Anthropic Races Toward Public Markets as Compute Costs Soar

Daniela Amodei doesn’t flinch. At the Bloomberg Tech conference this week, the Anthropic president and co-founder faced questions about artificial intelligence’s towering expenses. Her answer was direct. “It’s a very capital-intensive business to train AI models,” she said, according to a Bloomberg report. The public market, she added, is “very well-suited to that.”

Days earlier the company had confidentially filed paperwork for an initial public offering. The move came hot on the heels of a $65 billion funding round that pushed its valuation to $965 billion. Demand for that round was greatly oversubscribed, multiple investors told TechCrunch. Now the race to public listing is on. And Anthropic appears determined to reach the market ahead of rival OpenAI.

The filing gives the company flexibility. “It gives us the option to potentially go public after the SEC review,” Amodei told the audience in San Francisco, per Bloomberg. She declined to say more on timing. Market conditions will decide. Yet the signal is unmistakable. A firm once content to stay private now sees equity markets as essential fuel for its next phase.

Amodei and her brother Dario founded Anthropic in 2021 after leaving OpenAI. They brought with them a group of researchers worried about the pace and safety of AI development. The new venture promised a different path. One focused on constitutional AI. One that put guardrails first. That focus remains. But the financial reality has shifted. Training successive generations of Claude models demands staggering sums for chips, data centers and power.

Anthropic has raised tens of billions from Amazon and Google. Those partnerships provide cloud credits and distribution. Still the bills keep growing. Amodei has long argued her company spends less than some peers while delivering strong results. In a January interview she told CNBC that the next phase of progress belongs to those who achieve the most capability per dollar of compute. Not necessarily the biggest pre-training runs.

That efficiency bet has kept Anthropic competitive. Its Claude models rank among the strongest available. Enterprise customers like the thoughtful, less hallucinatory responses. Governments and regulated industries appreciate the safety emphasis. Yet competitors pour money into ever-larger clusters. OpenAI has signaled massive compute commitments. The gap in spending invites skepticism about returns. Amodei shrugs it off.

She pointed to disciplined capital allocation. The company prefers to have slightly more demand for its products than it can immediately serve rather than overbuying infrastructure, she has said in recent appearances. That approach shows restraint in an industry prone to exuberance. Investors appear to agree. The latest round sailed past expectations despite the enormous headline number.

The IPO filing itself was no surprise to close watchers. The New York Times reported the confidential submission on June 1, noting it sets the stage for one of the largest offerings in history. A public debut could come as soon as this fall, though nothing is locked. Anthropic’s statement was careful. The filing “gives us the option to go public” after regulatory review. Timing depends on markets and other factors.

Wall Street is gearing up. Morgan Stanley and Goldman Sachs are expected to lead the deal, according to people familiar with the matter cited by Bloomberg. The offering would join a select group of potential 2026 blockbusters that includes SpaceX and possibly OpenAI. NVIDIA chief Jensen Huang recently remarked that buying shares in these companies at IPO could feel like getting in early on Amazon, Google or Meta. His comments, widely shared on X in recent days, added fresh fuel to the hype.

But not everyone is convinced the economics will hold once public scrutiny intensifies. AI companies burn cash at extraordinary rates. Revenue grows fast yet profitability remains elusive for frontier labs. Amodei acknowledges the challenges. She also argues that public markets bring discipline. They demand clear reporting, steady progress and realistic guidance. For a company that has operated with heavy backers and limited disclosure, the transition will test its culture.

Safety remains central to her message. Anthropic has repeatedly delayed or limited release of its most powerful models to allow for better testing and controls. Recent reporting from Forbes Australia highlighted Amodei’s comments on keeping the advanced Mythos model under tight restrictions even as benchmarks suggest it leads the field. That caution wins praise from risk-conscious stakeholders. It can frustrate customers eager for the latest capabilities.

The tension sits at the heart of the business. Move too slowly and rivals capture market share. Move too quickly and unintended consequences multiply. Amodei has steered the company through this narrow channel since day one. Her steady presence provides continuity as the firm scales toward thousands of employees and billions in annual revenue.

Enterprise adoption of Claude has accelerated. Developers build agents and workflows around its API. Large corporations integrate it into knowledge work while citing its refusal to engage in harmful tasks. That reliability carries a premium. It also supports higher pricing in certain segments. The model helps explain why investors keep writing nine- and ten-figure checks.

Competition has only grown fiercer. OpenAI’s GPT series, Google’s Gemini, Meta’s open source efforts and a host of startups all vie for attention. Compute suppliers remain constrained. NVIDIA chips trade at a premium. New clusters take years to build. Power availability in desirable locations has become a binding constraint. Against that backdrop Amodei’s focus on algorithmic efficiency looks less like a philosophical choice and more like a survival imperative.

She has avoided direct comparisons on who might list first. The question came up at the conference. Her response stayed measured. The confidential filing simply preserves options. Yet the timing tells its own story. By moving now Anthropic forces OpenAI to consider its own plans. It also gives investors a clearer view into the financials of a genuine AI pure play before the full weight of 2026’s mega-offerings hits the calendar.

Analysts expect the S-1, when it becomes public, to reveal details on revenue trajectory, customer concentration, research spending and governance structure. Current investors will look for ways to monetize portions of their stakes. Employees holding equity will gain liquidity. The broader market will gain a new bellwether for AI economics. Success could open the door for additional lab listings. Failure would cast a long shadow.

Amodei shows little appetite for drama. Her comments this week were pragmatic. High costs are real. Public capital can help. The business requires patience and large bets. Returns will come from turning intelligence into products that solve expensive problems. She points to coding, scientific research and complex analysis as areas where Claude already delivers measurable value.

Still the numbers are eye-watering. A single training run for frontier models can cost hundreds of millions. Inference at scale adds more. Data center construction runs into billions. The $65 billion round announced last month, covered extensively by TechCrunch, reflects the scale now required to stay relevant. Few private companies have ever raised so much.

That reality explains the IPO pivot. Private rounds cannot continue indefinitely at these levels without pressuring valuations or diluting founders excessively. Public markets offer a different bargain. They provide permanent capital in exchange for transparency and quarterly accountability. For Anthropic the trade-off appears worth making.

Whether the public will embrace the stock is another question. Valuations at these heights leave little room for error. Any slowdown in progress or regulatory setback could trigger sharp moves. Amodei and her team know this. They have spent years building a reputation for seriousness in an industry often accused of overpromising.

The coming months will test that reputation. The SEC review process will probe disclosures on risks, competition, intellectual property and dependence on big cloud partners. Investors will demand clarity on the path to sustainable margins. Employees will watch how leadership balances growth with the original safety mission.

And through it all Daniela Amodei will likely maintain her calm demeanor. She has done so before. When the company split from OpenAI. When it accepted massive checks from tech giants. When it chose to hold back models that could have generated more immediate revenue. Consistency defines her approach.

The AI sector stands at an inflection point. Capital intensity has reached levels once associated with semiconductor fabrication or oil exploration. Returns remain theoretical for many applications. Yet the upside if models continue improving is immense. Amodei’s willingness to embrace public markets suggests she believes the industry’s biggest players must eventually operate under its scrutiny.

Her message this week was clear. The costs are high. The public market can handle them. Anthropic is ready to show its numbers. The rest will depend on execution, regulation and continued technical progress. For an industry that has operated largely behind closed doors, the transition promises to be revealing.

Watch the filings closely when they emerge. Listen to the earnings calls that will eventually follow. The numbers will tell their own story about whether efficiency can win against brute force scale. Amodei has placed her bet. Now the market gets its chance to judge.