Investors' imaginations ran wild in recent weeks wondering what Anthropic's business would look like under the hood when it finally filed for its long-anticipated initial public offering. The numbers are solid, it turns out, but investing in Anthropic will still require a leap of faith.

The artificial-intelligence lab released its S-1 IPO prospectus on Monday, setting in motion a likely November offering. The 318-page document reveals incredible growth, a diverse group of customers, and nascent profitability after some accounting adjustments. With some creative extrapolation and a healthy dose of optimism, it's possible to imagine a potential $2 trillion valuation.

It's the extrapolating that is the hard part, though.

Anthropic will have to balance and spin several plates simultaneously to keep growing revenue at a rate than can compensate for its eye-watering costs. In an industry as young as AI, and with competitors closing in from all sides, investors won't know for sure whether Anthropic can deliver that at the time of its IPO.

"There's not a whole lot of operating history here," Jordan Chalfin, head of technology at CreditSights, told Barron's this week.

But Anthropic does have a brief history of making venture-capital firms very rich. What started as a group of seven computer nerds, ditching their jobs at OpenAI to go out on their own, has morphed into one of the fastest-growing companies in the world -- all in less than six years.

The AI lab chalked up $16.3 billion in revenue in the first half of 2026, an incredible leap from the $1.2 billion in the same period in 2025. Just as important, it turned an adjusted operating profit of $1.6 billion in that period, an enormous improvement on its $1.8 billion loss a year ago. The adjustments strip out stock-based compensation, noncash charitable contributions, and a litigation charge.

Anthropic's implied gross margin -- revenue less the cost of servicing customers -- was 85% in the first half of the year, the kind of number traditionally associated with mature, thriving software companies. And it isn't reliant on just a few customers, either: Anthropic has almost 12,000 customers, each spending $100,000 on a trailing 12-month basis, up from 200 at the start of 2025.

"The business model is profitable and generating cash," Ben Reitzes, who leads technology coverage at Melius Research, wrote in a note Monday. "There wasn't a new company like this in your dad's bubble of the 1990s."

At the same time, Anthropic spent $9.7 billion on research and development in the first six months of 2026, including the costs associated with training AI models. The S-1 also disclosed as much as $800 billion in data-center and chip commitments over the next decade, so bringing costs down in the medium term isn't really an option.

"The principal risk is that revenue falls short while fixed obligations remain in place," Reitzes said. "A slowdown would expose the company and its financing partners to substantial contractual and credit risk."

Certain aspects of Anthropic's meteoric rise are definitely unsustainable. Quarter-over-quarter revenue growth, for instance, slowed to 45% in the third quarter from a triple-digit growth rate the period before, according to a person familiar with the matter. Anthropic will reveal its final third-quarter and October revenue data before its IPO.

Other aspects are likely unsustainable -- and may test the risk appetites of even the most adventurous of investors. That 85% gross profit margin bodes well for Anthropic's inference business, where it provides existing AI models to customers. But it also doesn't leave much room for improvement after the IPO.

Anthropic also faces a range of competitive challenges that will make it hard to sustain supercharged growth.

Open-weight models remain a significant threat. Companies can run these customizable models on their own hardware or external clouds, often at a much lower price per token -- the standard unit of measurement for an AI query -- than Anthropic's Claude or related products. These models may not be as advanced as so-called frontier models, but for many users, they are sufficient, and they are especially helpful where maintaining data privacy is important.

Anthropic acknowledged in its S-1 that such models "may capture a portion of the market for certain use cases, which could exert pricing pressure on our offerings."

Jay Goldberg, an analyst at Seaport Research Partners, called open-weight models Anthropic's most serious outside threat, in a recent research note: "These models set the floor on pricing, basically a few cents above the cost of electricity."

Pressure from the open-weight class also magnifies the substantial threat within Anthropic's control: What if the company deliberately slows down model development? CEO Dario Amodei has continually warned about the potential harms of AI, sometimes to the chagrin of more gung-ho leaders like Nvidia CEO Jensen Huang.

A letter from Amodei and his six co-founders at the top of the S-1 outlined fears of rogue actors using AI to launch cyberattacks or design biological weapons; governments using the technology to "deepen repression" and bolster surveillance states; and AI agents acting autonomously in ways humans don't intend, as some already have.

As a public-benefit corporation, or PBC, under Delaware law, Anthropic acknowledges it could make decisions that put public safety above profit maximization. Whether investors believe Amodei's warnings, a PBC of this size remains an untested experiment.

Given all the question marks hanging over Anthropic -- from its legal structure to its revenue durability -- patience may be the best play for investors. As Barron's argued before the SpaceX IPO in June, the largest IPOs tend to disappoint investors, not least because early private investors are eager to sell shares and lock in gains. (SpaceX stock is up from its $135 offering price, but has fallen from its first-day close of $160.95.)

Anthropic's S-1 could've spooked investors completely. That hasn't happened, and we contend it remains worth watching. If you think the company sustain its rapid ascent, by all means invest.

Our take: Waiting a few more quarters to see how the rapidly evolving AI industry takes shape may be the more prudent move.