Artificial intelligence research lab Anthropic has officially submitted its draft S-1 registration statement to the U.S. Securities and Exchange Commission (SEC) ahead of its initial public offering (IPO), disclosing its financial performance and compute expenditures for the first time.

Anthropic Unveils IPO Filing: $16.3B First-Half Revenue and Multi-Chip Compute Strategy — Silicon wafer discs manufactured in advanced semiconductor facilities for AI compute acceleration.

According to the filing, Anthropic recorded $16.3 billion in revenue for the first half of 2026, while spending $8.9 billion on cloud computing infrastructure and specialized AI accelerators to meet soaring enterprise demand for its Claude models.

A key revelation in the prospectus is Anthropic's 'fungible compute pool' strategy spanning four major silicon families. By diversifying its workload infrastructure across Nvidia GPUs, Google TPUs, Amazon Trainium, and AMD accelerators, the company has effectively mitigated hardware supply constraints and reduced dependence on single-vendor pricing.

Wall Street analysts and tech industry leaders view the detailed financial disclosures as a major milestone, offering the first transparent benchmark for the revenue growth and infrastructure economics of frontier AI model developers.

An Anthropic spokesperson stated that continuous compiler optimization across heterogeneous hardware architectures serves as a cornerstone for building a sustainable and cost-effective AI enterprise business.