I Followed the Money Behind 30 Companion Chatbot Companies. This Is How VC Funding Sees the Adult Content Line.

Why I counted the money

Most startups brag about their funding. Companion chatbot startups are a strange case: many of them sell romance and explicit content, the exact things that make traditional investors nervous, so it’s fair to wonder how they get funded at all, or whether they do. Everyone in this niche assumes VCs won’t touch the adult side, but nobody I could find had actually counted. Since I pay for and test these apps, I had a personal stake in the answer: the products I live in every day split visibly into two species, the polished safe ones and the subscriber-funded spicy ones, as if they were built with different kinds of money. So I stopped guessing and built the ledger: who raised, from whom, how much, and what happened next. The pattern showed up before I finished the spreadsheet, and it fits in four comparisons. Everything below is me walking through them.

Character.AI raised $193M. What did all that money buy?

Character.AI is the category’s capital champion: $43M seed in late 2021, then a $150M Series A from Andreessen Horowitz in March 2023 at a $1 billion valuation, $193M in total. So it’s the right place to ask the question funding articles never ask: what did the money actually buy?

Traction, unquestionably: roughly 20 million monthly users and session lengths around half an hour, engagement numbers social networks would kill for. But the business underneath ran harder: revenue was estimated around just $32M in 2024, with press estimates putting 2025 near $50M, still a fraction of what $193M-backed expectations imply, against the brutal compute costs of serving billions of free messages. And the strict all-ages moderation that made it fundable is the same thing its power users complain about most, something my own testing keeps bumping into from the other side. The ending says the most: in August 2024, Google paid for a licensing deal valuing the company at $2.5 billion, bought out the investors, and took founders Noam Shazeer and Daniel De Freitas back, with about 30 researchers. The investors got their exit. The product lost its founders. The biggest funding story in the companion chatbot industry ended with the money winning and the companion becoming collateral.

Replika built this category on just $11M

Now the control case. Replika invented much of this category in 2015 and has raised, lifetime, about $11 million, one-seventeenth of Character.AI’s single Series A. And yet: 10 million+ registered users, a decade of survival, a business that weathered the most violent content whiplash in the category (erotic roleplay removed in February 2023 under Italian regulatory pressure, partially restored after users grieved publicly), and it’s still here, a decade on, while much better-funded companion companies around it got acquired or shut down. The failures make the same point from both directions: Embodied raised $68.78M and shut down in December 2024 when a round collapsed, Dot folded with about $3.7M reported raised, while on the opaque side Soulmate (funding undisclosed) and Forever Voices (angel money, largely undisclosed) disappeared too. Funding did not guarantee durability, and the lack of it did not necessarily prevent it. Capital, it turns out, is not destiny in this market. Which raises the question the next chapter answers: if modest money can build a durable companion, what exactly is the big money selecting for?

The new VC money funds companionship without the AI girlfriend label

2025 and 2026 reopened the category’s wallet: five startups, $66.5M+, marquee firms back on the cap tables. And once I lined the five up, the selection rule was impossible to miss. Tolan ($30M, Khosla) builds alien companions programmed to refuse romance entirely. Shapes ($8M, Lightspeed) markets itself as the antidote to unhealthy one-on-one AI attachment. Status ($17M, General Catalyst and USV) turns characters into a gamified social world. CODE27 ($10M+) sells licensed anime IP in desktop hardware. Charms ($1.5M) builds a creator economy. Five rounds, five different engineering solutions to the same brand problem: how to be a companion company without being an AI girlfriend company. The market’s biggest use case, per 2025’s usage research, companionship and emotional support, is exactly what the new money funds, each time with the girlfriend part engineered out of the brand.

The investor web: a16z, YC, CoreWeave, and Google keep showing up

Build the investor-level table and the starkest number in the dataset surfaces first: ten-plus rounds led by tier-one firms on the SFW and wellness side (a16z, Khosla, Lightspeed, General Catalyst, USV), and zero tier-one-led rounds in the NSFW-friendly cohort. The adult side’s entire disclosed capital amounts to Dippy’s $2.1M accelerator check and SpicyChat’s roughly $1M angel round. Smaller patterns surface underneath. Y Combinator appears twice, a decade apart: it seeded Replika’s parent in the mid-2010s and backed Status in 2026, the same bet, companionship, at the category’s two safest moments. Andreessen Horowitz also appears twice, and the second one is the exception that proves every rule on this page: alongside its Character.AI lead sits a $2.1M check into Dippy, a spicier companion app, but through a16z’s Speedrun accelerator rather than the main fund. The closest any tier-one name gets to the adult side of this market is an accelerator’s arm’s length. Then there’s the capital that isn’t venture at all: Chai’s $55M came from CoreWeave, its GPU vendor, a supplier financing its own demand. And hovering above the whole table, the two names that ended the two biggest stories: Google and Microsoft, who didn’t fund the category so much as harvest it, extracting Character.AI’s and Inflection’s founders in twin licensing acquihires. In this market, Big Tech isn’t an investor. It’s the exit.

Half these companies have no disclosed funding, and most of the adult leaders are among them

Here’s where the story inverts. Half my dataset, 15 of 30 companies, has no disclosed funding, and it’s not the failing half. The cleanest case is Candy AI: $0 raised according to Latka’s company profile, a claimed $25M+ ARR, and an estimated 15-25 million monthly visits. Put it next to Character.AI and the title of this article compresses into one contrast: Character.AI raised $193M to build enormous scale; Candy AI reports $0 raised and appears to have built a substantial customer-funded business. One is investor-financed growth, the other looks like customer-financed growth, and only the first kind shows up in funding databases. JanitorAI runs one of the biggest character libraries on the internet, roughly 138 million visits a month by SEMrush’s count, on zero disclosures. Kindroid, Nomi, CrushOn, Muah: no disclosed funding on any of them. One caution before the story runs away with itself: a blank record is not proof of a bootstrap. Candy’s $0 is on the record; the other 14 are a mix of undisclosed, opaque, parent-funded, and possibly angel-funded off the record. What the group shares is the incentive: adult-priced subscriptions can plausibly fund their own growth, and businesses that bill through quiet names like “Dream AI” and “NDAI.CC” (I know because I pay them monthly) and operate from Malta and stranger places benefit from exactly the opacity that venture capital’s press-release culture destroys.

The border: capital and adult revenue keep to remarkably separate halves

So the pattern, stated plainly: institutional capital and adult revenue appear to occupy remarkably separate halves of the companion chatbot market. 99.25% of disclosed capital sits on one side of the content line, while several adult platforms appear to have reached substantial consumer scale with almost none of it. My read on why the halves stay separate, and it is a read: investors protect their names, and adult platforms protect their independence.

One honest limit before the conclusions: my dataset can’t tell me why individual investors made these decisions. What it can tell me is where the money actually went. Payment processors, app-store rules, reputational risk, and exit math are all plausible explanations for the border, but those are my interpretation, not findings from the data.

As someone who pays for and tests both kinds, I keep coming back to this: the border explains more about these products than any feature list. The funded apps behave like companies with boards: safe, moderated, positioned for the next round, and forever at risk of a Google-shaped ending. The bootstrapped apps behave like businesses with customers: spicier, scrappier, quicker to ship what users pay for, and accountable to nobody’s reputation but their biller name. That may be why the border is surprisingly stable: each half already has the money it needs. The funded side has found ways to build companionship without embracing adult content. The adult side shows almost no institutional backing in my dataset, yet several of its platforms appear capable of growing without institutional money at all, and investor cash would mostly just buy them a boss. From where I sit as a paying user, neither side seems under much pressure to become the other. The complete numbers behind this story, every total, median, cohort split, and timeline, are in the funding statistics page, and the raw dataset is yours to download.