Listen to the latest episode of Credit Exchange with Lisa Lee
Published in London & New York
10 Queen Street Place, London
1345 Avenue of the Americas, New York
Creditflux is an
company
© Creditflux Ltd 2026. All rights reserved. Available by subscription only.
News Analysis
Anthropic swims in uncertain AI waters
by Lisa Fu
As markets count down to its blockbuster IPO, Anthropic has given the public a peek inside its books, underlining the huge scale of demand for AI, but leaving lingering questions about the infrastructure needed to deliver on it.
For debt markets that have backed the AI buildout, the pre-IPO filing from the Claude developer has provided some reassuring evidence on the demand side. However, what also stands out is Anthropic’s heavy capital commitments to secure the required computing capacity to support the AI boom.
“The first point: they have a lot of cash pre-IPO,” said Steve Purdy, head of credit research at TCW. “Second point is, they do have a lot of compute that they want to purchase in the future, more than we would have liked. But it feels net-net positive.”
Breakneck revenue growth and other striking metrics in the pre-IPO paperwork point to an emerging AI customer base, said Malik Ahmed Khan, equity analyst for technology at Morningstar.
Investment has flowed into infrastructure such as data centres, cloud services and advanced chips on the assumption of sufficient AI demand going forward. AI companies need enough money coming in from customers to fulfil their various leasing agreements and contracts for critical hardware and services. The counterparties are also relying on Anthropic to pay them, so they can keep their own creditors happy.
Balance sheet safeguard
The sum of USD 130bn in cash and marketable securities on Anthropic’s balance sheet offers a safeguard, said Khan. The cash can be used to cover the company’s many existing costs if sales do not take off as planned, he added.
However, Anthropic’s S-1 statement leaves questions about whether demand over the long term can soak up the massive supply of AI infrastructure being built out today. “You didn’t get any amazing news, nor devastating news, from the S-1. We already knew [AI demand] was growing like crazy,” Purdy said. “Does that tell us what’s going to happen in five years? No.”
You didn’t get any amazing news, nor devastating news, from the S-1
Steve Purdy
Head of credit research
TCW
One credit manager cited concerns that the drive for AI infrastructure is turning into a bubble, while others worry about a glut if usage patterns deviate from expectations.
Anthropic has already struck long-dated commitments, many of which are set to come online in the next five or so years, with tech partners such as Amazon, Google, Microsoft and Broadcom.
Even AI bulls see a risk that mass adoption of AI technology will take longer than anticipated, casting doubt on whether the sales ramp-up will be fast enough to match infrastructure obligations, Khan said.
If the market was fully confident that Anthropic would be able to make large long-term payouts to chip makers, Broadcom stock would have surged after the pre-IPO filing, he said, but this was not the case.
Overestimating AI demand two to three years in the future by 10% to 20% may not cause a big problem. But the market fears an overshoot of 50% to 60% could cause a seriously large glut.
Let demand flow through
With the debt markets awash with supply, companies would get better funding costs if they slowed down the pace to allow demand to flow through, said Purdy. But instead, they appear to be prioritising outbuilding their competitors in the AI race, he says.
The results of the scramble will take time to emerge, notes Khan. “If we’re building too much capacity right now, we’re going to know whether that was too much, likely towards the end of 2027.”
