U.S. Treasury Yields Pull Back Near 5%: AI Financing Costs and Oil Prices Squeeze Pricing
nashnova research
The 10-year Treasury yield has pulled back from its early-October high to near 5%, yet AI companies' funding pressure has not eased — widening credit spreads are offsetting the decline in benchmark rates, while energy inflation and growth fears squeeze rate expectations from both sides.
Benchmark rates fell — why is AI funding still expensive?
The 10-year yield declined not because inflation cooled, but because markets began pricing in a slowdown in AI infrastructure buildout under high financing costs. This means → the very reason rates fell is itself bad news.
Lower Treasury yields ≠ cheaper corporate borrowing. In plain terms = Treasuries set the "base price"; companies pay an extra layer called the credit spread. The base dropped, but the spread widened — so actual borrowing costs may be higher.
Brent crude closed at $116.72 per barrel on October 23, the highest since April 29. September U.S. CPI rose 3.7% year-on-year; gasoline climbed 9.1% month-on-month, accounting for over half the monthly gain. This reflects energy inflation pushing rate expectations higher from the opposite direction.
CoreWeave vs. Microsoft — how far apart are their borrowing costs?
CoreWeave's 2031 bonds yield 15.76%; its five-year CDS — a form of insurance against default — trades at roughly 1,115 basis points. Media reports cite a five-year cumulative implied default probability of about 63%, up from around 30% a year ago. This means → the cost of insuring against a CoreWeave default has more than doubled.
A caveat: implied default probability bakes in recovery-rate assumptions and risk premia. It measures how expensive default protection is, not a direct prediction of actual default.
Microsoft's 2060 bonds yield about 6.36%, Amazon's 2076 bonds about 6.95%, and Meta's 2064 bonds about 7.48%. In plain terms = big tech's funding got pricier, but they are nowhere near a liquidity crisis.
The core difference is the business model. New cloud players like CoreWeave spend first, collect cash later, relying on continuous outside funding. The giants have mature cash flows; their pressure shows up as costlier new debt + squeezed free cash flow + higher return hurdles.
The Fed chair won't give a roadmap — who should markets listen to?
Fed Chair Kevin Warsh has refused to offer any forward guidance on the rate path since taking office in May, insisting that markets should judge rates from the economic data alone.
Other Fed officials' remarks, however, still move pricing. On October 14, a hot inflation print briefly pushed October hike odds from about 20% to 50%, before a string of officials signaling a pause erased most of the rise.
Neil Dutta, head of economics at Renaissance Macro Research, noted: "There is a disconnect between what Warsh says and what other officials signal. When that happens, markets risk sharp reversals at any given meeting."
What exactly did each official say, and how did hike odds shift?
New York Fed President John Williams said on September 29 that one more hike "later this year" might be appropriate → October hike odds fell from about 70% to 50%.
Fed Vice Chair Philip Jefferson said on October 1 that more time was needed to observe → odds dropped to roughly 25%. Fed Governor Christopher Waller said on October 8 that officials need not tighten at consecutive meetings → the pause expectation solidified further.
As of October 23, interest-rate swap pricing puts October hike odds below 15% and December hike odds at about 80%. This means → markets bet the Fed will stand pat in October but almost certainly hike in December.
Earnings week meets the FOMC — what is the market watching?
Alphabet and Microsoft report on October 28; Amazon reports after the close on the 29th. The key question: can AI capital spending deliver strong enough returns, and will the giants keep expanding AI capex beyond 2028–2030?
JPMorgan strategists forecast hyperscaler capex at roughly $950 billion in 2026, $1.4 trillion in 2027, and $3 trillion by 2030. This means → the market believes AI infrastructure spending will keep doubling — but so will the financing burden.
The FOMC announces its decision at 2:00 a.m. Beijing time on October 29; Warsh holds a press conference at 2:30 a.m. Jefferies forecasts the S&P 500 reaching 8,000 by end-2026, based on EPS of $373 (up 35% year-on-year) and a 21.5× P/E multiple.
Deutsche Bank's chief U.S. economist Matthew Luzzetti expects further hikes in the months ahead. Put simply = whether big tech earnings beat expectations is the most direct test of the current AI valuation thesis.
市场有风险，内容仅供研究参考，不构成投资建议。