Oracle just raised $25 billion with a record $129 billion order book — and still had to pay 40–58 basis points extra versus its last deal. Alphabet raised $32 billion across three currencies in under 24 hours, including the largest corporate bond sale in sterling market history. Meanwhile, US investment-grade spreads sit at their tightest level since 1998. When Vanguard’s co-head of credit told Bloomberg on February 20 that investors are “full,” it wasn’t a warning. It was a confirmation of the trade that sophisticated capital had been running since January.
February 21, 2026 · Credit Markets · Fixed Income · Institutional / Hedge Fund
Tags: investment-grade credit, AI bond issuance, credit spreads 2026, Vanguard credit outlook, Oracle bond deal, Alphabet bond deal, CDX payer swaptions, CDS, hyperscaler debt, spread widening, corporate bond market
By Navnoor Bawa · LinkedIn · YouTube: The Mathematical Trader
I. The Signal: What Vanguard Actually Said, and What It Means for Market Structure
On February 20, 2026, Bloomberg published a direct interview with Arvind Narayanan, Vanguard’s co-head of investment-grade credit and senior portfolio manager. His firm’s approximately $11.6 trillion in global AUM as of September 2025 makes it the largest or second-largest buyer in the US investment-grade primary market on any given week. Three statements carried operational weight beyond their surface reading.
“If there is broader macro weakness, broader hiccups in the market, then US IG could underperform because investors are full.”
— Arvind Narayanan, Co-Head of Investment-Grade Credit, Vanguard · Bloomberg, Feb 20, 2026
The second: “Hiding up in quality in times of uncertainty may not necessarily play out well this go-around.” The third: Vanguard is “looking really at global markets” to diversify away from US investment-grade.
Read for market-structure implications rather than portfolio guidance, these three statements mean one thing: the largest passive backstop in the primary market is reducing its absorption capacity — and saying why publicly. When the backstop bid withdraws, new-issue concessions widen. When concessions widen, secondary spreads follow. The trades that profit from this sequence were already live before Narayanan spoke. His interview was not the catalyst. It was the confirmation.
Apollo’s Torsten Slok had already documented the divergence signal on January 21 — a full 30 days earlier. Oracle’s CDS had already moved from 36 basis points to 139 basis points. The concession mechanism had already printed live in Oracle’s deal on February 2. By the time Vanguard confirmed the thesis publicly, the entry had been open for a month.
II. The Setup: Historically Extreme Valuations Meet a Record Supply Wave
Every trade in this piece rests on a verifiable, data-anchored market structure. Two variables define it: where spreads are, and what supply is coming.
Spread Valuations: The 2nd Percentile of a 20-Year Lookback
The Bloomberg US Investment-Grade Corporate Index option-adjusted spread (OAS) ended 2025 at 78 basis points — sitting in the 2nd percentile of a 20-year lookback, meaning spreads have been tighter on only 2% of all trading days over two decades. (Breckinridge Capital Advisors, Q1 2026 Corporate Bond Market Outlook)
By late January 2026, US investment-grade spreads compressed further to 71 basis points — the lowest reading since 1998, per Bloomberg index data. Simultaneously, yield premiums on global corporate debt across currencies and ratings fell below one percentage point for the first time since 2007, according to a Bloomberg gauge, while spreads on Asian investment-grade dollar notes touched a record low in the same week. (Bloomberg via Yahoo Finance, Jan 23, 2026)
Two independent Street-level forecasts anchor the widening thesis quantitatively. J.P. Morgan Global Research targets 110 basis points for US high-grade spreads by year-end 2026 — a 55% move from January 2026 levels. (J.P. Morgan 2026 Market Outlook) TD Securities’ Hans Mikkelsen targets 95 basis points, implying a 34% move from those same January levels. (Bloomberg, Jan 23, 2026) These are separate estimates from separate firms. The gap between them — 95 versus 110 basis points — is itself a calibrated measure of uncertainty in the widening call. Both are directionally identical.
The AI Bond Supply Wave: $400 Billion From Hyperscalers Alone
Morgan Stanley estimates hyperscalers will borrow $400 billion in US bonds in 2026, up from $165 billion in 2025 — a 142% sector-level increase in a single year. Morgan Stanley projects total US investment-grade gross issuance reaching a record $2.25 trillion for the full year. (Euronews, Feb 10, 2026; Breckinridge, citing Morgan Stanley Nov 2025 research)
Separately, TD Securities’ Mikkelsen forecasts up to $500 billion of AI-related issuance this year alone and projects that if realised, total US investment-grade offerings could reach $2.1 trillion or higher — TD’s own estimate, independent of and lower than Morgan Stanley’s. (Bloomberg, Jan 23, 2026) In Mikkelsen’s own words: tech companies “could issue more bonds for AI spending than the market can absorb at these tight spread levels.”
Demand is deep but finite. $490 billion flowed into taxable bond funds in 2025; foreign investors net-purchased $304 billion in US corporate bonds in the 12 months through October 2025. (Breckinridge, citing ICI and US Treasury International Capital data) The absorption pool is large. The incoming supply is larger.
The Concession Mechanism: Already Visible in Oracle and Alphabet Deal Data
Oracle’s $25 billion eight-part deal on February 2 set a record: Bloomberg confirmed it attracted more than $129 billion in orders — the most ever for such an offering. IFR, citing one fixed-income investor, reported bids reaching as high as $155 billion at the intraday peak before books closed. Despite that record demand, the deal still priced 40–58 basis points wider than Oracle’s September 2025 deal across equivalent tenors. Record demand coexisted with a large concession requirement. That is the new-issue concession dynamic in operation: when buyers are saturated with a credit, they require a spread premium to absorb even a record-setting book.
Alphabet’s multi-currency raise, executed February 9–10, 2026, raised approximately $32 billion in less than 24 hours across USD, sterling, and Swiss franc tranches. (Bloomberg via swissinfo.ch, Feb 10, 2026) The USD tranche was $20 billion, upsized from $15 billion, with over $100 billion in orders. The sterling tranche — which included a 100-year century bond, the first issued by a technology company since Motorola in 1997 — raised £5.5 billion ($7.5 billion), the largest corporate bond sale in the history of the sterling market, surpassing the previous record of £3 billion set by National Grid in 2016. (Bloomberg via swissinfo.ch, Feb 10, 2026; GlobalCapital, Feb 10, 2026)
“I think the fact that a 100-year bond comes out, you can’t get much more frothy than that.”
— Bill Blain, Windshift Capital · CNBC, February 12, 2026
Oracle and Alphabet together raised $57 billion in new investment-grade paper in three weeks. Both deals required concessions despite record oversubscription. The market was telling you in early February what Vanguard confirmed on February 20.
III. Five Executable Trades — All Live Before Vanguard Spoke
Trade 1 — Oracle 5-Year CDS: The AI Capex Bubble Hedge
Oracle’s 5-year credit default swap spread rose from approximately 36 basis points in June 2025 to 128 basis points by December 3 — the highest level since March 2009, per ICE Data Services prices cited by Bloomberg. That December 3 reading was the closing level after some compression from an earlier intraday peak: Bondblox confirmed the CDS had hit its highest level since 2009, with the peak print reaching approximately 139 basis points at its widest. The arithmetic across sources reconciles cleanly: Voya Investment Management, which publicly named Oracle the preferred AI capex cycle hedge in its December 2025 research note, stated the CDS had risen “roughly 310% since end of June, pushing perceived credit risk to a 16-year high.” A 310% move from 36 basis points implies a peak of approximately 148 basis points — directionally consistent with Bondblox’s ~139 basis point figure, given that these reflect different intraday and closing timestamps across the same volatile week, not a contradiction.
Trading volume in Oracle’s CDS reached an estimated $5 billion in seven weeks through November 2025, against near-zero activity in the prior period. Morgan Stanley warned Oracle’s CDS could approach 200 basis points — near the 2008 all-time high.
After Oracle priced its February $25 billion deal and confirmed no additional 2026 issuance, CDS spreads compressed sharply and Barclays upgraded Oracle’s debt to overweight. Traders who bought protection in the 36–50 basis point range in mid-2025 and closed in the 128–139 basis point range realised substantial gross P&L — the precise size depending on entry timing. Oracle carries over $100 billion in total debt (rated Baa2/BBB/BBB), making it the largest non-bank issuer in the Bloomberg Corporate Index. That is why its CDS functioned as a sector barometer for the entire AI capex cycle, not merely a single-name hedge.
Trade 2 — CDX.NA.IG Payer Swaptions: Asymmetric Index Protection Against Spread Widening
The index-level expression of the spread-widening thesis is buying protection on the CDX.NA.IG 5-Year Index (125 North American investment-grade reference names) via payer swaptions — options to enter a protection-paying position at a fixed strike.
At 2nd percentile valuations, the carry cost of outright CDX protection is punishing: you pay the running spread daily while waiting for widening. A 3–6 month payer swaption struck 10–15% out of the money costs a fraction of the carry drag while the payout on a spread-gap event is substantially asymmetric. TD’s Mikkelsen’s 95 basis point target implies a 34% spread widening from January 2026 levels; J.P. Morgan’s 110 basis point target implies 55%. Either scenario generates multi-hundred basis points of P&L on a well-structured payer swaption.
The trigger mechanism Narayanan identified is concrete: $400 billion of hyperscaler paper hitting a market where the largest passive buyer is openly reducing absorption capacity forces concession widening, which transmits from primary to secondary within hours, repricing CDX IG. Oracle’s February deal — priced 40–58 basis points wider than its September precedent despite a record order book — demonstrated this transmission mechanism empirically before the Vanguard interview was published.
Trade 3 — Long iTraxx Europe / Short CDX.NA.IG: The AI Supply Asymmetry Basis Trade
Vanguard’s explicit shift toward European and Canadian credit is the consensus institutional response to the US supply wall. Hedge funds express the same logic synthetically: sell protection on iTraxx Europe Main 5-Year (long European investment-grade) while buying protection on CDX.NA.IG 5-Year (short US investment-grade), sized in equal DV01. The trade profits as US spreads widen relative to European spreads.
The supply asymmetry is structural and documented. Morgan Stanley projects $400 billion of hyperscaler paper in US dollar markets in 2026; European investment-grade faces no equivalent supply wall. Janus Henderson’s December 2025 Fixed Income Outlook explicitly confirms AI debt supply is “skewed more to the US.”
The monetary policy divergence reinforces the basis simultaneously from two directions: the ECB cutting more aggressively suppresses European spreads, while the Fed on pause keeps US debt-service pressure elevated on highly levered hyperscalers. Both vectors move the CDX/iTraxx basis in the same direction.
Trade 4 — Receiver Swaps Paired With Long CDX Protection: The Macro Paired Hedge
Narayanan’s third stated strategy — increasing interest rate duration in anticipation of Fed accommodation — pairs with the credit short as a two-leg macro hedge: receive fixed in 10-year or 30-year USD interest rate swaps while holding long CDX IG protection. Both legs pay in the same scenario: a growth shock severe enough to force Fed accommodation, where credit spreads widen and long-end Treasuries rally simultaneously in a flight-to-quality move.
Bridgewater’s Co-CIO Greg Jensen, in a January 8, 2026, Institutional Investor interview, called 2026 “a dangerous year for interest rates” and warned that Fed cuts at the front end “could lead to a bubble and hurt the long end.” He described bond supply as “the worst in aggregate ever” and warned “continued use of fiscal policy could lead to more inflation.” Jensen’s Pure Alpha 18% fund returned 34% in 2025 — its best year since 2010 — making his framework one of the most validated macro frameworks entering this year.
The explicit risk for this leg: stagflation, where the Fed cuts the front end but the long end sells off on inflation expectations, generating duration losses that net against the credit short. Jensen names this his primary rate risk for 2026. Apollo frames it as its “1970s backdrop.” This leg requires explicit scenario analysis and defined loss limits — not just a directional spread-widening view.
Trade 5 — Hyperscaler CDS vs. Industrial CDS Dispersion: The Apollo Divergence Trade
The most actionable data signal in this cycle predates the Vanguard interview by 30 days. On January 21, 2026, Apollo Chief Economist Torsten Slok published a Daily Spark note using ICE BofA data in the 7-to-11-year maturity range showing that investment-grade credit spreads are actively widening for hyperscalers and simultaneously tightening for industrials. That divergence was visible in price data a full month before it appeared in institutional strategy disclosures.
The structure is a dispersion trade that requires no directional view on whether artificial intelligence succeeds or fails as a technology — only that the spread differential already in motion continues. Buy protection on hyperscaler single-name CDS (Oracle, Meta, Alphabet) while selling protection on industrial and utility investment-grade credits.
AllianceBernstein’s 2026 Credit Outlook forecasts exactly this continuation: “We see 2026 as a divergence year marked by softening corporate fundamentals and higher dispersion, which could be amplified by AI.” J.P. Morgan’s 2026 Market Outlook similarly calls for higher investment-grade dispersion, noting weaker fundamentals will be felt differentially across sectors.
The fundamental credit pressure on the hyperscaler side is acute. Pivotal Research projects Alphabet’s free cash flow to plunge from $73.3 billion in 2025 to approximately $8.2 billion in 2026 — a near-90% decline — against a $185 billion capex commitment. (CNBC, February 6, 2026) The four hyperscalers combined are projected to spend close to $700 billion on AI infrastructure in 2026, up more than 60% from 2025 levels, with Amazon projected to turn free cash flow negative by an estimated $17–28 billion. That is the precise credit deterioration that widens a hyperscaler’s CDS while leaving utility and industrial credits largely unaffected — and the reason Apollo’s January 21 divergence signal was not a coincidence.
📋 Want the full trade reference? The five trades above are summarised here. The complete Patreon reference document covers every trade in full institutional detail: exact entry windows with basis-point levels, exit mechanics, what each incorrect version of the trade looks like and why it fails, how to reconcile conflicting CDS data across Bloomberg / Bondblox / Voya, and seven pattern-recognition lessons extracted for future cycles. Written for re-reading under time pressure — not for one-time consumption.
IV. Why the “Rotate into Quality” Defence Fails This Cycle
Every trade above shares a structural prop that distinguishes this credit cycle from prior ones: the conventional quality rotation defence is broken. In standard recessions, portfolio managers rotate from BBB-rated into A-rated and AA-rated credits. The trade works because quality issuers historically face less new-issue supply pressure during stress. But Oracle (Baa2/BBB/BBB), Alphabet (AA+), Amazon (AA), and Meta (A) are the quality cohort — and they are simultaneously the entities flooding the market with $20–32 billion in new supply within days of each other. There is no safe harbour within investment-grade when the largest new-issue supply sources are also the highest-rated names.
Breckinridge documents the direct evidence: the spread differential between A-rated and BBB-rated investment-grade is already historically compressed (Z-score of negative 1.5 relative to the five-year average), meaning the quality spread cushion has been fully consumed. There is no premium to rotating up in quality because the quality spread has been arbitraged away by the same demand dynamics that created the crowding problem.
Janus Henderson explicitly identifies utilities — not AA-rated hyperscalers — as the correct quality rotation for this specific cycle. That is an endorsement of the dispersion trade in Trade 5, not a contradiction of the broader thesis. The quality answer in 2026 is sector rotation away from hyperscalers, not a ratings upgrade within them.
“It really shifts from a company that had no financial policy to one that is saying: ‘We clearly want to be investment grade and want to have a balanced approach between debt and equity.’”
— Brad Smith, Portfolio Manager, Janus Henderson Investors · IFR, February 2026
Smith’s observation about Oracle — the largest non-bank investment-grade issuer — captures the mechanism precisely. When the largest borrowers in the index are issuing $25–32 billion in jumbo deals specifically to avoid being downgraded to high yield, the market is no longer pricing credit risk at the margin. It is pricing survival. That repricing, when it transmits fully, is not gradual.
V. Risk Framework: Three Scenarios That Break These Trades
Risk 1 — Sustained Soft Landing and AI Revenue Monetisation. The primary failure mode across all five trades is a soft landing in which hyperscaler capital expenditure generates measurable revenue within the forecast horizon. If Oracle’s and Alphabet’s deals are followed by on-time data centre completions, growing cloud revenues, and improving debt coverage ratios, CDS protection bleeds carry and CDX payer swaptions expire worthless. Goldman Sachs’ John Sales, following Oracle’s record deal, said markets remain “wide open” for technology deals, arguing the key story is record demand, not supply saturation. The $490 billion in taxable bond fund inflows and $304 billion in foreign net purchases documented by Breckinridge do not disappear overnight. This is an asymmetric bet that supply overwhelms absorption faster than AI monetisation materialises — a probability, not a certainty.
Risk 2 — Independent European Shock. The iTraxx long leg assumes European credit outperforms US credit on a relative basis. An independent European shock — German manufacturing contraction deepening, peripheral fiscal stress re-emerging, a banking sector event — could widen iTraxx Main regardless of US conditions, invalidating the geographic basis trade without touching CDX. Apollo places a 20% recession probability in Europe versus 30% in the US. (Apollo 2026 Economic Outlook) That 20% tail is not negligible and requires explicit position sizing discipline on this leg.
Risk 3 — Stagflation and Duration Blowout. For the paired receiver swap and CDX protection trade in Trade 4, the worst-case scenario is stagflation — the Fed cuts the front end while long-end Treasuries sell off on inflation expectations. The credit short pays, but duration losses can net the two-leg position to zero or negative. Apollo names this the “1970s backdrop.” Bridgewater’s Jensen identifies it as his primary rate risk for 2026. (Institutional Investor, Jan 8, 2026) This leg requires explicit scenario analysis with defined loss limits before execution, not just a directional spread forecast.
Conclusion: The Market Signalled in January. Vanguard Confirmed It in February.
The timeline is unambiguous. On January 21, Apollo’s Torsten Slok documented the hyperscaler-vs-industrial spread divergence in real-time price data. Oracle’s CDS had already moved from 36 to 139 basis points by December 2025 — a 3.5x move to 16-year highs — before reversing on deal completion. On February 2, Oracle’s $25 billion deal produced a record $129 billion order book and still required a 40–58 basis point concession premium: the new-issue saturation mechanism demonstrated empirically. On February 9–10, Alphabet raised $32 billion across three currencies in under 24 hours, including the largest corporate bond sale in sterling market history. Secondary investment-grade trading averaged more than $61 billion per day in January, according to Crisil Coalition Greenwich data cited by Bloomberg on February 17 — up 11% year-over-year, the fingerprint of large-scale institutional repositioning in real time. On February 20, Vanguard’s co-head of investment-grade credit confirmed the institutional logic publicly.
Five trades capture the alpha embedded in this sequence. The Oracle CDS leg was already executed and largely closed before Narayanan spoke. The CDX investment-grade payer swaption, the geographic basis, the paired duration hedge, and the Apollo-documented hyperscaler-vs-industrial dispersion trade remain live. All five are anchored to verifiable, sourced primary data from Bloomberg, IFR, Apollo, Breckinridge, Bridgewater, Janus Henderson, AllianceBernstein, J.P. Morgan, and TD Securities.
The question is not whether the supply wall is real. Morgan Stanley’s $400 billion estimate is already corroborated by Oracle and Alphabet alone — $57 billion in three weeks, both deals requiring concessions despite record oversubscription. The question is whether the absorption capacity that has held spreads at the 2nd percentile of a 20-year lookback can hold through a year in which the largest and highest-rated borrowers in the index are issuing at a pace with no historical precedent.
The answer, per Vanguard’s own co-head of investment-grade credit: not necessarily.
📊 Want Deeper Quantitative Analysis?
This research involved weeks of data collection, source verification, and cross-referencing across Bloomberg, IFR, Apollo, Breckinridge, Bridgewater, and a dozen other primary sources. If you found value in this deep-dive, I publish exclusive quantitative research, trading strategies, and institutional-grade analysis on Patreon.
Start here — the full trade reference for this article: → Trade Reference: US IG Credit Crowding Thesis — Full Entry Levels, Exit Mechanics & Lessons
This document covers all five trades with exact entry windows, the incorrect versions of each trade and why they fail, reconciliation of conflicting CDS data across sources, and seven pattern-recognition lessons for future cycles.
By joining, you’ll be supporting independent financial research and motivating more content of this calibre.
→ Join the Patreon community for ongoing research
Primary Sources
Bloomberg — Vanguard Targets Canada, Europe to Hedge US High-Grade Debt Risks, Feb 20, 2026
Breckinridge Capital Advisors — Q1 2026 Corporate Bond Market Outlook
IFR — Oracle Raises US$25bn in Bond Offering as AI Funding Spree Continues, Feb 7, 2026
Bloomberg — Oracle Blockbuster Bond Sale to Usher in AI Debt Wave, Goldman Says, Feb 5, 2026
Euronews — Alphabet Plans to Sell Rare 100-Year Bond in Huge Multi-Currency Debt Raise, Feb 10, 2026
GlobalCapital — Alphabet Leaves No Record Unbroken With Sterling Debut, Feb 10, 2026
CNBC — Alphabet Set to Raise Over $30 Billion in Global Debt Sale, Feb 10, 2026
CNBC — Alphabet 100-Year Bond: Debt Fears and AI Credit Risk, Feb 12, 2026
CNBC — Tech AI Spending Approaches $700 Billion; Free Cash Flow at Risk, Feb 6, 2026
Bloomberg — Oracle Credit Fear Gauge Hits Highest Since 2009 on AI Bubble Fears, Dec 2, 2025
Institutional Investor — Behind Bridgewater’s Surge, Jan 8, 2026
Janus Henderson — Fixed Income Outlook: Building Resilience in 2026
AllianceBernstein — 2026 Credit Outlook: Growing Divergence Amid AI’s Big Build-Out
Bloomberg — Heavy Demand for Corporate Bonds Creates Record Trading Volume, Feb 17, 2026 (Crisil Coalition Greenwich data)
TipRanks — Why Oracle’s Record Bond Sale Clears the Way for More AI Debt, Feb 2026
Voya Investment Management — The Connection Between Oracle’s Credit Default Swaps and AI, Dec 2025
Bondblox — Oracle’s 5Y CDS Jumps to Its Highest Level Since 2009, Dec 2025
FRED — ICE BofA US Corporate Index Option-Adjusted Spread (live data)
This article is analytical commentary for informational purposes only. Nothing herein constitutes investment advice, a solicitation, or an offer to buy or sell any security. All data and quotations are sourced from the linked primary sources and are accurate as of their publication dates.
About the Author
Navnoor Bawa publishes institutional-grade quantitative research and trading strategy analysis at the intersection of fixed income, derivatives, and AI-driven market structure.
· LinkedIn — Connect for research updates and commentary · YouTube: The Mathematical Trader — Subscribe for video analysis and deep-dives · Full Trade Reference — This Article (Patreon) — Entry levels, exit mechanics, incorrect trades, and lessons · Patreon — Ongoing Research — Exclusive quantitative research and trading strategies
Cover photograph: Federalreserve, public domain, via Wikimedia Commons.