What to build in 2026: capital moved to vertical AI, and Airtable sold 89% below its peak
Capital has moved to vertical software that sells outcomes, and horizontal seat-based software is clearing its exits at a fraction of their 2021 marks.

Table of contents
Per-seat licensing is at odds with the buyers it now serves. A customer who uses AI to cut headcount removes seats, so the vendor loses revenue at the moment the customer succeeds. Buyers are pushing back with price cuts, vendors are moving to prices that do not depend on headcount, and the venture record shows where the capital went next.
The 2026 software selloff was read as a verdict on software. The evidence points at a narrower target: the way software is priced, and the headcount behind it.
For a founder choosing what to build in 2026, that distinction decides whether you are building in the part of the market the capital is funding or the part it has already repriced.
AI took 83% of the venture dollars in 2026 and 43% of the deals
AI took 82.7% of US venture deal value by September 2026, up from 36.7% in 2023. Its share of the number of deals has moved far less, from 21.8% to 43.2%. The gap between the two lines is the market: a handful of outsized AI deals, and a long tail of ordinary companies funded at ordinary prices.
The first quarter of 2026 was the largest in venture history, at $300 billion, and four companies took $188 billion of it, or 65%. Early-stage funding was up 41% year over year, so concentration is not the whole picture: the market funded more companies this year and put fewer dollars into most of them.
A per-seat price charges for the thing your buyer is trying to remove.
Per-seat pricing and the buyer it is losing
82% of SaaS CEOs say customers have asked for AI-related price cuts. 85% see AI as a direct threat to their own business model. Those CEOs expect agents to automate 41% of core workflows within two years, and they expect 35% of revenue to come from consumption or value-based pricing.
Carrie Osman, who runs Cruxy, put the mechanism this way: "Seat-based pricing was built for a world where humans did the work. AI doesn't log in and doesn't require a user license."
The vendors are leaving a pricing model that still fits their product. 97% of the 300 CEOs in that survey plan to retire seat-based pricing within two years, while 94% say seat-based pricing still matches the value their product delivers today. Buyers have moved first: fewer than one in five enterprise software decision makers prefer per-user pricing, against 43% who prefer consumption and 27% who prefer outcomes.
Houlihan Lokey's Q1 2026 update on AI in vertical software puts the same split in a table: horizontal software is "often bundled or priced incrementally per seat", while vertical software's monetisation is "evolving to premium modules, usage-based pricing, or outcome-based fees".
The reason a vendor gives up a model that still fits is in the first number. A customer who automates work removes the seats that paid for the software. The vendor's revenue falls when the customer succeeds, so the customer asks for a discount, and the vendor has to change the model or lose the account.
Where the capital went
Vertical applications overtook horizontal platforms for the first time in Q3 2026. Vertical startups took 42% of the software and AI venture capital in Q4 2025, up from 19% in Q1 2025.
Sequoia partner Julien Bek's March 2026 essay argues that the next trillion-dollar company will be a software company masquerading as a services firm, on the ratio that for every dollar spent on software, six go to services. That is the same split the funding shows, seen from the other side: the money is moving from the tool budget to the services budget.
The exits have already repriced the difference. Bending Spoons bought Airtable for $1.285 billion of enterprise value on roughly $480 million of revenue, about 2.7 times revenue, against the $11.7 billion valuation the company carried in 2021. Airtable's exit landed 89% below its peak. Miro agreed a price 92% below its Series C value. Private equity capital deployed into new software fell 65.7% year over year in Q2 2026.
A new layer is pricing per task. Startups now sell to agents themselves: phone numbers, payment credentials, liability insurance, permissions. The person holding the seat is no longer the only customer.
The objection is that private buyers did not follow the public market down. Private equity accelerated take-privates of OneStream, Dayforce and Clearwater at premiums. The names that led the recovery were identity, security and vertical software, while AI-exposed horizontal SaaS lagged by roughly 65 points. Jefferies' Brent Thill argues the disruption fears are overblown and that depressed valuations set up a software M&A boom. That is the mechanism this thesis predicts: private equity buying horizontal software at 2.7 times revenue.
Where the seat survives
Per-seat pricing is not disappearing everywhere, and the survey evidence says where it survives. Futurum's 2H 2026 survey has per-seat pricing finishing last for core software at 12.6%, behind consumption at 28.9% and outcomes at 22.2%, and finishing first for AI software at 42.3%.
The dividing line is simple. Software sold to the people doing the AI work keeps its seats, because those people are the buyers and they are not going away. Where the software is sold to the work itself, agents do the work and the seats disappear.
For a founder, that is a question about who your user is. A per-seat price is exposed when your users are the teams you are helping to automate, and it is safe when they are the people building the automation.
Three questions before you write the pricing page
- Is your product vertical or horizontal, and which side of that line is the capital on? Vertical applications took 42% of software and AI capital in Q4 2025 and overtook horizontal platforms in Q3 2026.
- Does your price rise when your customer hires, and fall when your customer automates? A per-seat price falls when the customer succeeds, and 82% of SaaS CEOs say customers already ask for AI-related price cuts.
- What is your exit worth if you build horizontal seat-based software? Airtable cleared at 89% below its peak and about 2.7 times revenue, and Miro agreed 92% below its Series C value.
Price the outcome, and the buyer stops counting your seats.
The line to hold
The line to hold is about the pricing model, and the technology choice is a separate decision. A per-seat price ties your revenue to the headcount your customer is trying to remove. Price the outcome, sell into a workflow that is specific to one industry, and the venture record gives you a market that is funding exactly that shape.
Working out what to build next is a numbers problem before it is a design problem. If you want to run it with someone, get in touch.
About the data
Four sources carry the numbers here. PitchBook's Q3 2026 AI Report and its Q3 2026 US VC Valuations Report supply the funding shares and the exit multiples. Euclid Ventures' Vertical Report 2026 supplies the quarterly split between vertical and horizontal software, covering 4,395 US and Canadian software and AI financings worth $186 billion in 2025. Carta's PE and SaaS work supplies the pullback in new software deployment, and L40° supplies the private-market recovery data and the Jefferies counter-argument.
Two surveys supply the pricing evidence: Futurum's 1H and 2H 2026 Enterprise Software Decision Makers surveys, and Cruxy's April 2026 survey of 300 B2B SaaS CEOs, reported by CFOtech.
Two 2026 sources supply the market context. Crunchbase's Q1 2026 record-quarter figures come from its coverage of the quarter. Houlihan Lokey's Q1 2026 vertical software update supplies the monetisation split between horizontal and vertical software.
Two caveats. The funding data is US, and while Australian founders compete for the same global capital and the same exits, the totals do not describe Australian capital. And the exits rest on two companies, Airtable and Miro, and two exits are not a market: if the fourth quarter of 2026 shows horizontal multiples recovering toward their 2021 marks, this argument weakens.
Sources
- PitchBook, Q3 2026 AI Report
- PitchBook, Q3 2026 US VC Valuations Report
- Euclid Ventures, The Vertical Report 2026
- Bending Spoons, Airtable acquisition announcement, SEC Form 8-K
- L40°, Who is buying SaaS companies in 2026?
- Futurum Group, outcome-based and hybrid AI pricing models
- Futurum Group, the pricing split in core software and AI
- CFOtech, SaaS chiefs expect pricing shift as AI threatens sales
- Fortune, this Sequoia partner thinks AI-enabled services are the new software
- Crunchbase News, what the record venture funding quarter means for your startup's fundraise