Newsletter · · Ashutosh Agarwal

Data-Center IPOs Stall as the 10-Year Yield Tops 5% and Bond Sales Hit a Record - The Capital-Markets Reopening - Week of September 29, 2026

The Capital-Markets Reopening for the week of September 29, 2026. Podcast synthesis on SB Energy, Holtec and AgriCo postponing IPOs as Oracle invokes force majeure, Oura and NScale heading to market, Goldman's record $2.3 trillion investment-grade bond forecast, the Paramount and Warner Bros. Discovery settlement, strain in private credit, and a 10-year Treasury yield above 5%.

The Capital-Markets Reopening

Week of September 29, 2026: Data-Center IPOs Stall as the 10-Year Yield Tops 5% and Bond Sales Hit a Record

Week of September 22 – September 29, 2026

TL;DR

  • The IPO window is shutting on AI companies with a story but no profits. SB Energy, the SoftBank-backed data-center developer, has postponed its IPO after its bankers "couldn't find enough buyers" at a $50 billion valuation. Nuclear-services firm Holtec and power producer AgriCo have delayed too, and Oracle rattled the whole group by invoking a legal escape clause on a big data-center project (Prof G Markets, September 28). This week the test is smart-ring maker Oura, which is set to raise $2.2 billion at a $15.6 billion valuation. It is a real business growing 123%.
  • Bond sales are running at record pace, and Goldman and Morgan Stanley are getting paid for it. Goldman's own credit team raised its 2026 forecast for U.S. investment-grade bond sales (bonds from the safest companies) to a record $2.3 trillion. It expects the big cloud companies alone to borrow $400 billion in 2027 (Alpha Exchange, September 23). Goldman and JPMorgan are leading the NScale IPO, and Morgan Stanley is third in a 26-bank syndicate (Run the Numbers, September 24).
  • Antitrust is no longer the thing that kills big deals. Interest rates might be. Paramount settled the state lawsuit against its roughly $110 billion Warner Bros. Discovery deal without selling off a single network. A Bloomberg Intelligence litigation analyst says 2026 has seen zero federal merger challenges without a settlement attached (Bloomberg Law, September 26). Meanwhile the 10-year Treasury yield rose to 5.1–5.16%, its highest since 2007, and a Goldman trading executive called the bond market the "number one kind of clear and present danger" (The Markets, September 25).

What's new

How to read this. Some speakers work inside the business they're talking about. We flag them as operator/insider. Everyone else is marked as an analyst, journalist or commentator (a pundit). Items are ranked by how useful they are for someone trading these stocks. Every podcast below aired between September 22 and September 28.

1. The AI-IPO window is splitting in two: real revenue gets through, big promises don't. The most useful market signal this week came from Prof G Markets, "Investors Are Turning Against Data Centers (Here's Why)" (September 28), with host Ed Elson and NYU professor Scott Galloway (both pundits). Citing the New York Times, Elson reported that SB Energy's bankers "couldn't find enough buyers at the company's targeted valuation of $50 billion." SB Energy was supposed to list this month. Now it will wait "until investor sentiment towards data centers improves."

Galloway ran the numbers, and they are brutal:

  • $214 million in revenue, down 8%
  • A $50 billion ask, or roughly 234 times sales
  • Zero AI data centers in operation
  • Only about 9% of contracted capacity actually under construction

His line: "they're trying to get a $50 billion valuation on a PowerPoint with a SoftBank logo."

SB Energy isn't alone. Holtec, the nuclear-services company, has postponed indefinitely, and its CEO said the business is "viewed as connected to data centers." Power producer AgriCo is also delaying. On Squawk on the Street (September 22), CNBC noted this is SB Energy's second delay.

The trigger was Oracle. It sent a "force majeure" notice (a legal clause that excuses a party when events outside its control stop it from performing) to the developer of its Project Jupiter data center. Oracle wants the option to delay payments if the site isn't running by 2028. Oracle stock fell 5%, and Blue Owl, which is financing the project, fell 7%. Galloway's reading: "when Oracle invokes force majeure, what they're really saying is that the God is credit markets." He also cited $68 billion of data-center projects blocked or delayed in three months, with 30–50% of this year's capacity facing delays.

The key line for bankers:

"When the bankers can't find a book, it means the smart money has stopped buying the story."

The other side of the split is Oura. The smart-ring maker plans to go public this week at a $15.6 billion valuation, raising $2.2 billion. It has 123% growth and 83% subscription renewals. Galloway predicted it will be "really well received."

Why it matters: this is the most direct signal yet on the ECM (equity capital markets, meaning the business of underwriting stock sales) fee pool into year-end. Buyers are sorting companies with cash flow from companies with "potential." If Oura trades well, the window stays open for quality names. If it breaks, expect Goldman and Morgan Stanley's fourth-quarter underwriting calendar to thin out.

2. NScale's IPO filing puts Goldman and JPMorgan at the top of the next big AI listing, on the NYSE. On Run the Numbers, "The NScale IPO | A Neocloud S1 Breakdown" (September 24), the host (pundit) went through the filing of NScale. It is a two-year-old British "neocloud," a company that rents out Nvidia chips by the hour. It will list on the New York Stock Exchange (owned by ICE) under the ticker NSCL. Goldman and JPMorgan are "lead left," the top underwriting spots that usually earn the biggest fees, "followed by Morgan Stanley and then 23 banks."

According to the Financial Times, the target valuation is about $35 billion, up from $14.6 billion in March. The headline number is $103.4 billion of signed contracts, including up to $43.8 billion from Microsoft and up to $44.6 billion from Anthropic. But only $2.6 billion of that, about 2%, is live today.

The financials behind it:

  • First-half revenue of $141 million, up 1,252%
  • Gross margin of negative 159% after depreciation
  • A net loss of "$1 billion and change"
  • Only $137 million of cash at June 30

Since then the company has lined up a lot of debt: an $830 million revolving credit line, $4.2 billion of new facilities, $2.5 billion of Dell equipment financing and $3.1 billion of convertible notes.

Why it matters: this is a triple fee event for the banks: the IPO, plus billions in debt and convertible deals around it. It is also a hard test of item 1. NScale looks more like SB Energy (big contracts, little live revenue) than like Oura. Its closest comparison, CoreWeave, aimed for $35 billion at its IPO and priced at $23 billion, according to the podcast's summary. The deal is live for the lead banks, but the price could come in well below target.

3. Goldman's credit chief: record bond sales, and the AI borrowing wave is only getting started. On Alpha Exchange, "Amanda Lynam, Chief Credit Strategist in Global Investment Research, Goldman Sachs" (September 23), Amanda Lynam (Goldman insider, though an analyst rather than a banker) gave the most detailed numbers of the week on the DCM (debt capital markets, the business of underwriting bond sales) boom.

Goldman just raised its 2026 forecast for U.S. investment-grade bond sales to $2.3 trillion, up from $2.1 trillion. That would be a record. Lynam said the team had held back because "we anticipated a summer slowdown that never materialized." For 2027 it pencils in $2.4 trillion.

The hyperscalers (Amazon, Microsoft, Google, Meta and the other giant cloud companies) are the engine:

  • Borrowing: $108 billion of global investment-grade bonds in 2025, $229 billion so far in 2026, a forecast of $250 billion for the full year and $400 billion in 2027
  • Spending: planned capex (spending on buildings, equipment and chips) of "upwards of $6 trillion" from 2026 through 2030
  • Share of the market: the hyperscalers are only about 40% of all AI-related bond sales

Her caution is about who buys the debt, not whether the companies can borrow. "We're not concerned at all, actually, about access to capital." But the regular bond market won't "do the entirety of the heavy lifting," and from 2028 private markets will have to take more.

The same week, CNBC reported Goldman is leading a junk-bond sale (bonds from riskier borrowers) of data-center debt tied to CoreWeave, yielding "in the 11s," meaning about 11% (Squawk on the Street, September 22).

Why it matters: last week, Bank of America's CEO said issuers were "holding back." Goldman's own data says the bond calendar didn't slow. It sped up. For GS, MS and JEF, underwriting fees from investment-grade and AI-infrastructure bonds are the steadiest fees in the business right now.

4. Paramount–Warner settled cheaply, and the antitrust regime is now openly deal-friendly. On Brew Markets, "Meta's New Muse & Paramount Cleared to Buy Warner Bros." (September 22), the hosts (pundits) laid out the settlement with California Attorney General Rob Bonta. Paramount must:

  • Spend an extra $300 million a year on U.S. production for five years
  • Release 30–32 films a year
  • Set up a five-person editorial-independence board over CBS News and CNN
  • Not sell its studio lots for five years

Paramount CEO David Ellison (operator/insider): "Bringing Paramount and Warner Brothers Discovery together will build a stronger Hollywood."

On the financing, Squawk on the Street (September 22) put the equity at $47 billion: $22 billion from the Ellison family and RedBird, and $25 billion from Saudi, Qatari and Abu Dhabi sovereign wealth funds. The FCC raised the foreign-ownership cap to 49.5% from 25%. A $49 billion bridge loan (short-term debt meant to be replaced with bonds) is to be refinanced by Bank of America, Apollo and Citi. Closing was described as "a couple of weeks away."

The bigger point came on Bloomberg Law, "Weekend Law: White House Media Ban, Paramount Settles & The Arch" (September 26). Jennifer Rhee, senior litigation analyst at Bloomberg Intelligence (pundit), called the settlement "widely favorable, in my view, to Paramount." She noted the 30-film promise means release, not make, which is "really not that different from what they're already each individually doing." She said rumors had DOJ staff opposing the settlement, "and the decision was made over their head." Under Biden, she said, "there probably would have been either... a challenge that was settled with a stronger settlement than this... or they would have actually gone to court." She also noted the combined company is expected to carry about $80 billion of debt.

On The Capitol Forum, "Antitrust in Trump 2.0 (Second Request)" (September 25), two former DOJ antitrust lawyers split the regulators in two. Karina Lubel of Brunswick said the DOJ is settling "early and often." Its new antitrust chief has reportedly questioned whether blocking a merger violates a company's due-process rights. The FTC, by contrast, is "business as usual" on merger review. Host Teddy Downey summed up the view in the market: dealmakers believe they can "just hire the right lobbyist with Trump ties and get my deal through," so the real risk now sits with states and overseas regulators. Wiley's Nick Giolis pushed back that betting on this would be "a catastrophically bad miscalculation."

Why it matters: for the advisory fee pool, a faster, cheaper path through regulators means more big deals get announced and more get completed. The risk to watch is politics. The same podcast flagged polls showing Democrats favored to win Congress in the midterms, which could bring tougher enforcement in 2027.

5. The 10-year at 5%+ is now the main risk, and Goldman's own people are saying so. On The Markets, "Why the Bond Market May Be the Stock Market's Biggest Risk" (September 25), recorded on Goldman's trading floor, Tony Kim, Goldman's global head of hedge fund coverage in its FICC and equities business (operator/insider), was blunt. FICC is fixed income, currencies and commodities, the bond-trading side of the bank.

  • "If you were to ask me what's the number one kind of clear and present danger for the stock market, I'd say it is the bond market."
  • Futures markets now imply "about four hikes, maybe a touch more."
  • Inflation has been above the Fed's target for 66 months.
  • Goldman forecasts 3.3% GDP growth for the third quarter.
  • Hyperscaler capex is going from about $150 billion in 2023 to "probably $1.3 trillion" next year.
  • His suggested hedge for stock risk is betting against bonds.

On the same day, Goldman Vice Chairman and former Dallas Fed President Robert Kaplan (Goldman insider) took a calmer view on Exchanges, "Why Markets May Be Pricing in Too Many Fed Rate Hikes" (September 23). He would "skip October," hike again in December to 4%–4.25%, which he sees as roughly "neutral," and stop there. For deals, his key point was this: "the Fed is doing really isn't going to slow down the AI build." CEOs "are much more focused on being able to issue equity or on the Treasury curve," and credit spreads have been "reasonably well behaved."

The damage is falling on smaller, rate-sensitive borrowers. On Yet Another Value Podcast, "September 2026 Random Ramblings" (September 25), host Andrew Walker (pundit) pointed to the 10-year at 5.1%, "moved up fast over the past three weeks." He showed how that changes merger math. A safe deal priced at $100 that closes in a year now trades around $94–95, compared with $98–99 in 2015. He also warned about a wave of refinancing: corporate debt issued in 2021–23 at 3–5% will roll over in 2028–31 at possibly 8–10%.

Why it matters: higher rates cut both ways for our names. They help trading desks (GS, MS). They hurt leveraged buyouts and small-company financing (JEF, EVR, MC). And they eventually create restructuring work (MC, EVR).

The debate

The question hasn't changed: is this a durable, multi-year reopening, or a fragile head-fake? This week gave each side new evidence.

The durable case. The biggest pool of fees, underwriting debt, is running at record levels. Goldman expects a record $2.3 trillion of investment-grade bond sales this year and $2.4 trillion next, with hyperscaler borrowing nearly doubling to $400 billion in 2027 (Alpha Exchange, September 23). Megadeals are getting through regulators, and cheaply: Paramount–Warner cleared with no forced sales of any business (Bloomberg Law, September 26). New deals keep coming. Royal Caribbean agreed to buy 50% of Sandals Resorts for about $3 billion, roughly 10 times next year's EBITDA (a standard measure of operating profit), funded with committed debt from Morgan Stanley (Behind the Stays, September 25). Goldman's Kaplan argues the market is pricing too many hikes and that Fed policy "isn't going to slow down the AI build" (Exchanges, September 23). IPO investors aren't on strike. They are being picky, which is healthy. Oura, a real business, is coming this week.

The fragile case. The IPO wave everyone counted on is AI-heavy, and it is jamming up. Anthropic slipped to November last week. This week SB Energy, Holtec and AgriCo postponed, OpenAI is described as delaying, and Oracle reached for a force majeure clause (Prof G Markets, September 28). Private credit, the industry that has been funding buyouts and data centers, is facing a run on its investor-friendly fund wrappers. Blackstone's $82 billion B-Cred fund got redemption requests for 10% of its shares, about $8 billion, against a 5% cap. It has had four straight quarters of markdowns to NAV (net asset value, the stated value of the fund's holdings). Marlton's James Elbaor (operator/insider, a private-credit investor) said "the product was sold as if the gate did not exist" (Monetary Matters, September 27). GMO's Henry Peabody (pundit) argues the Fed will have to hike "far higher than expected," because big AI borrowers and private credit don't respond to rate moves, so the pain lands on weaker borrowers (Monetary Matters, September 24). Biotech investors describe a "quieter" M&A market as rates rise (Biotech Hangout, September 25).

My read. Galloway's framing is the right one. This is a "K-shaped" market (one where the strong keep rising while the weak fall). Companies with cash flow get financed, and companies with only "potential" are "about to get very familiar with the term cost of capital." That split is exactly why the diversified banks look better than the pure advisers right now. Goldman and Morgan Stanley get paid for the record investment-grade bond calendar, for trading a jumpy bond market, and for lead spots on whatever IPOs do get done. The pure advisers (Evercore, Moelis) need deals to close, and the deals that are stalling are exactly the sponsor-backed and speculative ones. The deciding data point is this week's Oura IPO. A strong debut says the window is open for quality. A weak one says the head-fake camp is winning.

Stocks in play

  • Goldman Sachs (GS).
    • Bull: The most exposed to the parts of the market that are working. It is lead-left on NScale (Run the Numbers, September 24) and leading the CoreWeave-linked junk-bond deal (Squawk on the Street, September 22). Its research team is forecasting a record bond calendar (Alpha Exchange, September 23). Its trading desk is watching a volatile bond market, and volatility tends to lift trading revenue (The Markets, September 25). It is also reportedly bidding on "a large CLO provider" (CLOs are funds that pool corporate loans) (Monetary Matters, September 27).
    • Bear: Its IPO calendar leans heavily on AI names that buyers are now rejecting. NScale is gross-margin negative, and the Anthropic fee has already slipped.
    • Catalyst / number to watch: NScale pricing against its roughly $35 billion target. Third-quarter results in mid-October, mainly trading revenue and debt-underwriting fees.
  • Morgan Stanley (MS).
    • Bull: It is providing committed debt financing for Royal Caribbean's roughly $3 billion Sandals deal (Behind the Stays, September 25), and it holds the third spot on NScale's 26-bank syndicate (Run the Numbers, September 24). Its wealth arm's Jim LeCamp (insider) says the market is handling the hawkish Fed "due to strong earnings, revenues and margins" (Squawk on the Street, September 24).
    • Bear: It shares the exposure to a stalled AI-IPO pipeline, including Anthropic.
    • Catalyst / number to watch: Third-quarter equity-underwriting revenue and wealth net new assets (new client money coming in). Whether Oura and NScale price well.
  • Intercontinental Exchange (ICE).
    • Bull: The NYSE, which ICE owns, won the NScale listing (ticker NSCL) (Run the Numbers, September 24). Heavy bond-market volatility and record corporate bond sales support ICE's fixed-income data and trading businesses.
    • Bear: Mortgage rates of 6–7% and above, now that the 10-year is over 5%, keep its mortgage-technology unit stuck in a frozen housing market (Monetary Matters, September 24). Postponed data-center IPOs mean fewer new listings.
    • Catalyst / number to watch: NSCL's first day of trading; the listings count going into the fourth quarter.
  • Nasdaq (NDAQ).
    • Bull: Stocks are near record highs, and the IPO pipeline beyond AI (Oura this week, Whoop targeting an IPO within 18 months) is healthy (Prof G Markets, September 28).
    • Bear: It lost the marquee NScale listing to the NYSE. The AI-IPO backlog that should feed its listings business is postponing.
    • Catalyst / number to watch: Which exchange Oura lists on; any update on tokenized stocks after the SEC's move last week.
  • Evercore (EVR).
    • Bull: Megadeals are clearing antitrust review easily (Bloomberg Law, September 26). A coming wave of private-credit and BDC mergers could create new advisory work. BDCs are business development companies, listed funds that lend to mid-sized firms. Elbaor: "I fully expect to see a lot of merger activity within the BDC space and private credit space," likely starting next fiscal year (Monetary Matters, September 27).
    • Bear: A pure adviser with no trading or lending income to fall back on. Wider merger-arbitrage spreads show buyers now demand more reward for the time a deal takes to close (Yet Another Value Podcast, September 25).
    • Catalyst / number to watch: Third-quarter advisory revenue and the pace of deal completions.
  • Moelis (MC).
    • Bull: The best restructuring play among our names. Private credit is in trouble (B-Cred gating, four quarters of markdowns), GMO warns of stress for weaker borrowers (Monetary Matters, September 24), and a refinancing wall is building for 2028–31 (Yet Another Value Podcast, September 25). Galloway's "dark fiber" comparison (the telecom overbuild before the 2001 bust) is the kind of setup that eventually creates restructuring work.
    • Bear: Restructuring fees come later. Before then, the sponsor deals it depends on are the ones stalling.
    • Catalyst / number to watch: Restructuring's share of revenue; any stress in the data-center or neocloud debt markets.
  • Jefferies (JEF).
    • Bull: It has leveraged-finance exposure to a junk market that is still open for AI-infrastructure deals, with CoreWeave-linked debt pricing in the 11s (Squawk on the Street, September 22), and a high-yield market where new bonds are coming at 8–10% (The NAVigator, September 25).
    • Bear: The most exposed of our names to lower-quality borrowers and private-credit strain, which this week's podcasts showed spreading.
    • Catalyst / number to watch: Its fiscal third-quarter results (quarter ended August), the group's first report.

Read-throughs

  • Boutique advisers (EVR / MC / JEF). Two forces are pulling in opposite directions. Regulators are easy: no big federal merger challenges, and settlements "early and often" (The Capitol Forum, September 25). Financing is harder: a 10-year above 5% and wider merger-arbitrage spreads (Yet Another Value Podcast, September 25). Deal flow shows up in mid-market and strategic deals: Royal Caribbean–Sandals (Behind the Stays, September 25), and Luca Mining's up-to-$385 million purchase of a Mexican copper mine from Capstone, with Scotiabank advising the seller (Company Interviews, September 26). Bank-to-bank M&A has a new tailwind: the FDIC has proposed speeding bank-merger reviews to as few as five days, and PNC and Citizens say they're open to selective deals (The Banker Next Door, September 23). That helps the financial-institutions advisory teams at EVR and JEF.
  • Leveraged-finance and private-credit lenders. This is where the cracks are. Blackstone's B-Cred got $8 billion of redemption requests against a 5% gate. BlueRock's fund (BPRE) traded at a 38% discount to NAV when it listed. Blue Owl's OBDC stopped offering quarterly withdrawals and is winding down its portfolio. Asset-manager stocks are down sharply this year: Apollo and Blackstone about 12%, KKR 22% and Blue Owl 35% (Monetary Matters, September 27). Blue Owl also fell 7% on Oracle's force majeure notice (Prof G Markets, September 28). The silver lining for banks: as private credit pulls back, borrowers go back to the public bond and loan markets, which banks underwrite. Goldman's Lynam expects public investment-grade bonds to do "a lot more of the heavy lifting in 2027 and 2028" (Alpha Exchange, September 23). Aberdeen's Matt Kence notes data-center debt is only mid-single digits of the high-yield market but 20%+ of private credit (The NAVigator, September 25). In other words, the AI concentration risk sits with private lenders, not the public junk market.
  • PE sponsors cashing out. The IPO exit route is open only for quality. Oura (123% growth) goes this week, while SB Energy and similar story-stocks can't find buyers (Prof G Markets, September 28). Elbaor says private-equity marks haven't had their "leg drop" yet the way private-credit marks have (Monetary Matters, September 27). That suggests sponsors still face a gap between what they think companies are worth and what buyers will pay. Outside the U.S., the Dangote refinery IPO, billed as Africa's largest ever, is trying to raise up to $2 billion through October 13 (Next Africa, September 25).
  • Exchange listings and data revenue (ICE / NDAQ). The NYSE won NScale (Run the Numbers, September 24). But the listings pipeline from data-center and power companies (SB Energy, Holtec, AgriCo) is delayed. Shein is now aiming for Hong Kong after failed attempts in New York and London, at an expected $40–50 billion valuation, down from $60–100 billion (Networth and Chill, September 23). Recurring data revenue is the steadier story. A volatile bond market with record new issuance is good for fixed-income data and trading volumes.
  • Macro and rates. The Fed is signaling more hikes. Philadelphia Fed President Anna Paulson says more are likely warranted, and CNBC's Steve Liesman says the center of the Fed board now favors more (Squawk on the Street, September 24). Treasury's surprise bond-buyback plan failed to push long-term yields down, and gross U.S. debt has reached $40 trillion (WSJ's Take On the Week, September 27). PIMCO's Tiffany Wilding says Warsh now steers by "financial conditions," not the neutral rate. That means a stock market up 13% this year is itself a reason for him to keep hiking (PIMCO Pod, September 25).

What changed vs. last week

Last week's issue was built around the Fed's first hike in three years, Bank of America CEO Brian Moynihan's warning that issuers were "holding back" deals, and Anthropic pushing its IPO to November. Here's what moved:

  • The IPO slowdown spread beyond Anthropic. Last week it was one delayed mega-IPO. This week it's a pattern: SB Energy (postponed again, now indefinitely), Holtec, AgriCo, OpenAI described as delaying, and Oracle's force majeure notice (Prof G Markets, September 28). The "gray cloud over AI" we flagged last week now has specific names and dollar figures.
  • Partly contradicting Moynihan: the bond calendar didn't pause. Last week, issuers were said to be pulling deals. This week Goldman's research team raised its full-year investment-grade forecast to a record $2.3 trillion, saying the expected slowdown "never materialized" (Alpha Exchange, September 23). The pullback is showing up in speculative equity, not in high-grade debt.
  • Oura got bigger. Last week it had filed at an "$11 billion-plus" valuation. This week it is set to price at $15.6 billion, raising $2.2 billion (Prof G Markets, September 28). Quality companies are being rewarded.
  • Paramount–Warner went from settlement talks to a signed settlement. The final terms are lighter than the $1.5 billion California commitment floated last week: $300 million a year for five years, 30–32 films released a year, and no divestitures (Brew Markets, September 22). A federal judge still has to approve the consent decree (Breaking Points, September 23).
  • Rates went higher still. The 10-year moved from "back over 5%" last week to 5.1–5.16% (Yet Another Value Podcast, September 25; The Peter Schiff Show, September 27). Futures now price about four hikes (The Markets, September 25). A new counterweight came from Goldman's Kaplan, who argues the market is pricing too many (Exchanges, September 23).
  • New this week: private credit's troubles went mainstream. Last week J.P. Morgan's credit team warned the cushion was "thin." This week there were hard numbers on redemptions and gates at B-Cred, OBDC and BPRE (Monetary Matters, September 27).