Frequently asked questions

Which small caps survived the AI screen?

Two of eighteen: PubMatic (PUBM), a programmatic ad-tech infrastructure play, and Sprout Social (SPT), a social media management SaaS. Neither is a full five-pillar fit; both are near fits, earning their grades with debt-light balance sheets, positive and growing free cash flow, and leadership with a record of doing what it said. The full reports rate PUBM Positive at 76/100 and SPT Selective Positive at 68/100.

Why is the balance sheet the binding constraint in the AI era?

Because self-funding is survival: you need cash to retool the product while revenue wobbles without going back to the market. The exciting AI narratives, fuel cells, batteries, data-center power plays, all died on the iron-fortress and free-cash-flow filters. The market will forgive a growth stall in a cash-generative compounder; it will not forgive a cash burn with an AI slide deck.

Why does infrastructure beat application at small-cap size?

AI increases demand for underlying infrastructure while threatening the application layer's economics. PubMatic sells the infrastructure ad transactions run on; CEVA licenses the chip IP edge AI runs on. Small-cap application SaaS sits in the worst seat: too small to out-build the platforms, too exposed to the seat deflation AI brings. Own the toll road, not the toll booth's software.

We screened eighteen sub-$1 billion software and tech names through our five-pillar engine this week. The brief was simple: find small caps worth a deep dive, and pay special attention to how AI is treating them, as headwind or tailwind.

Two survived: PubMatic (PUBM), a programmatic ad-tech infrastructure play, and Sprout Social (SPT), a social media management SaaS. Neither is a full five-pillar fit. Both are near fits, and both earned their grades the hard way: debt-light balance sheets, positive and growing free cash flow, and leadership with a record of doing what it said.

But the interesting output of the exercise wasn't the two names. It was the pattern. Here is what two small caps teach about software in the AI era.

Why is AI a balance-sheet story first?

The binding constraint in the screen wasn't growth or moat. It was pillars four and five: the iron-fortress balance sheet and free cash flow. The exciting small-cap AI narratives, fuel cells, batteries, data-center power plays, all died on this filter. They were funding AI-era stories with dilution.

That is not a coincidence. In a technology transition, self-funding is survival. You need cash to retool the product while revenue wobbles, and you need to do it without going back to the market with your hand out. For small caps, fortress plus cash flow isn't just a pillar. It is the admission ticket to the AI era. The market will forgive a growth stall in a cash-generative compounder. It will not forgive a cash burn with an AI slide deck.

How does AI compress the moat pillar fastest?

AI lowers the cost of building software. That sentence should worry every small-cap software investor, because differentiation is the moat small caps live on, and AI is making features replicable at machine speed.

Big companies can outspend this compression. Small companies have to out-niche it. Both of our survivors graded only partial on moat, and for the same reason: their software features can be copied, so the moat has to come from somewhere the code can't reach. For PubMatic, that is distribution (integrations with Walmart Connect, PayPal, Amazon) and owned infrastructure that cuts cost per impression. For Sprout, it is two billion daily customer interactions feeding an AI tier the platforms don't have. The lesson: in the AI era, a small-cap moat built on features alone is a wasting asset. Moats built on distribution, data scale, and cost structure still compound.

Why does infrastructure beat application at this size?

The screen's best risk-adjusted AI exposure wasn't SaaS at all. It was pipes. PubMatic sells the infrastructure that ad transactions run on; CEVA, another near-fit from the screen, licenses the chip IP that edge AI runs on.

The logic is clean. AI increases demand for underlying infrastructure while threatening the application layer's economics. Small-cap application SaaS sits in the worst seat in the house: too small to out-build the platforms, too exposed to the seat deflation AI brings. If you want small-cap AI exposure, own the toll road, not the toll booth's software.

How are seat economics breaking?

This is the defining question for software investing right now. When one AI agent does the work of five seats, per-seat pricing breaks. Sprout's own filings warn that AI could "significantly reduce demand for our services," and they are right to say it: AI agents that write, schedule, and optimize posts natively inside the platforms eat the exact workflow Sprout charges per seat for.

But the same force is an ARPU driver at the enterprise end, where Sprout's Trellis AI tier turns its data scale into retention and higher per-seat pricing. Enterprise is now 61% of Sprout's subscription revenue and growing 20%. So the market is splitting SaaS into two buckets: tools whose job AI does natively (usually SMB point tools in a platform's blast radius) and systems where AI deepens workflow lock-in. Sprout sits on the fault line between those buckets, and its investment case is really one sentence: the enterprise engine must outrun the low-end decay.

For PubMatic, the AI question is different. It is not about demand, ad spend on the open web isn't shrinking because of AI. It is about market structure: if Google and The Trade Desk keep agentic ad budgets inside their walled gardens, open-web pipes get bypassed. Distribution fight, not demand fight.

Why is leadership candor the new diligence?

At this size you cannot diligence the technology the way you can diligence a hyperscaler's. So you diligence the operator's honesty about it. The "did they do what they said" test now has an AI clause: is management actually retooling the product, or just putting AI on the pitch deck?

PubMatic is shipping: AgenticOS with 80+ autonomous campaigns running, all five agency holding companies on board, cost per impression down 20% year over year. That is the former. Most small caps we looked at were the latter. When the technology is moving this fast, candor about what is real and what is slideware is the closest thing to a leading indicator you get.

The bottom line

The five-pillar engine did its job on these two names: it refused to call either a full fit precisely because the AI question is unresolved, and at this size you don't get paid to guess on unresolved questions. You get paid to watch the flips. For PubMatic: three consecutive quarters of double-digit growth with CTV and emerging streams above 65% of revenue. For Sprout: the enterprise mix outrunning low-end decay, with Trellis Plus driving measurable ARPU expansion.

Until then, the screen's verdict stands as the article's verdict: in small-cap software, AI rewards the balance sheet first, the infrastructure second, and the storytellers never.

Where is the detailed research?

Both deep dives are published in full on our research page: