Anthropic is incurring nearly $10 billion in compute expenses to support its AI operations, with $8.95 billion in committed long-term contracts, reflecting aggressive investment ahead of its planned IPO in November. Despite current operating losses, investors accept the heavy upfront costs as necessary for scaling, with Anthropic strategically balancing premium short-term and favorable long-term cloud contracts to secure capacity and maintain competitiveness.
The discussion centers around Anthropic’s substantial compute expenses, which are nearing $10 billion, reflecting the company’s aggressive ramp-up in spending to support its AI operations. The business model inherently requires significant compute power for every token generated, leading to high costs. While these expenses currently contribute to operating losses, the expectation is that over time, Anthropic will achieve operating leverage and eventually generate free cash flow. For now, the company is focused on securing as much compute capacity as possible through contractual commitments, some of which are non-cancellable, to avoid turning away potential customers.
The timing of Anthropic’s upcoming IPO roadshow and pricing is also addressed. The roadshow is scheduled for the week of November 9th, with the stock expected to begin trading just before Thanksgiving. Despite the large numbers reported in the third quarter, these figures have not deterred investors or underwriters. The market appears to understand the necessity of heavy upfront investment in compute resources to build a scalable AI business, and there is a general acceptance of the risks involved.
The release of Anthropic’s third-quarter numbers was highly anticipated, especially since the initial S-1 filing did not include them. The delay in reporting these figures was likely due to timing issues around the quarter-end. Now that the numbers are public, they provide clearer insight into the company’s rapid increase in compute commitments and spending from Q2 to Q3. This transparency helps investors better gauge the scale and pace of Anthropic’s growth and operational strategy.
A key highlight is Anthropic’s disclosure of approximately $8.95 billion in aggregate committed compute spend as of September 30th. This figure is significant in the tech industry, indicating long-term contractual obligations that could amount to around $80 billion annually when spread over ten years. Such massive spending underscores the company’s commitment to securing the necessary infrastructure to support its AI workloads and maintain competitive advantage.
Finally, the discussion touches on the nature of these compute commitments, which involve major cloud hyperscalers like Amazon, Google, and CoreWeave. Due to supply and demand dynamics, Anthropic is currently paying premium prices for compute capacity, especially for contracts without long-term commitments. However, longer-term agreements with providers like CoreWeave offer more favorable pricing. This strategic mix of contracts allows Anthropic to balance cost and capacity as it scales its AI operations ahead of its anticipated IPO.