Blacklisted

Circular financing among AI companies paying each other for the privilege of using one another’s semiconductors and to build data centers has fueled the AI trade, which in turn has buoyed the entire stock market.

So what happens when that circle is broken?

Last Friday, the Pentagon gave AI startup Anthropic until 5pm to loosen its AI guardrails, reportedly so that the Department of Defense could use Anthropic’s large language model Claude in autonomous weapons and to surveil American citizens. Anthropic refused to back down, so Defense Secretary Pete Hegseth designated the company “a supply chain risk to national security,” while President Trump ordered US government agencies to “immediately cease” using Anthropic products.

Meanwhile, archrival OpenAI swooped in and inked a deal with the US government, leaving Anthropic blacklisted and kicked to the curb.

A tangled AI web

Last week’s drama is a major blow to Anthropic. Claude was the only LLM that had been approved by the Defense Department for use in classified settings (until the Pentagon’s deal with xAI last month), which was an enormous competitive advantage, and it had a $200 million contract with the US military. That contract will likely be torn up and Claude’s security clearance demoted following Friday’s decision.

Here’s the real problem for investors like us: Anthropic may not be publicly traded, but its backers certainly are. Microsoft and Nvidia struck a deal late last year to invest up to a combined $15 billion in Anthropic. In exchange, Anthropic will buy up to $30 billion of cloud compute capacity from Microsoft, and utilize up to 1 gigawatt of computing power from Nvidia.

Meanwhile, Amazon invested $8 billion in Anthropic so that the startup will use Amazon Web Services as its primary cloud partner. But then Alphabet began throwing its weight around, supplying Anthropic with a million of its specialized TPU chips in a deal worth tens of billions of dollars, on top of the $3 billion the Google parent company has already invested in Anthropic.

Anthropic is clearly well liked by hyperscalers, but if it loses out on deals with government entities, it could be a blow to its allies in Silicon Valley. Not only will their stakes in Anthropic be worth less if the company’s business slows, Anthropic might not be able to pay what it owes for all the chips, cloud computing capacity, and power it has purchased from the tech behemoths.

That’s exactly what investors are already worried about. We’ve seen the market shift from pro-tech to HALO investments this year as investors fret that hyperscalers are spending too much money on tech that may ultimately not live up to expectations. Anthropic is a key part of the AI ecosystem, and if it starts to flounder, it could drag the rest of the AI trade—and the market—down with it.

Don’t panic just yet

Anthropic has vowed to contest the government’s decision in court, so the market could wait to see how things play out before any selling starts. And for now, the blacklisting has played in Anthropic’s favor: Chatbot users have revolted against OpenAI, which they suspect acquiesced to the government’s demands, despite CEO Sam Altman’s promises otherwise. OpenAI’s ChatGPT fell out of the top spot on Apple’s free app list over the weekend and was replaced by Claude, seemingly indicating that Anthropic is now a fan favorite.

How long that goodwill lasts, and whether or not Anthropic can capitalize on it, remains to be seen.—MR