$65 billion in a single round — numbers that take your breath away

The announcement came on May 28, 2026, and immediately rewrote the rankings. Series H — the name of the current funding round — brought the company $65 billion and pushed its valuation to $965 billion (post-money). For comparison: OpenAI last secured investment at a valuation of $730 billion, which was the previous record in the AI sector.

According to the official announcement, the funds will go toward advancing safety research, expanding computing capacity, and scaling products including Claude Code and Claude Cowork. The New York Times additionally reported that Anthropic is planning an IPO in autumn 2026 — an offering that could be one of the largest in tech history.

Its growth rate is also worth noting: as recently as March 2025, the company had a valuation of $61.5 billion, by September 2025 it had reached $183 billion, and in February 2026 it hit $380 billion. A subsequent jump of $585 billion in three months is unprecedented in the tech world.

From OpenAI defectors to the world's most valuable AI company

The story of Anthropic is at its core a story of leaving OpenAI. In 2021, the sibling duo of Dario and Daniela Amodei — then VP of Research and VP of Safety — left OpenAI because of disagreements about the company's direction. Together with several other researchers, they founded Anthropic with a clear mission: to build AI systems that are safe, helpful, and honest.

Today, five years later, Anthropic employs over 2,500 people, is headquartered in San Francisco, and has developed the Claude model line — from Claude 1 (2023) through Claude 3.5 Sonnet to the latest Claude Opus 4.8, which was announced on the same day as the record-breaking investment.

The company has gradually attracted other key figures from OpenAI: Jan Leike (head of the alignment team), John Schulman (OpenAI co-founder), and last year also Andrej Karpathy, one of the most well-known faces in AI research. The team is rounded out by Mike Krieger, co-founder of Instagram, as Chief Product Officer.

The chip strategy: Broadcom, Google, and the end of dependence on Nvidia

While most AI companies are existentially dependent on Nvidia's graphics chips, Anthropic has bet on diversification. In April 2026, it announced an expanded partnership with Broadcom and Google (Alphabet) — starting in 2027, it will begin using its own TPU chips (Tensor Processing Units) for training Claude models, jointly developed by Broadcom and Google.

Anthropic isn't putting all its eggs in one basket: Claude models are currently trained on three types of chips — Nvidia GPUs, Google TPUs, and Amazon Trainium chips. This strategy gives the company independence and bargaining power. "If we were locked into a single supplier, they would have virtually unlimited pricing power over us," CEO Dario Amodei commented on the company's approach.

Broadcom, which develops TPU chips together with Google, expects that revenue from custom AI chips will exceed $100 billion by the end of 2027. In the first quarter of fiscal year 2026, the AI semiconductor division posted revenue of $8.4 billion — year-over-year growth of 106 percent.

TPU vs GPU: what's the difference?

Simply put: GPUs (Graphics Processing Units) from Nvidia are general-purpose computing units originally designed for graphics, which turned out to be remarkably effective for training AI models as well. TPUs, by contrast, are specialized chips designed to measure for the specific needs of particular AI models — in this case, Claude. The result can be higher performance with lower energy consumption, but at the cost of less flexibility.

Claude vs ChatGPT: what it does better and what worse

The latest model Claude Opus 4.8 brought two key innovations. The first is a "dynamic workflow" tool that enables automating complex workflows across different applications. The second is improved "honesty" of the model — Claude 4.8 is, according to Anthropic, significantly better at admitting when it's unsure of an answer rather than making one up (which is a long-standing problem with all large language models, including ChatGPT).

In practical comparison, Claude excels particularly at:

• Programming — Claude Code and Claude Cowork are among the most widely used AI assistants for developers
• Analyzing long documents — thanks to its large context window, it can handle hundreds of pages of text
• Safety and ethics — the model is trained using the Constitutional AI method, which instills clear ethical boundaries

ChatGPT from OpenAI, by contrast, leads in multimodal capabilities (generating images directly in chat), a broader range of plugins and integrations, and voice features. Both models are available for free in the basic version — Claude via the claude.ai website, with the premium Claude Pro subscription costing $20 per month (roughly 460 CZK).

What does this mean for regular users and businesses in Czechia?

For Czech users, there is good news: Claude is fully available in the Czech Republic, both through the web interface and through mobile apps for iOS and Android. The model communicates fluently in Czech, understands the context of Czech affairs, and handles even complex tasks in Czech — from drafting contracts through document analysis to programming.

For businesses, the key offering is Claude for Work — an enterprise subscription that includes higher limits, team collaboration, and administrator tools. Pricing starts at $30 per user per month. An API is also available (from $0.15 per million input tokens for Claude 3.5 Haiku), enabling integration of Claude into custom applications.

An important note for European businesses: as an American company, Anthropic processes data on servers in the US, which may be problematic from a GDPR standpoint. However, the company offers data processing in European data centers through a partnership with Amazon Web Services (AWS), which operates servers including a region in Frankfurt.

With its growing valuation and planned IPO, Anthropic can be expected to continue investing in its European presence — including localization and support for smaller languages such as Czech.

Safety, ethics, and the dispute with the Pentagon

Anthropic has from the start positioned itself as a "safety-first" AI company — and it recently defended this stance fiercely. In February 2026, it rejected a request from the US Department of Defense to remove contractual restrictions preventing the use of Claude for fully autonomous weapons and domestic surveillance. In retaliation, the Pentagon labeled Anthropic a "supply chain risk" and banned all military contractors from working with the company.

Anthropic appealed the decision in court — and in March 2026 obtained a preliminary injunction that blocked the Pentagon's ruling. The judge called the case a "classic retaliation violating the First Amendment." Tech giants including Google, Amazon, Apple, and Microsoft stood behind Anthropic.

This conflict illustrates a fundamental question facing the entire industry: how to balance commercial success, national security, and the ethical boundaries of AI? For now, Anthropic is betting that trustworthiness is a competitive advantage — and investors are clearly paying for it.