Ed Elson warns investors are misreading Anthropic's IPO filing

Misinterpreting the S-1 filing

Anthropic is preparing for a major initial public offering (IPO) this November. As the company prepares to go public, investors are closely scrutinizing its S-1 filing, which is the official document submitted to regulators to describe the business.

Ed Elson, a retail trader and co-host of the Prof G Markets podcast, suggests that many people are looking at the wrong metrics. He believes the market is focusing too much on the fact that the company's operating losses have narrowed significantly over the last year.

Elson argues that while the narrowing losses look like a positive trend, they do not tell the full story of the company's financial trajectory. He notes that the reality of the company's future costs is buried deep within the 300-page document rather than being presented on the front page.

The narrowing loss trend

The financial data shows a notable shift in how much money Anthropic is losing on an operating basis. In the first half of last year, the company reported losses exceeding $2 billion. However, in the first half of this year, those losses dropped to less than $700 million.

Many investors view this as a sign of improving efficiency or a stabilizing business model. They see the smaller loss as evidence that the company is getting closer to a sustainable path toward profitability as it scales its operations.

Elson warns that this trend might be misleading because it doesn't account for the massive spending that is scheduled to begin soon. He suggests that the current numbers are a snapshot of the past rather than a reliable predictor of the costs required to maintain growth.

A massive spending surge

One of the most significant details in the filing involves the amount of capital Anthropic intends to spend on AI compute. During the first half of this year, the company spent $8.9 billion on compute resources, which are the processing powers needed to run AI models.

However, the filing indicates that Anthropic plans to spend as much as $150 billion on compute next year. This represents a massive jump from their current spending levels, signaling a period of intense capital investment.

Elson points out that these commitments are not yet active, which is why they do not appear on current income statements. He notes that not enough investors recognize that these commitments are about to exist.

Building independent infrastructure

A key driver for this increased spending is Anthropic's need for its own hardware and infrastructure. Currently, the company relies heavily on cloud providers to run its services, but many of those providers are also direct competitors, such as Google.

To protect its interests and ensure reliable access to resources, Anthropic will need to start building out its own compute infrastructure. This shift from renting space to owning the underlying hardware will require the heavy capital expenditures mentioned in the filing.

Elson notes that the company is being transparent about these needs in the S-1 document. He highlights that while the information is scattered, the company does not hide the fact that its spending is about to surge as it builds out this independent infrastructure.

Broader market implications

The scale of Anthropic's planned spending is part of a larger trend involving other major AI players. Elson has previously raised concerns about the total amount of capital that companies like SpaceX and OpenAI will be injecting into the market.

He suggests that these massive capital injections could potentially threaten the momentum that has fueled growth for tech stocks recently. If multiple giants are spending hundreds of billions at once, it could create a different environment for the broader tech sector.

For the average investor, the takeaway is that Anthropic's growth will be fueled by high costs. While the company is narrowing its losses today, the path forward involves a significant commitment of capital to build the physical foundation of its AI services.