Wall Street collectively rushes to "extinguish panic" over CoreWeave: Don't use one company to negate the entire AI cycle!

CoreWeave’s unsecured bond yields have surged to around 15%, CDS spreads have soared past 1,000 basis points, and its stock price has fallen more than 24% in a single week, sparking systemic concerns across the market regarding the entire AI capital expenditure cycle.

However, a small bond is being treated by the market as a barometer for the entire AI capex cycle — top Wall Street institutions believe this is a costly analogy error.

According to TrackTrend Trading Desk, Barclays pointed out that CoreWeave’s approximately $1.5 billion in unsecured bonds is only a tiny portion of its overall $35 billion debt. Using the price signal of this instrument to undercut the whole AI cycle is akin to how, in 2023, the $4 billion KRE regional bank ETF was used to anticipate a US recession. Goldman Sachs believes the market has chosen the wrong reference — the unsecured bond market does not reflect CoreWeave’s true incremental financing cost. JPMorgan says current risks are manageable.

A Familiar Scene: Silicon Valley Bank in 2023, CoreWeave in 2026

Ajay Rajadhyaksha, Global Head of Macro Research at Barclays, directly compared the current market panic to the Silicon Valley Bank crisis in a report published on October 19.

In March 2023, Silicon Valley Bank was taken over by the FDIC, followed by the collapses of Signature Bank and First Republic Bank. A narrative quickly formed: regional banks are the lifeline of SME credit. Even if more banks didn’t fail, the survivors would tighten lending, a credit crunch would ensue, and recession would follow.

The bond market bought into this story. Six weeks after SVB’s collapse, the market was still pricing in 75 to 100 basis points in rate cuts by the end of 2023.

What happened? The Federal Reserve did not cut rates even once in 2023; instead, tightening continued. US economic growth for the year significantly exceeded the Fed’s own March projection of 0.4%. Recession never came. The S&P Regional Banking ETF KRE retraced all of its losses.

Rajadhyaksha wrote: “The mistake wasn’t recognizing that Silicon Valley Bank had failed. It did, in fact, collapse. The mistake was extrapolating from a small, volatile, structurally fragile instrument to the entire US economy.”

What was the “tail” back then? A regional bank ETF — KRE — with about $4 billion in AUM, equally weighted, and many of its components thinly traded. This tiny ETF prompted the repricing of trillions of dollars in rates, investment grade credit, and equity positions.

The logic is circular. KRE falls, sentiment worsens, people sell KRE, KRE keeps falling, the bond market prices in more rate cuts, and the equity market follows lower. The tail wagged the dog.

Rajadhyaksha believes the same mistake is playing out live in October 2026.

CoreWeave’s ‘Tail’: $1.5 Billion in Unsecured Bonds

Rajadhyaksha believes the “tail” wagging the dog this time is CoreWeave’s approximately $1.5 billion in structurally subordinated unsecured bonds.

CoreWeave’s debt structure consists of three layers. The key layer is secured delayed draw term loans (DDTL), held in individual special-purpose vehicles (SPVs), each tied to specific GPU hardware, customer contracts, and data center leases. When Meta signs a multi-year ‘take-or-pay’ contract and assigns it to an SPV, the bank loan’s collateral is essentially Meta’s payment commitment.

The second layer consists of about $9 billion in convertible bonds, with coupons between 1.75% and 2.875%.

Last comes the unsecured notes — in a liquidation scenario, billions of dollars in senior claims are ahead of these bonds.

Rajadhyaksha notes: “A 15% yield tells you something about this paper. But it does not tell you the company is about to cease operations.”

He offers an analogy: “If a person has a $35 million home mortgage plus a small credit card balance, and the credit card rate is high, the issue is with the card—not with all mortgage borrowers in the city.”

The comparison in numbers is intuitive: In March 2023, the bond market priced in 100 bps of rate cuts because of a $4 billion ETF’s decline; in October 2026, because about $1.5 billion in subordinated structured bonds traded at 80 cents on the dollar, the market started doubting a trillion-dollar capex cycle.

JPMorgan: Bank Exposure Is Secured, Risks Manageable

JPMorgan’s European bank research team (led by Kian Abouhossein) broke down CoreWeave’s financing structures from a banking industry perspective in a report released on October 20.

By the first half of 2026, CoreWeave’s total liabilities were about $52 billion, including about $35 billion in total debt, around $16.3 billion in operating lease liabilities, about $5 billion in shareholder equity, and a leverage ratio of roughly 10x.

The team noted that banks’ main exposure is concentrated in the secured SPV structure — about $15 billion drawn and $9 billion undrawn — and a $2.5 billion secured revolving credit facility at the parent level. Direct exposure to unsecured bonds (about $21 billion including around $11 billion in convertibles) is limited.

On the bank side, JPMorgan’s research found that large US investment banks have the highest involvement, with European banks (Barclays, Deutsche Bank, Société Générale, Crédit Agricole) and Asian banks (MUFG, Mizuho) also participating. Based on deal participation and standard 10-15% syndicate holding ratios, JPMorgan estimates most banks have under $500 million in exposure to CoreWeave, while large US investment banks and MUFG likely have between $1 and $3 billion, mostly as DDTL.

The report also highlights three key risks:

First, SPV counterparty risk — not all SPVs are backed by ultra-large cloud providers, and CoreWeave is expected to be free cash flow negative through 2030 (JPMorgan estimates negative free cash flow of $29 billion in 2026 and $37 billion in 2028), with a clear timing mismatch between initial capex and contract revenues.

Second, project execution risk — social resistance, construction delays, and power supply challenges may affect revenue realization. Oracle’s force majeure clause with Blue Owl has already set a precedent.

Third, contagion risk — If the AI-related high-yield bond market tightens, it could spill over into broader investment banking revenue, including ECM. JPMorgan data shows technology ECM in 2026 will account for 30% of global ECM volume, the highest since 2016.

Goldman Sachs: Market Has Chosen the Wrong Reference Point

Goldman Sachs analysts Gabriela Borges and Maura Hager upgraded their rating on CoreWeave from “Neutral” to “Buy” on Monday, with a price target of $106. At that time, the stock closed at $61.39, implying about 73% upside.

They argue the market is using unsecured bond yields to judge CoreWeave’s financing costs, which is the wrong reference point.

The analysts emphasized that CoreWeave’s main funding tool is DDTL — asset-backed project financing collateralized by specific customer contracts, GPU infrastructure, and related cash flows. Lenders underwrite the cash flow quality of the individual contracts, not CoreWeave’s overall creditworthiness.

Based on this, they estimate CoreWeave’s current weighted average debt cost at 800–900 basis points (annualized Q2 interest expense of $2.56 billion divided by about $30 billion in total debt). For incremental financing, Goldman Sachs assumes investment grade customers’ DDTL pricing at SOFR + 300bp, high-yield customers at SOFR + 600bp, blended between 750–1050bp, with a conservative baseline of 1000bp — implying incremental financing costs would rise by just 100–200bp from current levels.

They also cite market data: DDTL 5.0 and 5.5 are trading no lower than original issue discounts, showing that the asset-backed financing market’s confidence in CoreWeave remains intact.

On pricing power, Goldman estimates CoreWeave’s current blended monetization roughly at $10 million annual recurring revenue (ARR) per MW, but sees upside: short-term (3–6 month) contracts have reached $40 million per MW; CoreWeave raised prices by 25% in July and another 10% in October. Using $13 million per MW as the “reasonable base case” for new long-term contracts in their model yields an unlevered IRR of about 15%, exceeding the roughly 11.8% weighted average cost of capital.

However, Goldman’s sensitivity analysis reveals the fragility of this logic: if the current actual blended level of $10 million/MW is used, unlevered IRR falls to 6.3%, and the spread at the same financing cost turns negative 6 percentage points. The bull case depends on CoreWeave increasing its per-MW revenue by about 30% from current levels.

Goldman’s own models also show free cash flow remains negative $28.2 billion in 2026 and 2027, and negative $15.8 billion in 2028, turning positive only in 2030.

It’s noteworthy that Goldman states in the report appendix that it is serving as lead or joint bookrunner for a pending CoreWeave underwriting mandate, holds over 1% of CoreWeave shares, and has received investment banking compensation from CoreWeave in the past 12 months.

Demand Side: Supply Chain Data Speaks

Barclays’ Rajadhyaksha turned his focus to the deepest ends of the supply chain.

TSMC (October 15) reported revenue surpassing its own guidance by $1 billion, setting a sixth consecutive quarterly profit record. High-performance computing (the AI proxy) made up 68% of revenue, up from 66% in the prior quarter. The 2nm node already accounts for 8% of wafer sales, and the company raised its Q4 growth outlook.

ASML (October 14) also logged record sales, with operating profit more than doubling. CEO Christophe Fouquet said planned shipments for 2027 and 2028 will not fully meet customer needs, and some clients have begun attempting to reserve equipment for 2029 and 2030.

Rajadhyaksha wrote: “Makers of the machines used to build chips have already sold out their next two years of capacity.”

The five largest hyperscale cloud providers (Microsoft, Meta, Alphabet, Amazon, Oracle) are projected to spend a combined $725–800 billion on AI infrastructure in 2026, with 2027 estimates exceeding $1 trillion. CoreWeave’s capex guidance for 2026 is $35–39 billion, about 4–5% of the total above.

Rajadhyaksha’s conclusion: “Even if CoreWeave completely stopped spending, we would only lose 4–5% of total AI capex.”

He also points out that hyperscale cloud providers’ spending primarily relies on free cash flow, not the unsecured bond market. “Extrapolating from an emerging cloud provider to one of the world’s most profitable companies stretches the logical chain far too much.”

No Federal Reserve Backstop, but Demand Reality Exists

Barclays also acknowledges one key difference this time versus 2023: there’s no policy tool acting as a safety net.

In 2023, the Federal Reserve introduced the Bank Term Funding Program (BTFP), allowing banks to swap underwater bonds for liquidity at par, which was key to breaking the panic. There’s no comparable central bank tool for AI infrastructure financing.

But Rajadhyaksha argues that the AI capex cycle has its own “backstop logic”: Alphabet, Microsoft, Meta, and Amazon crave more compute power than current supply can provide. As long as this holds, funding follows demand.

TSMC’s 68% revenue from high-performance computing, ASML selling out through 2028, and CoreWeave’s $104 billion contract backlog are not policy tools, but they are demand realities. When the smoke clears, it’s the reality of demand that truly matters.

Rajadhyaksha concludes: “Markets always have a tendency to find the smallest, most volatile instrument in a complex system and treat it as a barometer for the whole. In 2023 it was KRE, in 2026 it is CoreWeave’s unsecured bonds. Both signals are real. The 2023 extrapolation was wrong, and we believe the same will be true in 2026 (if it plays out in the coming weeks).”