Cathie Wood Sees a Hidden Moat Around Anthropic. It’s Not Just Its Models

LongbridgeAII'm LongbridgeAI, I can summarize articles.

Anthropic, PBC‘s (NASDAQ:ANT) biggest advantage may be something investors cannot see: the AI models it has developed but deliberately kept under wraps. Cathie Wood says the company is investing so aggressively in training that its internal technology could be far ahead of what competitors can access, creating a lead that goes beyond the popularity of its Claude chatbot.

The Advantage Behind Anthropic’s 85% Margin

Wood pointed to Anthropic’s financial disclosures as evidence of an unusual business model. The company reported an implied gross margin of roughly 85% in the first half of 2026, even as Wood said about 75% of its spending was going toward training AI models.

"The biggest surprise, of course, was their gross margins, 85%, even though I think they are spending 75% of their budget on training," Wood said in a Nov. 18 interview with WOLF Financial.

The distinction matters. Training involves building and improving AI models, while inference is the process of running those models to answer customer requests and generate revenue. Anthropic’s S-1 filing shows that its first-half 2026 revenue was $16.3 billion, against $2.38 billion in cost of revenue, implying a gross margin of about 85%.

But that figure does not capture the full cost of developing its technology. Anthropic also reported $9.74 billion in research and development expenses during the same period, much of which includes compute used for model development and research. The headline margin therefore should not be mistaken for the profitability of the entire business.

Read Also: Anthropic IPO: Can SpaceX's $19B Playbook Work?

Why Wood Sees a Potential Moat

Wood’s more consequential argument is that Anthropic may be building capabilities that rivals cannot readily replicate. She said training costs are falling about 75% annually, allowing companies to improve their models faster and more efficiently.

Anthropic, she says, is using that advantage to keep its most advanced models private rather than releasing them for others to study. She also pointed to the company restricting "distillation" — a process through which developers use the outputs of a more advanced model to train a less capable one.

"They could be a decade ahead of the open-weight models," Wood said. Open-weight models make their trained numerical parameters, or weights, available for others to download and adapt. She argues that competitors relying on these publicly accessible systems could struggle to match Anthropic’s privately developed technology.

Wood later clarified that "a decade ahead" was an analogy, not a literal prediction. AI advances that once took 18 months to two years are now happening in roughly two months, she said.

Investment Takeaway

For investors, the question is whether Anthropic can turn the training advantage Wood describes into a durable lead in enterprise AI, where businesses pay to use its models. Its 85% gross margin excludes research and development expenses and does not establish overall profitability.

Watch future disclosures for evidence that model improvements are translating into stronger customer adoption and revenue without costs rising just as quickly.

Read Also: Dan Ives Says Anthropic's Blockbuster IPO Is Fueling Wall Street's Tech Selloff

Courtesy Ark Invest

ANT Price Action: Anthropic, PBC shares were down 10.59% at $86.88 at last check Friday, according to Benzinga Pro data.

Photo: Courtesy Ark Invest

Login to unlock2,398characters for free

Due to copyright restrictions, please log in to your Longbridge account to view this content.
Thank you for your understanding and support of licensed content.