Anthropic’s planned public-market debut, reportedly under consideration for November 2026 at a valuation reaching as high as $2 trillion, would mark an extraordinary transformation for a company created by seven OpenAI executives after an abrupt December 2020 break. The group left after being removed from an all-hands video call and losing access to OpenAI’s internal systems, then built a company around the argument that increasingly powerful AI models required stronger safety controls and more cautious governance.
The scale now associated with Anthropic also illustrates the capital demands of frontier AI. Documents described in the account put the company’s 2025 revenue at $4.6 billion, alongside a $42 billion accounting loss. Roughly half of its operating expenditure was directed toward chips and cloud computing, according to the same materials, reflecting the cost of training and serving large models such as Claude.
A flotation at a multi-trillion-dollar valuation would place Anthropic among the largest potential technology listings in history. It would also intensify scrutiny of whether revenue growth can eventually offset the computing, data-center, and energy costs attached to the race to build more capable AI systems.
A dispute over OpenAI’s direction
Anthropic emerged from disagreements that had been developing inside OpenAI for years, centered in part on Dario Amodei, OpenAI’s former vice president of research. Amodei played a major role in work on GPT-2 and GPT-3, and had argued internally that scaling computing power, data, and model parameters would produce sharp gains in AI capability.
His 2017 internal paper, titled “Big Blob of Compute,” set out that thesis before large language models became the commercial centerpiece of the AI sector. The paper’s thinking helped steer research toward larger models, but it also raised the question that would later divide OpenAI staff: how should an organization control systems whose capabilities were advancing rapidly?
A major early dispute involved a fundraising proposal known internally as the “Countries Plan.” Five people familiar with the discussions described it as a concept to auction future rights connected to artificial general intelligence, or AGI, to government buyers including the United States and other major powers.
In a December 2017 email sent to OpenAI co-founders Greg Brockman and Ilya Sutskever, among others, Amodei warned that such an approach could aggravate an international AI arms race, facilitate human-rights abuses, and damage OpenAI’s moral credibility.
The disagreement was not solely about fundraising. It concerned which institutions should gain access to high-capability AI and whether commercial partnerships could coexist with the group’s safety commitments.
Governance and access battles expanded
Arguments over those questions continued during GPT-3’s private testing phase. OpenAI staff debated lists of approved customers, according to the account, as the company weighed limited access against demand for the model.
Microsoft’s proposed $1 billion investment in OpenAI became another pressure point. Amodei pushed to preserve a provision in OpenAI’s charter commonly referred to as “merge-and-assist.” The clause contemplated a scenario in which OpenAI could pause a competitive effort and help another project if it believed that project had a better chance of developing AGI safely.
The provision reflected an unusually cooperative view of a field now dominated by intense competition for chips, data-center capacity, researchers, and corporate customers. As commercial AI accelerated, preserving such a commitment became harder to reconcile with the incentives of a rapidly expanding technology company.
A separate 2020 episode involved a meeting with rapper Kanye West and a proposal for participation in GPT-3 testing. Jack Clark, OpenAI’s policy lead at the time, later blocked the proposal, according to the account. The incident added to internal concerns over who should receive early access to potentially powerful models and how those decisions should be made.
By then, Dario Amodei and his sister, Daniela Amodei, had established a close internal group working heavily on safety questions. Their colleagues used private Slack channels to coordinate work and discuss the direction of AI development.
The December 2020 split
Dario Amodei had considered joining DeepMind in 2019 and later discussed creating a new research lab with colleagues including Jared Kaplan, a physics professor. By 2020, a group of seven eventual Anthropic co-founders had organized around a private Slack channel called “BATNA,” shorthand for “best alternative to a negotiated agreement.”
The departing team and OpenAI negotiated separation terms before the split. The arrangement reportedly included a period in which the new venture could recruit OpenAI employees, followed by a one-year non-solicitation commitment. It also allowed the group to take some internal research materials and bound both sides to mutual non-disparagement terms.
At the December 2020 all-hands meeting, Amodei told staff that he and six other senior leaders would resign to create a safety-focused AI company. Within minutes, the seven were removed from the video call. Their email, Slack, and laptop access was disabled, with company devices remotely locked.
The group initially considered naming its new company Sparrow Systems before choosing Anthropic. In a 2021 founding vision document, Amodei envisioned a safety research laboratory of about 50 people and estimated that building AGI could require roughly $10 billion.
That forecast proved conservative in one respect: Anthropic’s staffing and infrastructure needs grew far beyond the early blueprint. The company had about 160 employees by the summer of 2023, according to the account. By winter 2025, it employed thousands, with restricted access and tighter physical security on floors used to train newer versions of Claude.
From safety lab to capital-intensive challenger
Anthropic’s expansion has placed it in the same resource contest that its founders once sought to approach more cautiously. Training frontier models depends on large supplies of advanced processors and cloud capacity, giving a small number of chip makers and infrastructure providers substantial influence over which AI developers can compete.
The company’s reported losses should be read alongside the high upfront cost of building AI infrastructure. Revenue can grow quickly when businesses adopt AI tools, while expenditure arrives earlier through long-term cloud contracts, hardware purchases, model training, and research hiring.
Concentration among customers may become a central issue in any public listing. The account says nearly one-quarter of Anthropic’s revenue comes from two unnamed buyers. Such dependence can create substantial exposure if a major customer reduces spending, renegotiates contracts, or develops competing systems internally.
Anthropic’s origins also remain visible in its internal culture. During OpenAI’s 2023 board crisis, Anthropic employees reportedly considered deleting a Slack message in which Dario Amodei called OpenAI “evil,” fearing that it could become public despite the earlier non-disparagement agreement.
For cryptocurrency markets, the more direct connection is likely to be competition for capital and computing resources rather than an automatic impact on token prices. AI-linked crypto assets have shown sharp swings during changes in public enthusiasm for the sector, but a public offering by Anthropic would not by itself determine demand for those tokens. The company’s prospective listing would instead offer public markets a large, conventional equity vehicle for exposure to frontier AI—one built by executives who left OpenAI over questions that remain unresolved: who controls advanced models, who benefits from them, and how much risk commercial competition should be allowed to create.
To explore how AI shapes modern finance and trading, dive into this deep-dive on AI–blockchain convergence now.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
