Anthropic posted a net loss of more than $80 billion in the first half of this year, according to its securities filing, ahead of a planned initial public offering next month. The loss reflects the massive infrastructure spending that has accompanied the artificial intelligence buildout, though concerns persist over whether AI can generate returns commensurate with the investment.
Up to $750 Billion for Compute Capacity
Anthropic's revenue reached $16.3 billion in the first half, up more than 1,300% from $1.1 billion a year earlier, according to the company's S-1 registration statement obtained by The Wall Street Journal on the 20th. Over the same period, however, the company recorded a net loss of $80.7 billion — nearly double the $42 billion annual loss it reported last year.
Infrastructure investment drove the widening loss. Anthropic spent $9 billion on computing capacity and infrastructure in the first half alone, already exceeding the $7.3 billion it spent for all of last year. The Journal reported that Anthropic's contracted commitments for future compute resources total $450 billion, describing them as "mostly contracts that cannot be terminated early." Including additional negotiations now under way, the company would need to commit as much as $750 billion to securing compute capacity alone.
Most of the first-half net loss, however, was not an actual cash outlay. Of the $80.7 billion, $72 billion stemmed from changes in the valuation of convertible financial instruments, including convertible bonds issued to raise funds for infrastructure investment. Because the value of such instruments is linked to the company's share price, the charges are structured to grow alongside Anthropic's valuation once it lists next month.
Big Tech 'Default Insurance' Trading Jumps 4.5-Fold
Debate over AI profitability continues. Credit risk concerns that began with CoreWeave, the leading neocloud provider, are gradually spreading to Big Tech. The Financial Times reported on the same day, citing a recent Morgan Stanley report, that trading in credit default swaps tied to bonds issued by Oracle, SpaceX, Alphabet, Amazon, Meta, Broadcom and Nvidia averaged $9 billion a week last month. That is roughly 4.5 times the level of less than $2 billion in July, two months earlier.
A credit default swap functions as insurance purchased in advance against the possibility that the referenced company defaults on its debt. CDS trading volumes have risen along with the expansion of AI infrastructure investment. The increase in default insurance trading indicates that the pace of high-risk fundraising has accelerated further. The FT noted that "a borrowing spree to fund trillions of dollars of AI infrastructure spending is fueling demand for insurance-like derivatives."
According to Morgan Stanley, AI-related bond issuance to finance infrastructure such as data centers and semiconductors has reached $466 billion so far this year, more than four times the $101 billion issued over the same period last year.
Morgan Stanley also projected that negative free cash flow at AI-focused Big Tech companies would hit a record of about $540 billion next year, including leases. Free cash flow is operating cash flow minus capital expenditures.