The same few names have dominated most of the AI discourse in 2025 and 2026. The hardware narrative is dominated by Nvidia. The majority of enterprise AI news is absorbed by Google and Microsoft. There is a lot of attention paid to the infrastructure buildout, which includes hyperscalers announcing capital expenditure figures that would have seemed unreal five years ago, data centers humming in the Virginia suburbs, and submarine cables being laid at record speed. The smaller businesses that sit just one layer below that infrastructure, creating products that real businesses use on a daily basis, growing their revenue at remarkable rates, and trading at valuations that don’t fully reflect what might be coming, are often overlooked.
Analysts who focus more on numbers and less on narrative have shown a renewed interest in three names in particular: UiPath, GitLab, and SentinelOne. Outside of enterprise technology circles, none of them are well-known. Each of the three has experienced difficult periods in the market. And because of what their revenue trajectories are subtly telling anyone who is willing to pay attention, all three are now priced in a way that makes the potential upside genuinely intriguing, not because of hype.
Of the three, UiPath is the oldest and, in some respects, the most misinterpreted. Robotic process automation, the somewhat unglamorous industry of employing software bots to perform repetitive tasks like data entry or invoice processing, is how the company established its reputation. For a while, that seemed like outdated technology that would be rendered obsolete by generative AI. The outcome of that argument has not been what critics had anticipated. Rather, UiPath has discovered that it possesses a truly valuable asset: years of expertise overseeing automated processes in intricate corporate settings, which turns out to be precisely the basis required to coordinate AI agents.
Managing multiple agents from various vendors, such as Salesforce agents, Microsoft Copilot integrations, and custom-built tools, is becoming increasingly difficult for businesses. For that orchestration problem, UiPath has developed a platform. The stock is priced as if the AI agent cycle hasn’t arrived yet, trading at a forward price-to-sales multiple of about 5, with revenue just beginning to show signs of re-acceleration. That gap might close more quickly than the market currently anticipates.
The situation at GitLab is different because the market has actively penalized the company for a crime it did not commit. The argument against it is simple: if AI is capable of writing code, fewer human developers will require GitLab’s platform, and the company will become smaller. The bears have not been persuaded by more than two years of annual revenue growth of 25 to 35 percent. Despite having gross margins of 87 percent, which most software companies would consider exceptional, GitLab’s forward price-to-sales multiple has been compressed below 5.5 due to this stubbornness.
The bears don’t seem to understand that more AI-generated code doesn’t always translate into less need for the supporting tools. As development teams expand rather than contract, GitLab is gaining new users, observing current clients upgrade to higher-tier subscriptions, and witnessing seat expansion. Additionally, the company has introduced its own AI agents that support not only the code-writing phase but the entire software development lifecycle. As AI becomes more ingrained in software development, there is a plausible scenario in which GitLab’s revenue growth actually accelerates. If that occurs, it would be extremely difficult to defend the current valuation in retrospect.
In a way, SentinelOne is the easiest of the three. This AI-powered cybersecurity company is expanding its revenue more quickly than the majority of its larger competitors. Its forward price-to-sales multiple is less than 4.5, which is significantly less than what similar companies like CrowdStrike command. The discount has continued in part because SentinelOne isn’t yet profitable in the conventional sense and in part because AI anxiety, or the worry that current security products aren’t strong enough to fend off AI-enabled attacks, has put pressure on the cybersecurity industry as a whole.
Although SentinelOne’s recent actions indicate that it is taking the issue seriously rather than ignoring it, the concern is legitimate. As it fully ramps up, the Lenovo partnership, which pre-installs its Singularity Platform on the devices of the largest enterprise computer vendor in the world, may prove to be a significant distribution agreement. Prompt Security’s acquisition adds an AI data leakage protection capability that the majority of its rivals are still working hard to develop. Although the exact moment when operating leverage takes effect and the stock starts to reflect the underlying growth is still unknown, the circumstances are coming together.
One type of market curiosity is seeing these three trade at these valuations while their revenue curves continue to rise. It’s the kind of thing that eventually resolves, usually in one direction or another, and usually more quickly than the quiet period that preceded it suggested was feasible. Each scenario carries a real risk: if the overall economy declines more severely than current projections indicate, enterprise sales cycles may stall, competition may become more fierce, and the AI spending environment may change.
None of these are certain. However, these three names are about as close to under-the-radar as the industry currently offers for investors who think the next stage of AI adoption involves the messy, complex work of actually deploying and managing the technology inside real organizations—rather than just building the chips and the data centers. There is a revenue trajectory. There is the valuation room. Nobody can say for sure whether the parabolic moment will occur in 2026 or later. However, the setup appears more honest than most, at least based on the numbers.
