Venture datasets now support opposite stories about the same quarter.1 Scenarios let us trace where the signals from Part 1 could lead under two open questions. Will capital keep concentrating in a few companies? Will AI agents take over the decision to invest?
Key points
- Capital concentration and AI decision-making are the two uncertainties most likely to shape venture's next decade.
- AI drew 82.7% of US deal value through Q3 2026, and rounds of $100 million or more took 87.5% in the first half.23
- Cheaper company formation pulls against concentration as solo founders start 36.3% of new US startups.4
- The share of private investment firms using AI for investment decisions doubled to 28% in a year.5
- Each of the four futures has evidence behind it today, and they differ in who holds capital and who decides.
Two uncertainties
Concentrated capital or dispersed capital
Dollars are piling into a few companies. AI took 82.7% of US deal value in the first nine months of 2026.2 Rounds of $100 million or more took 87.5% of US value in the first half.3 Q3 2026 set a record for billion-dollar rounds.6 The forces behind dispersal are also strong. Solo founders started 36.3% of new startups in H1 2025, up from 23.7% in 2019.4 Series A software teams shrank from 22 to 15 people between 2022 and 2024.7 One in five Atlas startups charged a first customer within 30 days in 2025.8 Cheaper companies need smaller checks and could spread capital across many more bets. The data cannot settle this yet. PitchBook counted its second busiest US deal quarter in Q3 2026. Crunchbase coverage showed North American deal counts at their lowest since 2023.1
Human partners or AI agents
The second question is who decides. The share of private investors using AI for investment decisions doubled to 28% in a year.5 DVC replaced its analysts with AI agents in a $75 million fund.9 Its 170 limited partners source deals in exchange for a share of carried interest.9 Delaware's proposed Artificial Intelligence Company could give AI-run funds and firms legal standing. Courts, regulators, and limited partners have not yet decided whether machine judgment can carry fiduciary duty.
Four futures
The Kingmakers' Club
In 2036 a handful of mega-firms raise most LP money and write most of the checks. Human partners still decide, and their names still move markets. Capital flows to a short list of AI labs and the platforms around them. Seed checks keep growing from today's $3 million median, and fewer companies get one.10 High entry prices force each company to show more revenue before its Series A.11 The winners are large firms, frontier labs, and founders with elite networks. The losers are first-time fund managers, non-AI sectors, and founders outside the network. Today's signals point here. Rounds of $100 million or more already take 87.5% of US value.3
The Allocator Machines
In 2036 the largest firms run agentic platforms that source, screen, and price deals across the whole market. A few model-driven allocators manage most venture capital. Fees fall as agents replace analysts and associates.9 AI-run funds hold capital as legal entities under wrappers like Delaware's AIC. Shared models herd money toward the same companies, and megarounds grow larger. Founders pitch to software before they meet a person. The winners are platform owners and LPs who get cheaper access. The losers are mid-sized firms and junior investors whose training path disappears. Human founders also compete with AI-run companies the allocators favor for speed and data. Early signs include AI-assisted decisions doubling to 28%5 and record billion-dollar rounds.6
The Workshop Economy
In 2036 a company costs little to start and less to run. Solo founders with AI tools reach revenue in weeks and stay small.8 Many never raise a priced round. Thousands of small funds, angels, and syndicates write checks of $250,000 to $1 million into revenue-backed companies. Seed inflation reverses as check sizes follow lower capital needs. Products are easy to copy, and investors pay more for judgment of the founder. The winners are founders who own most of their company and small funds that know a niche. The losers are mega-firms built for billion-dollar deployment and the old ladder from pre-seed to Series B. Today's signals point here too. Solo founders start over a third of new startups,4 and Series A teams shrank by a third.7
The Agent Swarm
In 2036 agent-run funds write thousands of small checks into agent-run companies. An AI company can incorporate, launch, and sell without a human founder. Formation soars, and most new entities never hire anyone. Launches flood every marketplace. Platform gatekeepers decide who gets seen. Funds pay human networks in carry for deal flow and judgment the models lack.9 The human founder becomes an owner of agents and the person who carries legal liability. The winners are platform owners, fast operators, and networks that verify which companies are real. The losers are founders of labor-heavy services and investors who screen by hand. Apple received 557,000 new app submissions in 2025, the most since 2016.12 US business applications set a record near 5.67 million in 2025. High-propensity filings stayed flat.13
What holds in every future
- Founders who reach revenue with a small team keep their options open whichever way capital flows.
- Investors who can judge a founder's distribution edge hold value in any market where products are cheap to copy.
- LPs should check managers against several datasets because one quarter's global totals differ by about $28 billion across sources.1
- Firms should train junior investors on purpose because agents remove the jobs that once trained them.9
Signposts
- Rounds of $100 million or more falling below 70% of US deal value for a full year would favor the two dispersed futures.3
- The median US seed round rising past $5 million would favor the Kingmakers' Club.10
- AI-assisted investment decisions spreading to half of surveyed firms would favor the Allocator Machines.5
- Solo-led companies winning well above 14.7% of priced-round cash would favor the Workshop Economy.4
- Delaware enacting the AIC and a major LP backing an AI-run fund would favor the two AI futures.
Sources
- Venture datasets disagree on Q3 2026 totals and deal count direction · grey-lit ↩
- AI takes 82.7% of US venture deal value through Q3 2026 · journalism ↩
- Rounds of $100M or more take 87.5% of US venture value in H1 2026 · journalism ↩
- Solo founders start 36.3% of new US startups, up from 23.7% in 2019 · grey-lit ↩
- Share of private investors using AI for investment decisions doubles to 28% · grey-lit ↩
- Crunchbase counts a record number of $1B-plus rounds in Q3 2026 · journalism ↩
- Average Series A software team shrinks from 22 to 15 people · expert ↩
- Stripe Atlas startups reach first revenue faster, 20% within 30 days · disclosure ↩
- DVC replaces analysts with AI agents and carry-sharing LPs in $75M fund · journalism ↩
- US median seed round triples since 2018 to about $3M · journalism ↩
- Carta median seed pre-money reaches $16M, up 14% in a year · journalism ↩
- App Store new submissions hit 557,000 in 2025, the most since 2016 · journalism ↩
- US business applications set a record while serious filings stay flat · institutional ↩