The Solo-Founder Shift

Sources#

The State of Solo Founding (Solo Founders Report 2025)

Summary#

Carta's proprietary cap-table dataset on tens of thousands of U.S. companies, published as The State of Solo Founding (December 2025) by Solo Founders with Carta Insights. The measured claim: the share of new U.S. startups on Carta founded by a single person rose from 23.7% in 2019 to 36.3% in H1 2025.

Read the figures and the framing as two different sources. The numbers are Carta cap-table data — the same instrument class as the platform's other primaries. The publisher is an organization that sells a paid Solo Founders Program on the same page and whose thesis is "Today, solo founding is considered odd. Soon it will be the default." Interpretation throughout is attributed to the publisher; only the figures are carried here. The most useful consequence of reading it this way is that the dataset repeatedly declines to support the framing, which is where the interesting findings are.

The series, and two caveats the report does not make#

Share of U.S. startups on Carta founded by a solo founder (transcribed from the chart image during ingest — the prose gives only the endpoints):

2019	2020	2021	2022	2023	2024	H1 2025
23.7%	24.5%	26.0%	27.2%	27.8%	30.5%	36.3%

Two problems with the headline, neither raised in the source:

The final point is a half-year measured against six full years, and it is by far the steepest jump in the series — +5.8pp, against +2.7pp for the largest full-year step. The "over one-third of new companies" headline rests entirely on that non-comparable point. Every earlier year moves by 0.6–2.7pp; if H1 2025 behaved like any prior year it would read ~32%.
The denominator is companies on Carta. The series moves with Carta's own customer mix as well as with founder behavior, and the report does not address platform-composition drift. Solo founders are cheaper and likelier to adopt a free cap-table tier, which would inflate the trend without any change in the underlying population.

The rise before 2025 is nonetheless steady and predates ChatGPT (+3.5pp from 2019 to 2022, i.e. most of the pre-2024 movement happened before generative AI was available) — which is itself evidence against the report's AI-first explanation, and is not remarked on. Founder commentary attributes the shift squarely to AI (Daniel Francis: "It's definitely a testament to AI making things much easier"); the pre-2022 slope says falling company-formation costs were already doing the work. Peter Walker (Carta) gives the more defensible version: "This trend reflects technology lowering the cost of company creation."

The firm-scale twin of the solo-authorship rebound#

The structural parallel with The Solo-Authorship Rebound is close enough to be worth stating as its own finding, because the two sources share an instrument, a period, a proposed mechanism — and a weakness.

	The Solo-Authorship Rebound (Matsui)	This page (Carta)
Unit	a paper	a company
Instrument	left tail of the author-count distribution	left tail of the founder-count distribution
Finding	decades-long decline in solo authorship halts/reverses at end-2022	solo-founded share rises steadily 2019→H1 2025
Mechanism claimed	LLMs substitute for the coauthor's execution work	AI substitutes for the co-founder's execution work
Core weakness	~half the pooled break is venue composition in OpenAlex	trend confounded by platform composition on Carta
Evidence quality	interrupted time series, no untreated unit, heavily stress-tested	descriptive trend, no counterfactual, no robustness work at all

That last row is the asymmetry that matters. Matsui stress-tests his break against composition standardization, history conditioning, cutoff scans, donut specifications and a pandemic-rebound test, and still concedes the causal claim is not identified. Carta's series has none of that apparatus. So this is the weaker of the two measurements by a wide margin, and it is pointed at the more consequential claim. Cite it as a trend that exists, not as an AI effect.

The Solo-Authorship Rebound already recorded Zuckerberg's prediction that firm sizes shrink into "a larger number of companies with fewer people each" as "a prediction where this page has measurement." This page is the firm-scale measurement that comparison was missing — and it partially supports the prediction on company formation while saying nothing about company size, which is what the prediction is actually about.

Where the data contradict the framing#

Three places the dataset declines to support the publisher's thesis. These are the findings worth carrying.

1. Solo founders hire earlier — the organization of one is a waypoint, not a destination#

Median days from incorporation to first hire: 399 for solo founders, 480 for multi-founder companies. The report presents this as a solo-founder virtue ("solo founders tend to hire faster"), and it plainly is a difference. But note what it does to the surrounding narrative: the solo founder's first move is to stop being solo. A co-founded team starts with two people and adds a third at ~16 months; a solo founder starts with one and adds a second at ~13 months. Nobody is running a company of one for long.

This directly qualifies the vault's "one-person company" thread. AI-Native Organization's "organization of one" and Founder as Agent Orchestrator both describe a founder whose leverage comes from agents rather than headcount; Paul Klein IV, quoted in this very report, states the constraint that breaks it — "By default, a solo founder is single-threaded. You must become multi-threaded. You have to have someone building and selling the product while you're hiring." The binding limit is the founder's serial attention, not implementation capacity, and agents do not relieve it. Read against Implementation Abundance Inverts Product Work: implementation got cheap, and the first thing the freed-up founder buys is still a person.

One caveat cutting the other way: both groups are waiting longer to hire than they did a few years ago, which is the trend consistent with AI-substituted execution. The gap between solo and multi is old; the lengthening applies to everyone.

2. Solo founders do not spend their extra equity on talent — the measured null the publisher disputes#

For hires made in 2023–24, median equity grants to the first five employees are near-identical across solo- and multi-founder companies. This is the report's own stated surprise, and it is a real finding: solo founders hold roughly double the founder equity and do not convert any of it into larger early-employee grants.

"There's very little difference in equity grants, which runs counter to the theory that solo founders would be more generous with ownership. We expected equity to be a lever they'd pull to outcompete multi-founder companies for the best talent. Perhaps this is an underused strategy that solo founders could lean on, but largely aren't yet." — Peter Walker, Carta

The publisher then contradicts his own dataset in the same document, which is the sharpest internal disagreement in the source and is recorded here unresolved:

"The Carta data does not align with my experience with solo founders over the last 6 years or the Solo Founders Program in 2025. In both cases, the solo founders I've worked with often gave 2-5x the equity for early employees than the median numbers in the previous chart." — Julian Weisser, CEO of Solo Founders

Both claims can hold without conflict, and the reason is a selection problem the report does not name: Weisser's comparison set is founders who joined an organization that advocates exactly this practice, measured against a platform-wide median. His observation is evidence about his program, not about solo founders. Carta's median is the population fact.

The explanation the report offers for the null is more interesting than the dispute: solo founders "often encounter frameworks built for co-founder teams that have already split the company two or three ways," so they anchor on grant sizes calibrated for a cap table they don't have. Charles Hudson (Precursor Ventures) puts it precisely — "3% is only too much in the context of two founders." If that is right, the null is norm inertia, not preference: a measurable case of advice outliving the structure it was designed for.

3. No efficiency claim is available from this data#

The report's ownership findings are real and clean:

Dilution is near-identical across solo- and multi-founder companies at every early stage in 2024. Walker's reading: "when a solo founder can convince a VC they are a great investment, they don't pay a tax."
Round sizes are near-identical at Series A, slightly lower at Priced Seed and Series B — the seed gap read as a mild team-evaluation bias that fades once there is a business to evaluate.
By Series B, solo founders hold a roughly 50% larger personal stake; median ownership at exit is 75% greater than lead founders in multi-founder companies (2019–H1 2025).
Solo-founded companies took longer to exit in earlier cohorts but exit slightly faster recently.

What none of it is: a measure of output. Carta's dataset contains cap tables, not revenue. So this source characterizes the lean tail's structure — how it is owned, how it is funded, when it hires — and says nothing whatever about its efficiency.

That distinction is load-bearing for AI Investment Story, Not Efficiency Story, whose open question asks specifically about "the deliberately-lean solo-founder tail's RPE." This source does not answer it and cannot. It supplies the population base rate for the tail (the tail is much bigger than assumed — over a third of new companies) without supplying a single revenue-per-head figure for it. The report is careful about the adjacent version of the same trap, and the caveat deserves quoting because it is the most rigorous sentence in the document: "more ownership at exit only implies a better outcome when exit values are comparable, which this data does not capture."

Where the money goes#

Solo-led companies were 30% of startups founded in 2024 but received 14.7% of cash raised in priced equity rounds that year — a ratio the report treats as improving (Miura-Ko: from "around 10% in 2019 to the mid-teens more recently") and frames as investor attitudes thawing.

Walker names the more likely mechanic: the gap "is impacted less by the number of people starting companies as solo founders and more by the number of founders getting through VC filters to raise capital." And cash raised is explicitly a lagging indicator — newer solo cohorts have not reached the later stages where round sizes move aggregates. Both readings predict the ratio closes without any change in investor preference, purely from cohort maturation. Nothing here distinguishes thawing from arithmetic.

Two structural notes:

Solo-founded companies are less likely to raise a pre-seed (SAFE/note) in their first year, and reach their first priced round sooner — in every cohort from 2018–2024. The report suggests they skip the pre-seed stage rather than delay financing; Walker adds that a solo founder's lower burn buys more unfinanced tinkering time.
Sector mix is broadly similar, with solo over-represented in Consumer Products and Services (17.0% vs 12.0%) and under-represented in Pharmaceuticals and Biotech (5.3% vs 8.6%). The biotech gap is the one with an obvious mechanism — deep technical expertise usually has to be paired with commercial experience — and it is the firm-scale echo of the field ordering in The Solo-Authorship Rebound, where laboratory- and instrument-organized disciplines are exactly the ones that show no solo rebound. Two independent datasets, one on papers and one on companies, put wet-lab work in the same position: the execution that AI cannot absorb.

Connections#

The Solo-Authorship Rebound — the same left-tail instrument one unit down, and the better-identified of the pair: solo authorship halting its decline at end-2022, with the same AI-substitution mechanism, the same composition weakness, and far more robustness work. The biotech/wet-lab gap here matches its field ordering exactly
AI Investment Story, Not Efficiency Story — supplies the population base rate for the deliberately-lean tail that page's open question names, and pointedly not its RPE: Carta holds cap tables, not revenue
AI-Native Startup Lifecycle — the staffing dimension from the founding-team side; the lifecycle's headcount bands start at "seed → 10 engineers," and this measures what the team looks like before the first of them arrives
Founder as Agent Orchestrator — the qualification: the orchestrator-founder is real but short-lived on this data, hiring a first employee at a median 399 days. Agents relieve implementation, not the founder's serial attention
AI-Native Organization — the "organization of one" reframed as a waypoint: solo founders are the fastest, not the slowest, to add a second person
Implementation Abundance Inverts Product Work — the economics behind the trend and its limit: implementation got cheap, and the first thing the freed-up founder buys is still a person
Balance-of-Power Superintelligence — the firm-scale prediction this partially tests: Zuckerberg forecasts more companies with fewer people each. Company formation supports him; company size is untouched by this data
The Tragedy of the Cognitive Commons — the apprenticeship reading, weaker here than in the paper case: a co-founder slot that did not form is not obviously a training slot destroyed, since the solo founder hires sooner than the team does
AI and Market Power — founder characteristics measured on a much larger, non-platform population: across 382,108 VC-backed or patenting start-ups in the OECD Start-up Database, having a serial founder is the single largest founder effect in either OECD regression — +4.2 to +5.0pp on the probability of an acquisition exit (base rate 7.58%) and +0.93 to +1.88 log points on lifetime VC raised, against +0.54 for a PhD founder. Carta's series counts how many founders; that one prices which founders, and it does so on the exit and funding margins this page can only describe structurally
Emergent — the celebrated lean-tail exhibit, and the reminder that tail structure is not tail efficiency

Open Questions#

Does the H1 2025 jump to 36.3% survive a full-year datapoint, or is it a half-year artifact? Every prior step is 0.6–2.7pp and this one is 5.8pp. Trigger: Carta's 2025 full-year or 2026 update to this series.
Is the employee-equity null a real population fact or a median artifact? Carta reports near-identical medians; the publisher claims 2–5× among founders in his own program. A distributional cut — variance or upper decile of first-five grants, split by founding-team size — would settle it, and neither party publishes one.
Does the solo-founded tail differ from co-founded companies on revenue per head? This dataset cannot say — it holds cap tables, not revenue — and it is the missing half of AI Investment Story, Not Efficiency Story's tail question.

Sources#

The State of Solo Founding (Solo Founders Report 2025) — The State of Solo Founding (Solo Founders Report 2025), Julian Weisser & Kieran Ryan (Solo Founders) with Peter Walker & Hamza Shad (Carta Insights), 2025-12-09, empirical for the Carta cap-table data on tens of thousands of U.S. companies. COI is direct and disclosed in-file: the publisher sells a paid Solo Founders Program on the same page and its stated thesis is that solo founding is the future — all interpretation above is attributed, and the three sections where the data contradict the framing are the ones carried. Ingested from solofounders.com/report (the ~43K-char full web report) rather than the ~4.6K-char Carta landing-page excerpt; a 100-page PDF behind an email gate was not taken. Chart-parse note: the report is an interactive tabbed document rendering one chart image per tab, and only the share-of-solo-founded-companies chart was captured and transcribed (the annual series in the table above, which the prose gives only as endpoints) — no other chart is cited here, and the raw file records the same restriction. Figures otherwise quoted from prose