Anthropic Just Passed OpenAI's Valuation. Here is the Real Reason.
Tens of millions of people now open Claude the way they'd open a search bar — for drafting emails, debugging code, talking through a decision. Far fewer could tell you anything about the company behind it. That gap got a lot more interesting on May 28, 2026, when Anthropic closed a funding round that — according to Bloomberg and TechCrunch — pushed its valuation past $965 billion, surpassing rival OpenAI's valuation for the first time. Here's what's actually inside that number, and the part of Anthropic's own corporate structure that almost nobody outside AI policy circles ever talks about.
Anthropic's May 2026 Series H round valued the company at $965 billion — but the more consequential story is in how the company is structured to govern itself as it scales.
Anthropic was founded in 2021 by siblings Dario Amodei and Daniela Amodei, along with several colleagues who had worked together at OpenAI. Dario had served as OpenAI's VP of Research.
The founding story that's been widely reported: the group left after disagreements about the pace of commercialization relative to the level of safety work being built into increasingly capable models. Whatever the exact internal dynamics, the company that emerged made AI safety research — not just AI capability — a stated part of its core identity from day one.
🏢 What Anthropic Actually Is, in One Paragraph
Anthropic is an AI research and product company best known for Claude, its family of large language models, available at claude.ai, through mobile apps, via a developer API, and through enterprise cloud platforms including AWS Bedrock and Google Cloud Vertex AI. The company describes its mission as building reliable, interpretable, and steerable AI systems. Structurally, Anthropic is a Delaware Public Benefit Corporation (PBC) — a legal status that explicitly lets its board weigh public benefit alongside shareholder returns — paired with an unusual oversight body called the Long-Term Benefit Trust. As of May 2026, it's also one of the most valuable privately held companies in the world, reportedly preparing for a possible IPO.
The Valuation Climb — How Anthropic Got Here
📈 Anthropic's Valuation Trajectory — Late 2025 to Mid-2026
The Series H round — $65 billion raised — was co-led by Capital Group, Coatue, D1 Capital Partners, GIC, ICONIQ, and XN, with major participation from Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital, among dozens of other institutional investors. According to Anthropic's own announcement, the company's run-rate revenue crossed $47 billion around the time of the round.
According to Bloomberg, this round valued Anthropic above rival OpenAI for the first timeThe Claude Code Numbers — The Part of This Story That's Genuinely Wild
Buried inside Anthropic's own funding announcements is a statistic that's almost stranger than the valuation figure itself.
A coding assistant from a single company reportedly contributing to a meaningful share of the entire world's open-source commit activity, in well under a year of general availability, is the kind of number that would normally be the headline of its own article. Here, it's a footnote in a funding press release.
The Governance Structure Nobody Explains Well — PBC + The Long-Term Benefit Trust
This is the part of the Anthropic story that gets the least mainstream coverage, and it's arguably the most important for understanding how the company says it intends to behave as it gets larger and more powerful.
⚖️ The Two-Part Governance Structure
Delaware Public Benefit Corporation — directors may legally weigh public benefit alongside shareholder returns
"The responsible development and maintenance of advanced AI for the long-term benefit of humanity"
Independent trustees with no financial stake, intended to eventually elect a portion of the board
Can stockholders override or dissolve the Trust's authority via an undisclosed supermajority threshold?
🔬 The Critique Almost No Mainstream Article Mentions
Anthropic's PBC status and the Long-Term Benefit Trust are genuinely unusual — most AI labs don't build anything like this into their corporate paperwork. But a line of analysis that's circulated for a couple of years now, including a widely discussed 2024 LessWrong post and references in academic legal commentary, raises a specific structural question: Anthropic's stockholders can reportedly override or dissolve the Trust's authority through a supermajority vote — and the exact threshold for that vote has not been made public, because Anthropic has not published the full Trust Agreement. Given that Anthropic's investor base includes Google and Amazon among dozens of others, critics have pointed out that a sufficiently aligned bloc of major investors could, in theory, constitute such a supermajority. None of this means the Trust is meaningless — it has real, documented roles in Anthropic's Responsible Scaling Policy, including being consulted on policy changes and receiving regular safety updates. But the durability of the safeguard against a determined stockholder majority is a genuinely open, publicly debated question — and it's almost never mentioned in coverage that focuses on valuation numbers.
Constitutional AI and the Responsible Scaling Policy
Two specific frameworks show up repeatedly in Anthropic's public communications, and they're worth understanding on their own terms.
Constitutional AI — Training Claude Against a Written Set of Principles
Constitutional AI is Anthropic's approach to shaping model behavior using a written "constitution" — a set of principles drawn partly from sources like UN human rights frameworks and industry trust-and-safety norms. During training, the model is prompted to critique and revise its own outputs against these principles, and that AI-generated feedback is used alongside traditional human feedback. The practical upshot: a lot of Claude's behavior on sensitive topics is shaped by an explicit, at least partially documented framework — not purely ad hoc human labeling decisions made one prompt at a time.
Responsible Scaling Policy (RSP) — "Safety Levels" Borrowed From Biosafety Labs
Anthropic's Responsible Scaling Policy, first published in 2023 and now on version 3.0 (February 2026), defines a tiered framework of "AI Safety Levels" (ASL) explicitly modeled on the Biosafety Level (BSL) system used in biological research labs. Each tier specifies required security and safety measures that must be in place before Anthropic will train or deploy a model whose capabilities cross certain thresholds. Changes to the RSP require board approval "in consultation with" the Long-Term Benefit Trust — tying the two frameworks together.
What Most Anthropic Coverage Leaves Out Entirely
⚡ 1. Even the Name "Claude" Has Never Been Officially Confirmed by Anthropic
It's widely reported — including on Wikipedia and in encyclopedia entries — that Claude is named after Claude Shannon, the "father of information theory," whose 1948 work on probabilistic models of language anticipated the statistical foundations underlying modern LLMs, and whose master's thesis on Boolean logic helped make digital computing possible in the first place. It's a fitting namesake on multiple levels. The detail almost nobody mentions: Anthropic itself has reportedly never issued a formal public statement confirming this is the reason for the name. It's "widely accepted within the industry," per multiple sources — but one of the most-used product names in consumer tech currently rests on inference and internal anecdote rather than an official company statement.
⚡ 2. MCP Quietly Became Industry Infrastructure — Including for Competitors
In November 2024, two Anthropic engineers — David Soria Parra and Justin Spahr-Summers — released the Model Context Protocol (MCP), an open standard for connecting AI assistants to external tools, files, and data sources. Before MCP, every AI app needed bespoke, one-off integrations to connect to things like databases or business software. MCP defined a common protocol with SDKs across more than ten programming languages. What's notable: it was released as genuinely open infrastructure, and within roughly a year it saw adoption well beyond Anthropic's own products — including by other AI labs and developer tool ecosystems. A safety-focused lab ended up shipping one of the more consequential pieces of shared AI infrastructure of the decade, almost as a side project relative to the headlines about model releases and funding rounds.
🔬 3. Anthropic Has Been Trying to "Read the Mind" of Its Own Models
A significant and growing share of Anthropic's published research falls under mechanistic interpretability — literally trying to reverse-engineer what's happening inside a trained model, neuron by neuron and feature by feature. In 2024, this research produced a public demo that got a lot of attention for being genuinely strange: researchers identified a specific internal "feature" inside a Claude model that corresponded to the concept of the Golden Gate Bridge, then artificially amplified it — producing a version of the model that would bring up the Golden Gate Bridge in almost any conversation, regardless of topic. It was a vivid, slightly funny public demonstration of a serious research direction: if you can identify and manipulate individual concepts inside a model, you've made real progress on a problem — understanding what's actually happening inside these systems — that the entire field has historically treated as a black box.
⚡ 4. The Series H Investor List Reads Like a Semiconductor Supply Chain
Most coverage of Anthropic's funding rounds focuses on the venture firms. Less noted: the Series H round reportedly included strategic investment from Samsung, SK Hynix, and Micron — major memory and chip manufacturers. Their participation is a signal worth sitting with: the AI boom isn't just a software story funded by software-focused venture capital anymore. It's increasingly a hardware supply-chain story, where the companies that make the memory chips going into AI data centers have a direct financial stake in a frontier AI lab's continued growth — and vice versa. When the inputs to AI (chips, memory, power) and the AI labs themselves start showing up on each other's cap tables and customer lists, it's a structural shift worth tracking independently of any single valuation headline.
The Honest Assessment — What Stands Out About Anthropic's Approach
✅ What's Genuinely Distinctive
- PBC structure + Long-Term Benefit Trust is a real attempt at novel AI governance — most labs have nothing comparable
- Responsible Scaling Policy is a published, versioned, board-accountable framework — not just a blog post
- Constitutional AI makes a meaningful part of model behavior shaping explicit and documented
- MCP was released as genuinely open infrastructure that benefits the whole ecosystem, not just Anthropic
- Mechanistic interpretability research is published openly, including findings that aren't flattering or simple
- Claude Code's enterprise adoption and GitHub-commit share reflect substantial real-world utility, not just hype
⚠️ Open Questions Worth Tracking
- The Long-Term Benefit Trust's actual enforceability against a stockholder supermajority remains publicly undisclosed
- Major investors (Google, Amazon, and now chip manufacturers) create overlapping commercial relationships and incentives
- A near-$1 trillion valuation and IPO preparations create pressure that's historically hard for any company to fully insulate from product and safety decisions
- Constitutional AI's specific constitution document has been only partially public over time
- Rapid revenue growth (run-rate crossing $47B) means the company is scaling commercial operations very fast alongside its safety commitments
- As with any frontier lab, independent verification of internal safety processes is inherently limited from the outside
⚠️ What an IPO Would Actually Change
Anthropic reportedly engaged advisors for IPO preparation as part of its funding trajectory. Going public would not, by itself, change Anthropic's PBC status — Delaware PBCs can and do go public while retaining that structure. But it would introduce new pressures: quarterly earnings cycles, public shareholder activism, and analyst expectations that operate on very different timescales than multi-year safety research programs. The PBC structure gives Anthropic's board legal latitude to weigh public benefit against shareholder returns — it doesn't eliminate the pressure that comes from having a much larger, more liquid, more vocal shareholder base. Whether the governance structure described above holds up under that pressure is likely to become a much more concrete, testable question if and when an IPO actually happens — rather than the largely theoretical one it has been so far.
🔧 Want to see what frontier AI models can actually do when properly instructed?
Stop limiting Claude and GPT with basic questions. Use the free AI Super Prompt Generator at Solid AI Tech to instantly engineer high-precision, research-backed system prompts built for your exact professional workflows. No sign-up required.
Try the Free AI Super Prompt Generator →Frequently Asked Questions
What is Anthropic and what does it do?
Anthropic is an AI safety and research company founded in 2021 by siblings Dario and Daniela Amodei and former OpenAI colleagues. Its flagship product is Claude, available via claude.ai, mobile apps, an API, and enterprise platforms (AWS Bedrock, Google Cloud Vertex AI). Anthropic is structured as a Delaware Public Benefit Corporation with a stated mission of building reliable, interpretable, and steerable AI for the long-term benefit of humanity.
What is Constitutional AI?
Constitutional AI is Anthropic's training approach where Claude critiques and revises its own outputs against a written set of principles (a "constitution") drawn from sources like UN human rights frameworks and industry safety norms. This AI-generated feedback (RLAIF) is combined with human feedback (RLHF) during training, shaping much of Claude's behavior on sensitive topics through an explicit, documented framework rather than purely case-by-case labeling.
What is Anthropic's corporate governance structure?
Anthropic is a Delaware Public Benefit Corporation (PBC), legally allowing directors to weigh public benefit alongside shareholder returns. It also created the Long-Term Benefit Trust (LTBT) — independent trustees intended to eventually elect part of the board to keep safety considerations represented. A documented 2024 critique argues stockholders can reportedly override the Trust via an undisclosed supermajority threshold, and the full Trust Agreement hasn't been published — raising open questions about long-term enforceability, especially given major investors like Google and Amazon.
How much is Anthropic worth and who has invested in it?
As of late May 2026, Anthropic's Series H round raised $65 billion at a $965 billion post-money valuation — per Bloomberg, surpassing OpenAI's valuation for the first time. The round was co-led by Capital Group, Coatue, D1 Capital Partners, GIC, ICONIQ, and XN, with Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital among many other participants, plus strategic investment from Samsung, SK Hynix, and Micron. Run-rate revenue reportedly crossed $47 billion around the announcement, following a $380B valuation in February 2026 (Series G).
What is the Model Context Protocol (MCP)?
MCP is an open standard Anthropic released in November 2024 (created by engineers David Soria Parra and Justin Spahr-Summers) for connecting AI assistants to external tools, files, and data sources via a common protocol instead of bespoke integrations. Released with SDKs across many languages, it was adopted relatively quickly across the broader AI industry — including by other labs and developer tool ecosystems — becoming one of the more consequential open infrastructure contributions from any frontier AI lab.
