Record round and who's behind it

The financing was led by a quartet of prominent Silicon Valley investors: Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital. The round also included Capital Group, Coatue, GIC (Singapore's sovereign wealth fund), D1 Capital Partners, ICONIQ, and Fidelity. It also includes $15 billion from previously committed investments by cloud giants — of which $5 billion from Amazon, which remains Anthropic's primary cloud and training partner.

A new development is the entry of semiconductor manufacturers Micron, Samsung, and SK Hynix — companies whose memory and storage chips are essential for running models. These are strategic infrastructure partners that will help Anthropic scale computing capacity according to growing demand. And that demand is enormous: the company announced that its annual run-rate revenue in May exceeded $47 billion.

Claude is on all three major clouds for the very first time

An important milestone is that Claude has become the first frontier model available simultaneously on Amazon Web Services, Google Cloud, and Microsoft Azure. That alone shows how deeply Anthropic has already penetrated the enterprise ecosystem. AWS remains the primary cloud partner, but the company recently signed a deal with Google and Broadcom for 5 gigawatts of next-generation TPU capacity and with SpaceX for access to GPUs in the Colossus 1 and 2 supercomputers.

For European and Czech companies, this means Claude is no longer tied to a single cloud platform. If your company runs on Azure, Google Cloud, or AWS, you have direct access to the model via API.

Claude Opus 4.8: A smaller leap, but an important shift in honesty

Alongside the funding announcement, Anthropic released Claude Opus 4.8 — a new version of its most powerful model. Pricing remains the same as its predecessor: $5 per million input tokens and $25 per million output tokens. Fast mode is now three times cheaper than before.

The improvements aren't as dramatic as with previous generations, but note one thing: according to internal tests, Opus 4.8 is four times less prone to "overlooking" errors in its own code. Early testers from companies like Cursor, Devin (Cognition), Databricks, Hebbia, and Thomson Reuters confirm that the model better recognizes uncertainty and is less likely to confidently assert nonsense.

In agent benchmarks, Opus 4.8 even beats GPT-5.5 — on the Online-Mind2Web test for browser control it scored 84%, on the Super-Agent benchmark it was the only one to complete all cases, and on Terminal-Bench 2.1 for programming tasks it scored 78.9%. For the very first time, the model broke the 10% threshold on the rigorous Legal Agent Benchmark for legal practice.

Opus 4.8 handles Czech very well — just like its predecessors. Claude is among the models with the strongest support for Slavic languages, and Czech has traditionally achieved high quality with it, whether for translations, text summarization, or content creation.

Mythos heading to all customers

Anthropic also confirmed that Mythos — its most powerful model yet with advanced cybersecurity capabilities — will be available to all customers within weeks, not just selected partners within the Project Glasswing program. It is Mythos that has stirred the most concern among experts and politicians in recent months, because it can find software vulnerabilities at a pace that human teams cannot match.

The announcement comes at a time when Anthropic is also engaged in a legal dispute with the Pentagon. The U.S. Department of Defense has labeled the company a supplier risk — according to Anthropic, an unconstitutional retaliation for refusing to give the military unrestricted access to its models.

SpaceX, IPO, and the trillion-dollar club

Anthropic is not the only AI giant heading for the stock market. SpaceX, which absorbed Elon Musk's xAI in February, could — according to Reuters — begin trading as early as June 12, 2026 with a target valuation of approximately $1.75 billion. Both OpenAI and Anthropic are planning their own IPOs later this year. Together, all three companies could create a combined market value exceeding $4 billion — more than Germany's GDP.

For investors, it's an opportunity but also a risk: valuations are growing at a pace unprecedented in tech history. OpenAI reached its current value in ten years, Anthropic in five. Yet both companies are still massively investing in infrastructure, with profits nowhere in sight.

What it means for Czech companies and users

Increasing competition among Anthropic, OpenAI, and Google means one thing: models are getting faster, cheaper, and more accessible. Claude is now available through the web interface at claude.ai, via API on all major clouds, and thanks to Claude Code also as a developer tool. Czech companies that use Claude — for example Ecomail, which integrated it into its email marketing tool — thus benefit from every intergenerational improvement without needing to change anything.

In the context of the EU AI Act, it's important that Anthropic has prioritized safety from the beginning. The company has a detailed Responsible Scaling Policy, model cards (system cards) with safety test results, and for Mythos it is even delaying public release until additional cybersecurity safeguards are complete. This is exactly the type of approach that European regulation requires.