• Buyside diligence deepens on AI impacts
• IPO hopefuls aim to prove real adoption
• Similar scrutiny extends to late-stage private funding

IPO-bound companies are encountering a new kind of investor scrutiny as firms preparing to list increasingly find themselves pressed to show how artificial intelligence fits into their business models as a core test of credibility.

What began as an easy buzzword scattered through investor presentations has become a central element of due diligence. Across sectors from enterprise software to industrials, issuers are being asked to prove how they capture the benefits of AI while also accounting for the risks it introduces.

Market participants say investor curiosity on the topic, from early looks through to the roadshow, has grown far more granular just as the technology spills over every aspect of government, business, and daily life.

Rather than accepting broad claims, the market now interrogates management teams and their advisers about how AI interacts with revenue drivers, where it delivers efficiencies, and where it might erode margins. That level of diligence is already influencing valuations and peer comparisons before filings reach the SEC.

The shift mirrors the same no-nonsense approach playing out in the earnings season for listed AI names, whose usage metrics and capital intensity have been dissected instead of rewarding generic exposure or loose marketing language.

If a company cannot explain how it uses AI to improve efficiency or customer experience, said Steve Maletzky, managing director at William Blair, the market begins to question its longer-term durability and competitive moat. Maletzky believes AI cuts both ways: a credible strategy signals resilience, while the absence of one suggests obsolescence.

One ECM banker added that nearly every IPO this year has contained some AI component because investors now expect to understand how companies use the technology to defend their economics as much as to expand them.

The July IPO of Figma demonstrated how these expectations have reshaped investor behavior.

The San Francisco-based design-software firm priced at an equity valuation of about USD 19.3bn. Shares rose strongly on debut and have remained well above issue price, making it one of the most successful software listings of the year. In the run-up to the offering, the market debated whether new AI-native design tools could threaten Figma’s business model.

Emerging AI applications were seen as potential competitors able to automate functions that currently drive subscription revenue, this news service reported. Even so, investors were comfortable backing the listing and saw promise in the company’s projections. One buysider described the AI issue back then as a medium-term risk to monitor rather than a near-term concern, but found its projections convincing enough to go ahead.

The same investor appetite has benefited the picks-and-shovels side of the AI economy. When CoreWeave went public earlier in the year, its positioning as a critical infrastructure enabler rather than another AI application drew heavy demand. The cloud-compute provider was marketed as a way for funds to gain exposure to the hardware backbone powering AI models.

The offering’s strong reception showed how the market now differentiate between companies developing AI products and those building the systems that make them possible.

That distinction is also shaping the decisions of other high-profile infrastructure players. Cerebras Systems withdrew its S-1 filing after securing a substantial late-stage round, with CEO Andrew Feldman citing the need to revise its disclosures to reflect “the portrait of a company that is becoming unrecognizable from its first filing.” In this sense, business models in the AI hardware segment are evolving and issuers are recalibrating their equity stories quicker than ever before returning to the public market.

Beneath the enthusiasm lies a more complex question of valuation. Steven Halperin, managing director and head of equity public markets at Moelis, called this the “dot-com 2000 stage of AI,” a period of capital chasing exposure without clarity on which players will endure. Investors are willing to do the work, he said, but long-dated projections tied to AI adoption curves between 2028 and 2030 remain difficult to underwrite.

Many issuers are finding that presenting an AI story is no longer enough and that they must show how it converts into earnings. The same forensic mindset that shapes the performance of public AI stocks is filtering into IPO due diligence.

Regulators are moving in the same direction. Ran Ben-Tzur, a partner at Fenwick who has advised on several major listings this year including Figma and CoreWeave, said the SEC has made AI claims a standard focus of comment letters. When companies describe themselves as AI-enabled, staff now request details on whether models are proprietary or licensed, how they are built into products, and what evidence supports any claims of superiority. “They want transparency,” Ben-Tzur said. “The SEC wants issuers to clearly and accurately disclose their use of AI and the associated risks.”

AI references appear in nearly every S-1 filing, and the Commission continues to push for precision. From the SEC’s perspective, scrutiny is intensifying as companies make assertions about AI features or models, said PwC IPO services leader Mike Bellin. “From the SEC lens, there’s a lot of scrutiny and good questions as companies make claims in documents around certain AI features or models,” Bellin said. “For issuers, the challenge is to demonstrate opportunity without overpromising.”

Companies today are effectively expected to deliver two AI stories, Shari Mager, US capital markets readiness leader at KMPG, said. The first is how AI features in their product offering or suite of services, and the second is how they use it internally to operate more effectively and efficiently. “Those two stories resonate very strongly in the market,” she said, “but there has to be teeth behind them.” Mager added that investors are increasingly trying to separate genuine implementation from cosmetic references inserted to stay on trend. “That’s what we’re seeing,” she said. “Investors are weeding out what’s just commentary on AI to get it in there versus what’s real, and companies need to be focused on that to differentiate and show value.”

That insistence on authenticity is shaping how advisers prepare issuers. Kevin Friedmann, who leads the corporate, M&A, and securities Chicago team at Norton Rose Fulbright, said AI has lost strength as a differentiator, functioning now as a basic expectation rather than a badge of innovation. Valuation upside, he noted, is reserved for companies that can prove a defensible advantage such as AI infrastructure or drug-discovery algorithms.

Some issuers have responded by tying AI credentials directly to measurable outcomes, but found that this may still not be enough to offset other less convincing aspects of their stories. Navan displayed an 11-point gross-margin gain over two years linked to automation in approvals and expense data. This showed how AI integration can deliver financial improvement, but the stock’s early trading slide also reminded the market of its disclosures around decelerating growth and net losses.

Private investors are applying the same discipline as they deploy capital into late-stage companies touting AI potential. Funds now triangulate AI claims by speaking directly with customers to determine whether the technology drives measurable outcomes, said Legion VC founder and general partner Tim Hoag. “You talk to enough customers and they’ll tell you whether something is really AI or not,” he said.

His team also weighs revenue acceleration against valuation multiples to judge whether the implied growth premium is justified. “We’ve seen companies go from USD 10m to USD 50m in ARR within a year, which changes the calculus,” he said. “It’s not just about the label, it’s about what the data shows.”

Moelis’ Halperin views this process as a necessary filter. Just as the early internet era produced lasting winners only after a speculative shakeout, AI’s eventual beneficiaries will be those able to turn narrative into measurable economics.

“Investors still want exposure,” he said, “but they are only beginning to build the frameworks needed to value it properly.”

The market may be searching for more robust valuation models, yet it is unwilling to finance stories that lack them. At the same time, AI no longer guarantees investor enthusiasm, but a company without a credible strategy risks failing the test entirely.