-over
- Quinn Emanuel. "Client Alert: Emerging Litigation Risks in Financing AI Data Centers Boom." August 2026 (Oracle SPV financings, Moody's commentary). https://www.quinnemanuel.com/the-firm/publications/client-alert-emerging-litigation-risks-in-financing-ai-data-centers-boom/
- Refacto. "The AI Buildout's Shadow Balance Sheet." August 2026. https://refacto.substack.com/p/hyperscaler-and-neocloud-off-balance
- Traders Union. "Meta lease exposure reshapes credit view on AI data center buildout" (citing the Financial Times). October 2026. https://tradersunion.com/news/financial-news/show/3762236-meta-lease-exposure-ai-datacenter-credit/
- Bloomberg via EnergyNow. "The $3 Trillion AI Data Center Build-Out Becomes All-Consuming For Debt Markets." February 2026. https://energynow.com/2026/02/the-3-trillion-ai-data-center-build-out-becomes-all-consuming-for-debt-markets/
- The State of AI. "Is AI Data Center Debt the Next Subprime?" 20 October 2026. https://www.thestateofai.com/news/ai-data-center-debt-subprime-pension-funds
Banks, insurers and private credit
- Insurance Journal (Bloomberg). "HSBC Seeks to Hedge Data Center Debt With Credit Risk Insurance." 30 October 2026. https://www.insurancejournal.com/news/international/2026/10/30/890898.htm
- Claims Journal (Bloomberg). "Barclays Turns to Insurers to Hedge Billions of Data Center Debt." 30 October 2026. https://www.claimsjournal.com/news/national/2026/10/30/340977.htm
- CNBC. "AI debt concerns grow as ING reportedly seeks buyers for $500 million of data center loans." 6 November 2026. https://www.cnbc.com/2026/11/06/ing-ai-data-center-loans-debt.html
- CNBC. "AI data center boom 'stress tests' insurers as private capital floods in." April 2026. https://www.cnbc.com/2026/04/06/ai-data-centers-financing-insurance-deals-gpu-debt.html
- Forbes (Dara Abasiita). "AI Debt Is Flowing Into Assets Insurance Regulators Struggle To Track." 8 October 2026. https://www.forbes.com/sites/daraabasiita/2026/10/08/ai-debt-is-flowing-into-assets-insurance-regulators-struggle-to-track/
- Forbes (Mayra Rodriguez Valladares). "The Half-Trillion-Dollar Silicon Gamble: Inside The AI Debt Boom." 22 October 2026. https://www.forbes.com/sites/mayrarodriguezvalladares/2026/10/22/the-half-trillion-dollar-silicon-gamble-inside-the-ai-debt-boom/
- Forbes (Mayra Rodriguez Valladares). "The AI Data Center Boom Has A Wall Street Risk Problem." 30 October 2026. https://www.forbes.com/sites/mayrarodriguezvalladares/2026/10/30/the-ai-data-center-boom-has-a-wall-street-risk-problem/
- AI Tech Model. "Who Holds AI Debt: 401(k)s, Insurers, BDCs and Banks." October 2026. https://aitechmodel.com/who-holds-ai-debt-retail-401k-insurers/
The midterms and data-centre policy
- CNBC. "Top Democrats call to rein in AI buildout after midterms sweep." 5 November 2026. https://www.cnbc.com/2026/11/05/democrats-data-center-restrictions-midterms.html
- CNBC. "AI data center stocks and Musk's SpaceX decline as Texas governor result raises risks of a slowdown." 4 November 2026. https://www.cnbc.com/2026/11/04/ai-stocks-texas-governor-data-centers.html
- CNN. "America got the anti-data-center wave it voted for. Here come the economic consequences." 5 November 2026. https://www.cnn.com/2026/11/05/economy/data-center-texas-election
- Fortune. "Democrat wins will slow data center construction and force tech to the negotiating table." 4 November 2026. https://fortune.com/2026/11/04/democrats-midterm-election-2026-anti-data-center-governors-texas-georgia-ohio-ai-infrastructure-big-tech-spending-moratorium-regulations/
- Fox Business. "How Democrats' midterm gains could reshape the political landscape for AI" (UBS estimates). 5 November 2026. https://www.foxbusiness.com/politics/new-political-landscape-facing-ai-after-democrats-midterm-gains
- Yahoo Finance. "Democrats' victories in Texas laid bare voter unease over data centers." 4 November 2026. https://finance.yahoo.com/economy/article/democrats-victories-in-texas-laid-bare-voter-unease-over-data-centers--and-signal-new-hurdles-for-the-industry-142918051.html
Rates and the Fed
- Quartz. "Richmond Fed president said more interest rate hikes are coming — but gradually." 5 November 2026. https://qz.com/richmond-fed-barkin-interest-rate-hikes-gradual-110526
- Invezz. "Fed rate hike odds jump after October jobs report as officials signal more tightening." 6 November 2026. https://invezz.com/en-ae/news/2026/11/06/fed-rate-hike-odds-jump-after-october-jobs-report-as-officials-signal-more-tightening/
- NPR. "Fed expected to hold interest rates steady but a hike is anticipated by year's end." 28 October 2026. https://www.npr.org/2026/10/28/nx-s1-6017978/fed-expected-to-hold-interest-rates-steady-but-a-hike-is-anticipated-by-years-end
- Rallies.ai (Reuters summary). "Fed's Hammack, Barkin favor gradual rate-hike path to lower inflation." 5 November 2026. https://rallies.ai/news/feds-hammack-barkin-favor-gradual-rate-hike-path-to-lower-inflation-4db698c746b617ca
Asia
- FourWeekMBA. "SoftBank Bond Falls Most in 10 Weeks as AI Rout Hits Credit" (FINRA TRACE data). 6 November 2026. https://fourweekmba.com/ai-softbank-bond-falls-most-in-10-weeks-as-ai-rout-hits-credit/
- ChainCatcher. "AI stock sell-off spreads to the credit market, SoftBank leads the decline in Asian dollar bonds." 6 November 2026. https://www.chaincatcher.com/en/article/2302342
- Daily Guardian (Reuters). "Investors dive into bonds as global AI tech selloff continues." 6 November 2026. https://dailyguardian.ca/investors-dive-into-bonds-as-global-ai-tech-selloff-continues/
- Moneyweb (Bloomberg). "Tencent tells investors its $5bn bond mainly for AI capex." October 2026. https://www.moneyweb.co.za/news/companies-and-deals/tencent-tells-investors-its-5bn-bond-mainly-for-ai-capex/
- Briefs Finance. "Tencent taps dollar and offshore yuan markets for roughly $5B to fuel AI buildout." October 2026. https://www.briefs.co/news/tencent-taps-dollar-and-offshore-yuan-markets-for-roughly-5b/
Delta Backwaters is an independent student-led research initiative. All materials published are strictly for informational and educational purposes and do not constitute investment advice.