Anthropic Aims for Nearly $1 Trillion IPO Valuation
Anthropic is reportedly in intensive communication with potential investors, planning to launch its IPO as early as September or early October, with a private valuation of approximately $965 billion.
Investors have focused on three main risks during the pre-roadshow: competition from low-cost models in China, strained relations with the Trump administration, and local resistance to data center construction in the U.S.
Anthropic executives downplayed the impact of China's low-cost models, asserting that enterprises will continue to pay for the strongest cutting-edge models with high-value users; the company will maintain its competitive core on model capability, safety, and enterprise products.
The company stated that its annualized revenue run rate surpassed $47 billion in May. This metric annualizes the latest monthly revenue and is not confirmed actual revenue for the past 12 months; its annualized run rate was $30 billion in April.
Computing costs are rising rapidly: Anthropic reportedly pays SpaceX $1.25 billion per month for full access to the Memphis Colossus 1 data center's computing capacity, gaining an additional 300 megawatts of capacity within a month.
In market mechanisms, IPO investors are purchasing future cash flows supported by Claude Code and rapidly growing enterprise demand, but they must bear the risks of model price declines, massive computing commitments, and delays in infrastructure licensing. Funding will shift from the secondary market and private investors to the public market, with the company benefiting from expanded financing channels, while its valuation will be pressured by whether revenue growth can continue to outpace computing and capital expenditure growth.
Source: Public Information
ABAB AI Insight
Anthropic's financing path has shifted from a model development company to a user of large-scale infrastructure. The company completed approximately $30 billion in financing in February 2026, with a valuation of $380 billion; in May, it further completed $65 billion in financing, raising its valuation to about $965 billion. The core of the short-term valuation jump is not a single model release, but the acceleration of revenue run rate driven by enterprise subscriptions and Claude Code.
Claude Code has changed Anthropic's capital structure. The product's annualized revenue exceeded $2.5 billion in February 2026, doubling since the beginning of the year; the number of customers with annual spending over $100,000 has increased sevenfold year-on-year. The code agent converts the token consumption of general models into high-frequency, rigid expenditures in engineering budgets, making it easier to form predictable revenue compared to consumer-facing chat products.
In historical comparison, traditional SaaS companies first establish software revenue and then gradually expand cloud resources; Anthropic, however, locks in heavy asset computing contracts in the high revenue growth phase. Unlike early cloud computing companies like Amazon and Microsoft, which used self-built infrastructure to serve external customers, Anthropic needs to purchase computing resources from suppliers like Google Cloud, AWS, and SpaceX to maintain model training and inference supply.
This represents capital concentration. The competition in cutting-edge AI is no longer just about algorithms, but is determined by financing ability, GPU supply, power capacity, data center licensing, and enterprise distribution. Companies that can sign multi-billion dollar long-term computing contracts before revenue is fully realized will secure their model iteration and service delivery capabilities; competitors unable to obtain equivalent capital and infrastructure are more likely to be forced to turn to open-source, low-cost, or vertical application markets.
ABAB News · Cognitive Laws
The moat of AI is first computing contracts, then model capability
Revenue growth proves demand, capital expenditure determines survival
The closer the valuation is to a trillion, the less growth can rely on the past