Anthropic President and co-founder Daniela Amodei is pushing back against doubts over whether the artificial intelligence boom can produce strong financial returns, arguing that the technology is still early in its commercial cycle as the Claude maker moves toward a closely watched public listing.
Speaking at Bloomberg Tech in San Francisco, Amodei said companies building frontier AI models will need access to large pools of capital because training and running advanced systems remains expensive. Her comments come as Anthropic has confidentially filed for a U.S. initial public offering, a move that could test public-market appetite for AI companies at a time when investors are paying closer attention to infrastructure costs, token usage, and profitability.
“It’s a really big upfront cost to train the models and to serve inference on them,” Amodei said. “My guess is that over time, the sort of core set of companies that are working to advance the frontier are just going to need access to capital, and I think the public market is very well suited to that.”
Anthropic Moves Toward Wall Street
Anthropic’s confidential IPO filing marks a major step for the San Francisco-based AI company, founded in 2021 by former OpenAI employees, including Daniela Amodei and her brother, CEO Dario Amodei. The company is best known for Claude, its family of AI models used for coding, research, writing, enterprise automation, and customer support.
The filing follows a major funding round that lifted Anthropic’s valuation sharply. The IPO terms, including the number of shares and expected pricing, have not been publicly disclosed.
Why Capital Is Central to the AI Race
Amodei’s remarks point to the core reason Anthropic may turn to public markets: frontier AI companies need enormous financial resources before returns become fully visible. Unlike traditional software firms, AI labs must spend heavily on chips, cloud capacity, model training, safety research, and inference delivery.
That cost structure has become one of the biggest questions around the AI boom. Investors are no longer asking only whether AI models are impressive. They are asking whether companies can convert model usage into durable profit after accounting for the cost of serving queries, running coding agents, and supporting enterprise customers.
Anthropic faces the same economic challenge as its rivals: every advanced model interaction consumes compute. That makes AI revenue different from classic software subscriptions, where margins often improve sharply once the product is built.
Amodei Says the Market Is Still Early
Amodei pushed back against the idea that current concerns over AI spending prove the technology is overhyped. She argued that the market is still in an early stage and that current tools should not be treated as the final form of AI.
“I actually think there’s a lot more distance to go still for what the models will be able to do two to four, six to eight years in the future,” Amodei said.
That view matters because much of the current debate around AI returns is based on today’s productivity gains. Companies are experimenting with AI assistants, coding agents, document tools, customer-service automation, and workflow copilots, but many are still working out how to measure real business impact.
Token Usage Becomes a Pressure Point
Her comments also come amid growing discussion around “tokenmaxxing,” a Silicon Valley term used to describe companies or employees trying to maximize AI token usage as a visible sign of adoption. Tokens are the units AI systems process when they read prompts and generate responses. Since many AI companies charge based on token volume, heavy usage can quickly become expensive.
The debate has raised a difficult question for businesses: does higher AI usage mean higher productivity, or simply higher cost?
Amodei acknowledged that some workers feel pressure to use AI tools because employers want adoption to rise. “Today there’s this feeling that’s like, ‘Oh, like AI, you know, the leaderboards, and it’s like I have to use it, and what am I going to use it for?’” she said.
She added that Anthropic does not run a forced usage culture around its own products. “But there’s not like, ‘You must use AI and you must use Claude,’” she said.
Claude Code Highlights the Cost Debate
Claude Code, Anthropic’s AI coding product, has become one of the tools most closely associated with heavy token usage. Coding agents can consume more tokens than a simple chatbot because they read files, reason through problems, generate code, test outputs, and work through several steps autonomously.
That makes coding one of the clearest examples of both AI’s promise and its cost problem. If an AI tool helps a developer complete complex work faster, the return can be meaningful. But if a company measures success only by token volume, it may reward consumption rather than useful output.
Amodei’s response suggests Anthropic wants the market to move beyond token counts as a crude productivity metric. “My hope is that over time it’ll be more incorporated into the day-to-day of how humans do our work, how we communicate together, and that there will actually be a lot more value realized in a way that feels really good to people,” Amodei said.
A Major Test for Public Investors
Anthropic’s potential listing could become a major test of how public investors value frontier AI companies. Private investors have shown strong appetite, but public markets usually demand more transparency around revenue, margins, cash burn, customer concentration, compute commitments, and long-term profitability.
If Anthropic performs well, it could strengthen confidence in the AI sector and open the door for other major AI listings. If investors push back, it could force a broader reassessment of how much value should be assigned to companies that still require massive infrastructure spending.
Investors Want Proof Beyond Hype
The key question is whether Anthropic can show that AI demand is turning into sustainable economics. The company has benefited from enterprise interest in Claude, particularly among developers and businesses looking for advanced reasoning and coding support. But the costs of training and serving frontier models remain steep.
That is why Amodei’s message matters. She is not denying that AI is expensive. Instead, she is arguing that the expense is part of building the frontier, and that public markets are better suited to funding companies with long-term infrastructure demands.
What Comes Next
Anthropic’s IPO process is still at an early stage, and the company has not released a public prospectus. When it does, investors will be able to examine the company’s financials, risk factors, customer base, infrastructure commitments, and growth assumptions in more detail.
Until then, Amodei’s comments offer the clearest view of how Anthropic wants to frame the debate. The company sees AI as a long-term infrastructure race, not a short-term software fad. It believes model capability still has significant room to improve, and it sees public capital as a logical next step for companies trying to build at the frontier. The IPO will show whether public investors are ready to make the same bet.
Comments