Two Roads to a Trillion: Anthropic and OpenAI Sprint to IPO With Opposite Strategies

Daniela Amodei stood on stage at the Bloomberg Tech conference in San Francisco and spelled out the difference between her company and its biggest rival in three sentences. “It’s a very capital-intensive business to train AI models,” she said. “The public market is very well-suited to that.” [1]

Days earlier, Anthropic had filed its confidential S-1 with the SEC, jumping ahead of OpenAI in a closely watched race to go public. The filing came right after the company closed a $65 billion Series H at a $965 billion valuation — the first time Anthropic had surpassed OpenAI’s roughly $852 billion private mark [2]. Multiple investors told TechCrunch the round was heavily oversubscribed, a clear sign of how hungry the private market is for AI exposure [2].

Amodei, who co-founded Anthropic in 2021 with her brother Dario and five other former OpenAI employees, built the company around a strategy that sounds almost contrarian in an industry built on trillion-dollar compute commitments. The phrase she keeps coming back to is “do more with less.”

The Efficiency Thesis

Anthropic runs on disciplined spending in a sector where the market leader acts like capital is free. OpenAI has committed roughly $1.4 trillion to compute infrastructure and data centers. Anthropic’s total compute commitments sit at about $100 billion [3].

“Anthropic has always had a fraction of what our competitors have had in terms of compute and capital,” Amodei told CNBC in January 2026. “And yet, pretty consistently, we’ve had the most powerful, most performant models for the majority of the past several years” [3].

The idea is not that scaling laws are wrong — Dario Amodei helped pioneer those during his time at OpenAI and Baidu. The bet is that brute-force size is not the only way forward, and that algorithmic efficiency, training data quality, and post-training reasoning can close the gap with less capital [3].

Anthropic’s training costs are estimated at $4.1 billion, dramatically lower than OpenAI’s range of $13 billion to $25 billion by similar measures [4]. The company is also available across every major cloud provider, giving enterprises flexibility without lock-in [3].

“It's a very capital-intensive business to train AI models,' she said. 'The public market is very well-suited to that.”

The Revenue Story

That efficiency bet is starting to show in the numbers. Anthropic’s annualized revenue hit $47 billion in May 2026, up from about $9 billion at the end of 2025 [2].

The company projects $10.9 billion in revenue for Q2 2026 alone, more than double Q1’s $4.8 billion. It also expects its first-ever operating profit of $559 million for that period [4].

The contrast with OpenAI is sharp. OpenAI is projected to lose $14 billion in 2026 and won’t reach positive cash flow until 2029 or 2030, according to HSBC estimates [4]. Its computing expenditure alone could hit $121 billion in 2028, with a loss of $74 billion that year [4].

Anthropic’s path to profit runs through the enterprise. Roughly 85 percent of its revenue comes from business customers, compared to OpenAI’s heavier mix of consumers. Over 500 companies now spend more than $1 million annually with Anthropic, and eight of the Fortune 10 are customers [4].

Enterprise contracts bring in three to five times more revenue per token than consumer subscriptions. Their query patterns are more predictable and cheaper to serve. The margins are better from day one [4].

The IPO Chess Match

The confidential S-1 filing is a strategic move as much as a financial one. If Anthropic gets to public markets before OpenAI, it gets to set the narrative for how AI companies should be valued — its growth story, its cost structure, its customer base, its path to profit. A first mover could cool investor appetite for whoever follows [5].

Prediction markets moved fast. Anthropic’s implied probability of confirming an IPO before November 1, 2026 jumped from 67.7 percent to 92.5 percent after the filing announcement, before settling at 88 percent. Its odds of reaching public markets ahead of OpenAI are now 78 percent [6].

Amodei would not comment directly on the rivalry when asked at Bloomberg Tech. “It gives us the option to potentially go public after the SEC review,” she said [1]. The company has not disclosed the size or terms of the planned offering.

The xAI Pivot

Anthropic is not building its own data centers. Instead, it struck a surprising compute deal with Elon Musk’s xAI, disclosed in SpaceX’s S-1 filing to cost $1.25 billion per month [2]. The deal fits Amodei’s stated preference: “We would much rather be on the side of having a little bit more demand for the product than we’re able to serve than the inverse” [2].

But the arrangement creates a dependency. A single provider supplying over a billion dollars per month in compute is concentrated risk, especially when that provider’s parent company is preparing its own IPO at the same time.

SpaceX is expected to debut on Nasdaq on June 12 [6]. Anthropic has also signed compute deals with Akamai Technologies, but the xAI agreement is by far the biggest.

The Profitability Question

Forbes framed the contrast between the two labs as a test of whether the market rewards discipline or scale. “The AI profitability race comes down to whether and how quickly the cost of serving intelligence can be brought below the revenue generated by deploying it,” wrote analyst Paulo Carvalho.

“Anthropic has always had a fraction of what our competitors have had in terms of compute and capital,' Amodei told CNBC in January 2026. 'And yet, pretty consistently, we've had the most powerful, most performant models for the majority of the past several years”

“As of this week, one of these companies has demonstrated that it can. The other is asking public market investors to believe it will” [4].

The Amazon comparison is useful but not exact. Amazon lost roughly $3 billion cumulatively over six years before turning its first annual profit in 2003. OpenAI is on track to pile up hundreds of billions in losses before reaching positive cash flow around 2029 or 2030 — a scale difference of roughly 100 times [4].

Anthropic’s projections show it reaching $17 billion in positive cash flow by 2028, with gross margins at 77 percent. Those numbers look more like enterprise software economics than infrastructure build-out [4].

What Comes Next

The confidential filing lets Anthropic keep its detailed financials private while the SEC reviews the paperwork. The company can pull the filing at any point. But the message is clear: Anthropic wants to be the first major AI lab answerable to public shareholders [5].

There are complications. Anthropic’s cybersecurity model Mythos has found thousands of high-severity vulnerabilities across every major operating system — which shows immense value but also raises opacity concerns for SEC review [6]. The company also faces a copyright lawsuit and a “supply chain risk” designation from the Pentagon [6].

Amodei’s pitch to investors is that the business world is still figuring out how to deploy AI effectively. The goal is for these tools to become a seamless part of daily work.

If she is right, the revenue growth that produced $47 billion in annualized run rate in under five years is just the start. If she is wrong, the margin for error is thin — but thinner for the company that spent less getting here.

“Regardless of how good the technology is,” she told CNBC, “it takes time for that to be used in a business or sort of personal context. The real question to me is: How quickly can businesses in particular, but also individuals, leverage the technology?” [3]