Solo founders are 36% of Carta's tracked startups now. The AI that built your product armed everyone else too.
Carta's own December 2025 report puts a real number on something I'd only felt anecdotally: from 2019 to the first half of 2025, the share of new US startups on its platform with a single founder climbed from 23.7% to 36.3%. Peter Walker, who runs insights at Carta, called it a thirteen point rise in about five years and named the cause directly: AI is lowering the cost of starting a company. I read that report the same week I renewed my Cursor subscription for the fourth month running, and it landed less like validation and more like a warning.
What the Carta number actually says
The 36.3% figure is Carta's own, pulled from tens of thousands of US companies on its cap table platform plus interviews it ran with Solo Founders, the research group it partnered with for the report. That part checks out directly from the source.
What I could not verify is the specific "double since 2016" framing that shows up in some of the secondary write ups covering this data, including the piece that pointed me toward this topic. Those pieces cite solo founders at somewhere around 18% in 2016, climbing to today's third plus of new startups. Carta's own published report only shows the 2019 to 2025 window, so I'm treating "roughly double in a decade" as a widely repeated estimate rather than a number I can trace back to Carta's own chart. The verified version is dramatic enough on its own: a thirteen point jump in five years, and the trend accelerated rather than leveled off through the first half of 2025.
Two more data points from the same report matter more than the headline stat. Solo founders now hire their first employee at a median of 399 days after incorporation, against 480 days for multi founder teams, meaning solo builders are running longer before they need help, not just starting alone. And solo led companies represented 30% of 2024 incorporations but pulled in only 14.7% of the cash raised in priced equity rounds that year. Building alone has normalized faster than funding alone has.
What's actually driving it
The mechanism is not mysterious, and it's the same one every solo builder reading this already lives inside. Five to eight years ago, going solo meant you were the founder, the engineer, the designer, and the support line, and each of those roles ate real hours you had to find somewhere. Today an AI coding tool writes most of the boilerplate, drafts the landing page copy, and can hold a decent first pass at a support conversation. The tasks that used to require hiring, or at minimum a co-founder to split the load with, now fit inside one person's calendar because a model is doing the parts that used to take a specialist.
That's the whole story behind the Carta number. It isn't that more people suddenly want to build companies alone. It's that building alone stopped requiring a team's worth of skills crammed into one head.
The stack that makes this possible, and what it actually costs
The rough shape of the 2026 solo stack is familiar to anyone in this space: an AI coding tool, a backend for data and auth, a merchant of record for payments, an analytics tool, and transactional email. Here's what I found when I priced it out properly rather than repeating the "under $20 a month" line that circulates in indie hacker posts.
- AI coding tool: Cursor Pro is $20 a month ($16 a month billed annually). Lovable's Pro tier runs $25 a month ($21 annually).
- Backend (data and auth): Supabase's free tier covers 500MB of database, 50,000 monthly active users, and unlimited API requests, so this is genuinely $0 until you have real traffic.
- Payments: Polar and Lemon Squeezy both work as a merchant of record, meaning they handle sales tax and VAT for you. Polar's current published rate is 5% plus 50 cents per transaction on its free Starter plan (accounts created before its late May 2026 pricing change stay grandfathered at 4% plus 40 cents). This is a cut of revenue, not a fixed bill, so it costs nothing before you have a paying customer.
- Analytics: Plausible has no permanent free tier anymore, just a 30 day trial. Its cheapest paid plan is $9 a month for up to 10,000 monthly pageviews.
- Email: Resend's free tier covers 3,000 emails a month, capped at 100 a day, which is plenty for transactional email on a pre-revenue product.
Add it up honestly and the floor is closer to $20 to $29 a month once you're paying for a coding tool and want real analytics, not the flat "under $20" figure that gets repeated as a talking point. You can get under $20 if you lean on Cursor's or Lovable's free tiers while prototyping and skip paid analytics until you have traffic worth measuring. Either way, compare that to what the same solo build cost in 2018: a few hundred dollars a month in hosting and infrastructure before a single customer, weeks of your own time hand rolling authentication instead of dropping in Supabase Auth, and no AI pair programmer to turn a rough feature spec into working code in an afternoon. The dollar cost dropped by an order of magnitude. The time cost dropped by more than that.
The part that doesn't show up in the "AI lets you build alone" posts
Here's the part most of these posts skip past: if the cost of building an MVP fell this far for you, it fell exactly as far for the other fifty people who had your same idea this week. The build side of starting a company just got commoditized, and commoditized things stop being where competitive advantage lives. When everyone can ship a working product in a weekend, shipping a working product in a weekend is no longer the hard part or the differentiated part. It's table stakes.
What doesn't get cheaper at the same rate is distribution: an actual audience that trusts you, a specific niche you understand better than a generic AI-assisted competitor does, a channel you've spent real months building instead of a landing page you spun up in an afternoon. That part is still slow, still expensive in time, and still genuinely yours in a way a Cursor subscription never will be. The uncomfortable math is that the thing getting cheaper (build) was never the scarce resource, and the thing that stayed expensive (distribution and a real relationship with a specific audience) was the scarce resource all along. AI just made that gap more obvious by removing the excuse that building was hard.
The honest take
Here's where I have to implicate myself, because I don't think I'm an exception to this. I checked my own hours this week before writing this: build side (coding, testing, fixing bugs in my own tools) versus distribution side (writing, replying to people, actually reaching anyone who isn't already following me), and it was not close to 50/50. It skewed hard toward build, the way it has skewed for months, because build feels like productive work and distribution feels like uncertain work with no guaranteed payoff on any given day.
So the concrete thing I'd actually do, and the thing I'm telling myself to do starting this week: track your hours for seven days, split them into build and distribution, and look at the ratio once you have something that actually works. If it isn't close to 50/50 by the time your product functions, that ratio is the actual problem, not your feature list, not your onboarding flow, not the next integration you were about to add. The build queue will always have one more item on it. That's exactly why it's the easier place to hide.
Where I could be wrong: some categories genuinely do reward more build time even post-product-market-fit, deep technical products with real moats being the obvious case, and not every solo builder is competing in a crowded, AI-flattened niche where distribution is the whole game. But for the standard indie SaaS or content-adjacent tool that this stack is built for, I think the ratio test holds, and I think most of us, myself included, already know which side of it we're avoiding.
Author
Lukas
@lukcombinator