Last week’s investor panic over AI names leaves model giant Anthropic with the unenviable choice of having to list well under its reported valuation target or crashing the market further and perhaps delaying its IPO for good.

Democratic victories in Congress and across state governments, notably in Texas, are calling into question the viability of the US’s data centre buildout.

After a short rally the day following the election, stock markets turned aggressively on Thursday 5 November.

The Nasdaq 100 fell over 6% in trading in a single day, with SpaceX one of the largest losers, crashing well below its USD 135 a share IPO price. Nvidia also took a heavy beating – it will likely do so again, today (albeit for different reasons…)

Other AI-linked hyperscaler stocks including Microsoft and Google parent Alphabet were also the targets of heavy selling.

The share price of neoclouds like CoreWeave and Nebius have also taken a battering since the election on fears Democrats could slow down data centre construction across the US.

The Nasdaq 100 recovered a little on Friday 6 November but closed the day 13% below its most recent high.

“This is a continuation of the initial tremors,” noted an equity placement agent working in the sector. “But there could be much more to go down from here.”

Reports of huge pain across funds in October, even before the November sell-off, shows a market that is ill-prepared to take on a USD 100bn listing, a tall task at the best of times. But when investors are being hammered on all sides, it makes an IPO run exceedingly perilous, for an issuer and the investors that take part.

Misanthropic investors

Anthropic, on the verge of launching a historic IPO, is particularly exposed to this risk due to its need for increasing computing capacity.

Fewer data centres will mean hyperscalers and cloud providers with existing compute infrastructure will be able to charge a premium to frontier LLM developers like Anthropic.

In its IPO prospectus, Anthropic revealed USD 895.4bn in non-cancellable AI infrastructure commitments to many of these cloud providers.

These agreements include strict “take-or-pay” true-up mechanisms, which mean that if Anthropic fails to consume its required dollar amount of compute in each billing cycle, it could be subject to cash penalties.

If compute prices rise because of reduced data centre construction, Anthropic’s near USD 1tn investment would buy it far less capacity, meaning the company might have to top-up in the expensive third-party spot market.

“If Democrats halt data centre buildout, Anthropic is staring now at potentially exponentially higher compute costs, which makes its unit economics even worse. That is going to hit its growth margins significantly,” noted an ECM investor looking at the IPO.

Several market participants speaking to ECM Pulse acknowledged that this market is perhaps the worst Anthropic could have wished for to launch its historic IPO.

Number crunching conducted by ECM Pulse before Thursday’s stock market rout, fittingly enough with the verified assistance of a proprietary AI agent, pointed to an IPO valuation of between USD 1.2tn-USD 1.45tn as a more attractive clearing level for Anthropic than the USD 2tn regularly touted at mid-year.

The top at USD 1.45tn, slightly under the already-trimmed USD 1.5tn post-summer target, baked in a slight discount given recent pressure on the sector and peer OpenAI’s discounted private capital raise.

This analysis was conducted using a basket of large cap hyperscaler peers, alongside SpaceX and Palantir.

Market moves last week, and an assumption of higher compute costs increasing Anthropic’s weighted cost of capital, smashes all those assumptions.

The model now points to an IPO valuation at around USD 1tn for Anthropic. Even then the risks are huge, with stress testing pushing it below the USD 1tn mark.

Several market participants speaking to ECM Pulse, and press reports at the tail end of last week, all validate a USD 1tn valuation.

“Clearly, it’s a very challenging environment for Anthropic to go to market. They can probably get a deal done but it would be at a far lower valuation, well below USD 1.5tn I would say,” said the investor. “Buying Anthropic at the end of the year would be a very difficult call for a lot of investors, so a deal is difficult.

“But what does it do if it doesn’t go? It could launch another private round to get some of the funds it needs, but that still isn’t great. Also, if you pull the deal, the market will likely get another shock wave.”

An advisor familiar with the deal shared the sentiment that Anthropic would be “crazy” to launch an IPO in this market, albeit for the fact that they need the funds and access to public markets.

Hyperscalers like Meta and Alphabet, which are building direct LLM competitors to Anthropic’s Claude, are already public and have been tapping equity capital markets for tens of billions of dollars of funding this year.

Not listing now could be terminal for Anthropic’s growth plans – but coming to market at half the USD 2tn level that Anthropic was targeting before the summer screams a forced sale and could herald weak aftermarket performance.

The roadshow is scheduled to kick off this week, one ECM advisor told us. “We’ll see if it happens. If it does, listening to investor feedback will be critical.

“It’s important to price the deal correctly, or it could be a disaster. From what I’m hearing, USD 1.5tn is too high. I could see Anthropic lowering both the valuation and the number of shares they sell, like OpenAI did in its last funding round.”

For Anthropic there are almost no good options on the table and its decision in the next days, or weeks, could not have higher stakes for the company or the next wave of AI investment.