When Building Is Easy, Distribution Is the Moat: Who Sells the AI-Software Flood
Special report, July 19, 2026. Brought to you by SignalDeck.live.
The wrinkle
Yesterday’s piece, The Cheap-Model Winners, mapped who profits when AI models get cheap and everyone can self-host: the clouds, the data and inference infrastructure, the app-software franchises, and the silicon underneath. This is the follow-up it promised, because one angle turned out big enough to stand on its own.
Here is the wrinkle: AI is collapsing the cost of building software toward zero, but the cost of acquiring a paying human has not fallen a cent. The marginal cost of building software is going to zero; the marginal cost of a paying customer is not. When any two-person team can prompt an app into existence over a weekend, the app itself stops being the moat. What’s scarce is what it always was underneath: attention, distribution, and the machinery that converts a stranger into a paying customer. A flood of AI-built products is coming, and every one of them will need to be seen, trusted, and sold.
Yesterday’s report covered the big ad-buying rails: AppLovin’s AI ad engine, Meta and Google’s distribution monopolies, Shopify’s agentic commerce push. This piece goes one layer down, to the dedicated go-to-market and marketing-software layer that the coming product flood will run on. And the sharpest tell that this layer matters? The incumbents are buying it. Adobe acquired Semrush for roughly $1.9 billion (closed April 2026). Publicis agreed to acquire LiveRamp for roughly $2.5 billion in equity value (May 2026). When the giants pay premiums for marketing visibility and identity data, they’re telling you which asset they think is scarce.
The scoreboard
Sorted by latest revenue growth, fastest first.
RDDT, community platform: ads + AI data licensing. Q1 2026 rev $663M, +69% YoY. FCF $311M (~47% margin). 126.8M daily active uniques, +17%. Standalone.
ZETA, AI-native marketing cloud. Q1 2026 rev $396.3M, +50% YoY. FCF $41.7M (10.5% margin). 189 super-scaled customers at $1.7M ARPU. Standalone; GAAP-unprofitable.
BRZE, enterprise customer-engagement messaging. Q1 FY27 (Apr-end) rev $211M, +30% YoY. Record FCF $27M (~13% margin). 2,713 customers, 349 at $500K+ ARR. Standalone; GAAP-unprofitable.
KVYO, e-commerce marketing automation (email/SMS). Q1 2026 rev $358M, +28% YoY. TTM FCF $212M. 196,000+ customers, Shopify-native. Standalone.
HUBS, all-in-one SMB marketing/sales/CRM. Q1 2026 rev $881M, +23% YoY. FCF $154M (17% margin). 299,458 customers, 135+ countries. Standalone.
TWLO, communications APIs + Segment customer-data platform. Q1 2026 rev $1.407B, +20% reported / +16% organic. FCF $132M (~9% margin). 402,000+ accounts, 10M+ developers. Standalone.
AMPL, product/growth analytics. Q1 2026 rev $93.5M, +17%. FCF −$13.2M (−14% margin). ARR $374M, +17%. Standalone.
TTD, independent ad-buying platform (open internet + streaming TV). Q1 2026 rev $689M, +12% YoY. FCF $276M (~40% margin). 95% client retention, ~$10-12M avg client budget. Standalone.
RAMP, data clean rooms / identity. FY2026 (Mar-end) rev $813M, +9%. 20% TTM FCF margin. Being acquired by Publicis, ~$2.5B.
Ad & attention rails
The Trade Desk (TTD) is the biggest independent demand-side platform, software for buying digital ads programmatically across the open internet and connected TV (streaming-TV ads), and it is the most contested story in this report. The fundamentals still read like a franchise: Q1 2026 revenue of $689 million, up 12%, free cash flow of $276 million, 95% client retention, roughly $1.4 billion in cash, and its Kokai AI buying platform at 85% client adoption. But growth has decelerated hard, +26% in FY24, +18% in FY25, +12% now (with Q2 guided to roughly +8%), and the stock is down roughly three-quarters year-to-date on a stack of real problems: Amazon’s DSP taking share with proprietary retail data, a fee dispute with Publicis, and two CFO departures. The narrative says TTD is an AI beneficiary; the market is pricing it as if AI-armed walled gardens are a threat to TTD specifically. Both can’t be right, and we’re not handicapping which, we’re flagging that the disagreement itself is the story.
Reddit (RDDT) is the cleanest expression of this thesis, because it monetizes the one thing AI cannot manufacture: authentic human opinion at scale. Q1 2026 revenue was $663 million, up 69%, ads alone $625 million, up 74%, with GAAP net income of $204 million (a 31% margin), $311 million in free cash flow, and $2.77 billion in cash. It gets paid twice: advertisers (up 75% year-over-year) buy access to 126.8 million daily uniques, while AI labs pay for the data itself, roughly $60 million a year from Google and $70 million from OpenAI, now being renegotiated toward usage-based terms. Reddit is the second most-visible domain in Google and appears in roughly 21% of Google AI Overviews (a third-party estimate); AI answers cite Reddit rather than replace it. The risks are concentration (94% of revenue is ads) and deceleration (guidance implies ~+44%), but this is the rare name where the AI flood makes the underlying asset scarcer, not cheaper.
Go-to-market software
HubSpot (HUBS) is the default go-to-market operating system for the small and mid-sized businesses that AI-built products will mostly be: Q1 2026 revenue of $881 million, up 23%, $3.45 billion in ARR, 299,458 customers across 135+ countries, and $154 million in free cash flow (17% margin). Its most important move isn’t a product, it’s a pricing decision: in April 2026 it shifted its Breeze Customer and Prospecting AI agents to outcome-based pricing ($0.50 per resolved conversation, $1 per qualified lead), you pay for results, not seats. That’s a direct answer to the seat-compression threat hanging over all software. Net revenue retention of 103% is the soft spot to watch. The bigger structural bet is platformization: as SMBs consolidate marketing, sales, and service onto one platform (the same ‘buy one vendor, not ten’ force reshaping cyber), HubSpot is the clear all-in-one SMB winner, consistently rated easier to use and better-liked than Salesforce down-market. Salesforce’s counter is its own moat, though: it owns Slack, both a network-effects asset and the natural agentic surface where its AI agents live and act, a conversational layer HubSpot lacks. Net it out: the SMB race tilts to HubSpot on simplicity and one-roof consolidation; the enterprise race tilts to Salesforce on the Slack network and its agentic surface.
Klaviyo (KVYO) is the marketing-automation layer (email and SMS) for e-commerce and direct-to-consumer brands, and it sits natively on Shopify, meaning it’s plumbed directly into the agentic-commerce rails yesterday’s report covered. Q1 2026 revenue was $358 million, up 28%, with 110% net revenue retention, a record 16.4% non-GAAP operating margin, its first GAAP profit, and $212 million in trailing free cash flow across 196,000+ customers. A $500 million buyback signals management thinks the market is underpricing that combination. If AI-built storefronts proliferate on Shopify, Klaviyo is how they’ll talk to customers.
Braze (BRZE) does cross-channel customer engagement for the enterprise: Q1 FY27 (ended April 30) revenue of $211 million, up 30%, its fourth straight quarter of accelerating growth, with 110% dollar-based net retention and 349 customers paying $500K+ a year. The honest caveat: it’s still GAAP-unprofitable, with a $27.5 million operating loss, though narrowing, alongside record free cash flow of $27 million. Its Project Catalyst beta, agentic generation of message variants, points at the interesting future: marketing software that does the work, not dashboards that report on it.
Delivery, data & AI-visibility
Twilio (TWLO) is the last mile: communications APIs (send SMS, voice, or WhatsApp messages from any app) plus Segment, a customer data platform that unifies a company’s first-party data. Q1 2026 revenue was $1.407 billion, up 20% reported and 16% organic, its fastest organic growth since 2022, with a 19.8% non-GAAP operating margin, $132 million in free cash flow, 402,000+ accounts, and 10 million+ developers. Its Agent Connect layer wires AI agents directly into voice and messaging: when AI-built products need to actually reach a human, this is the pipe. The risk is a commoditizing SMS/voice core and structurally thin margins.
Zeta Global (ZETA) is the highest-beta name here: an AI-native marketing cloud whose Q1 2026 revenue of $396.3 million grew 50% year-over-year, accelerating, its 19th consecutive beat-and-raise, tied to the launch of Athena, AI agents that execute campaigns end to end rather than assist a human operator. Adjusted EBITDA was $66.1 million (16.7% margin) with $41.7 million in free cash flow, and 189 super-scaled customers at $1.7 million average revenue each, up 21%. It’s still GAAP-unprofitable (a $13.2 million net loss, improving), so it carries the most execution risk alongside the most direct exposure to the thesis.
Semrush (SEMR, now an Adobe unit) effectively invented AEO, “answer engine optimization,” the practice of tracking and improving whether your brand gets cited inside ChatGPT, Google AI Overviews, and Perplexity answers, a category that didn’t exist three years ago. Its AI ARR went from $4 million to $38 million in 2025, $50K+/year customers grew 74%, and FY2025 revenue was $443.6 million, up 18%. Adobe paid roughly $1.9 billion, a 74%+ premium, to own that category, which closed the pure-play trade but validated the thesis louder than any earnings call could.
Two brief notes. LiveRamp (RAMP), data clean rooms, which let companies match first-party customer data without exposing raw personal information, grew FY2026 revenue 9% to $813 million with 20% free-cash-flow margins, and is being acquired by Publicis for $38.50 a share (~$2.5 billion), which framed the deal as “data co-creation for smarter agents.” Amplitude (AMPL), product and growth analytics, is the cautionary contrast: revenue of $93.5 million grew 17%, but free cash flow is negative $13.2 million and worsening because AI-inference costs are a drag. It’s the one name here where AI is currently a cost center, not a growth or pricing lever.
Most innovative in this layer
Reddit, for monetizing scarce authentic human signal twice, ads and AI data-licensing, while AI answers cite it rather than cannibalize it. Semrush, for inventing AEO and getting acquired at a 74% premium for it. Zeta and Braze, for pushing marketing from dashboards humans read to agents that do the marketing. HubSpot, for the bravest pricing move in software, charging for AI-agent outcomes, not seats. And The Trade Desk’s UID2, an open-source identity standard for the open internet, genuinely innovative, though as noted, the market currently disputes whether TTD captures the value of its own invention.
The counter-risk
Be clear-eyed: martech is not a free ride on this thesis. First, the seat-compression risk needs the same precision as yesterday’s piece: what bites is per-tool seats and the pricing model, not total headcount, and much of this layer isn’t even seat-priced. Braze bills mostly on monthly active users, Klaviyo on profiles and messages, Zeta on consumption, and HubSpot’s per-seat exposure sits in its Sales and Service hubs while Marketing Hub is contact-tiered. That makes martech structurally less seat-exposed than horizontal SaaS, and HubSpot’s outcome-pricing pivot is the hedge, not yet the proof. Second, several names are GAAP-unprofitable (Braze, Zeta), and The Trade Desk is decelerating hard, down roughly three-quarters year-to-date on Amazon DSP share loss, a Publicis fee dispute, and CFO churn, a reminder that “rails” can get disintermediated by bigger rails. Third, consolidation is eating the best pure-plays: Semrush is gone into Adobe, LiveRamp is going into Publicis, and each deal shrinks the investable surface even as it validates the category. Fourth, and sharpest: the platforms’ own AI could absorb this layer. Meta’s Advantage+ and Google’s Performance Max already automate targeting, creative, and budgets inside the walled gardens, and assistant-native commerce points the same way; if the platform’s AI does the marketing, the dedicated martech layer risks shrinking to plumbing. The tailwind names, Reddit, Klaviyo, Zeta, Twilio, have the flood at their backs; The Trade Desk and Amplitude have real questions to answer first, and no one here is immune to the agent that does the job the software was sold to assist.
Bottom line
When everyone can build, nobody can be found. The scarce resource in an AI-software flood is not code, it’s a paying customer’s attention, and the companies that own the machinery of being found own the choke point. The incumbents’ checkbooks already agree, Adobe bought the AI-visibility category, Publicis bought the identity layer. Watch the ones with the flood at their backs (Reddit, Klaviyo, Zeta, Twilio) and mind the ones still answering hard questions (The Trade Desk, Amplitude), and remember the whole layer sells picks and shovels against a seat-compression wave that can wash over it too.
📊 Read this on SignalDeck.live.
Not financial advice. For informational and educational purposes only. Do your own research.