AI capex bond market worries grow as tech borrowing costs rise
Alphabet’s higher spending plan has sharpened bond investor scrutiny of AI data-center debt across Google, Amazon, Meta and Oracle.
By Amanda Ross · Deals Correspondent
· 3 min read
The ai capex bond market debate intensified this week after Alphabet raised its capital-spending forecast, prompting higher yields and wider scrutiny of the debt used to fund artificial intelligence infrastructure. CNBC reported that fixed-income investors are demanding more compensation to lend to major technology companies as data-center construction, power costs and financing needs rise.
Credit spreads for Google, Amazon and Meta have widened, according to CNBC. A credit spread is the extra yield investors require above a comparable government bond to hold a company’s debt, and a wider spread signals that lenders want more reward for perceived risk.
Alphabet’s higher capex outlook raised concern that other large cloud and internet companies, often called hyperscalers, could also lift spending plans. Those companies are building large data centers across the United States to support AI workloads, a process that requires heavy investment in chips, buildings, networking equipment and electricity supply.
Why is AI capex worrying bond investors?
AI capex is worrying bond investors because the upfront cost of data centers is rising before the long-term returns are clear. If companies borrow more while free cash flow is pressured by construction and energy expenses, lenders may seek higher yields or stronger protections.
Power is one driver of the spending increase. GE Vernova Chief Executive Scott Strazik told CNBC that he expects the current inflationary environment to persist, partly because of heightened geopolitical conditions. CNBC also reported that oil moved above $100 a barrel this week.
Mizuho analysts wrote in a Friday client note that higher capex targets are testing investor tolerance, CNBC reported. The analysts said companies previously viewed as exceptionally strong credits are facing a sharp increase in AI-related costs, and added that hyperscalers are on pace to spend more on capital expenditures than they generate in free cash flow by next year.
A credit-fund portfolio manager told CNBC, on condition of anonymity to discuss private conversations, that the issue has become a subject of intense debate between bond and equity investors exposed to the largest technology companies.
Oracle becomes a hedge for AI debt concerns
Oracle’s five-year credit default swap has returned to a multi-year high, according to CNBC. A credit default swap, or CDS, is a contract that can pay out if a borrower defaults, and its price is often used as a market gauge of credit risk.
Barclays credit analyst Andrew Keches wrote in a Wednesday note that Oracle’s CDS is again being treated as a proxy for concerns tied to AI debt. Keches said the instrument has drawn interest beyond Oracle’s own fundamentals because it is liquid and linked in investors’ minds to AI capital spending, OpenAI execution risk and broader data-center investment themes.
Oracle is facing investor questions about how much it will use debt markets as data-center building and leasing become more capital intensive, CNBC reported. Earlier this month, S&P Global cut Oracle’s credit rating to BBB-, one step above junk status.
Oracle executives remain confident in the company’s AI position, CNBC reported. The company has expanding relationships with OpenAI and with hyperscalers including Meta and Nvidia on cloud architecture.
The higher yield environment may also affect future technology financing. CNBC reported that Meta is seeking financing for a $12 billion Texas data center, with pricing expected early next week, citing a person familiar with the talks who requested anonymity. The Financial Times reported that the deal is expected to price at a higher borrowing rate than prior projects, indicating that investors are seeking greater compensation while questioning returns on the buildout.
This story draws on original reporting from CNBC.