AI Companion Funding Statistics 2026: Where the Money Actually Went

1. Published estimates put the AI companion market anywhere from $4.2 billion to $37 billion in 2025, heading toward $100B-$550B over the next decade.

Honesty ladder first: these ranges disagree because “AI companion” has no settled definition. Grand View pegs the broad market at $36.79B in 2025 growing 31% annually toward $317.96B by 2033; GM Insights sizes companion apps at $14.1B in 2024 toward $115.3B by 2034; the narrowest app-only definition (Research and Markets) says $4.24B in 2025. Every one of these is a forecast of a broadly defined future, so this page reports the range, not a false precision.

2. The reality check: observed consumer spending inside dedicated companion apps ran around $120 million in 2025, against forecasts in the tens of billions.

The most important market-size fact is the gap between models and receipts: app-store intelligence tracked roughly $120M of actual in-app companion spending in 2025, two orders of magnitude below the analyst figures. Much real revenue hides outside app stores, on the web, where adult platforms live precisely to escape app-store rules, which means the truest market number is unknowable, and web-first NSFW revenue (the Candy AIs of the world) is systematically missing from every official estimate. Note the rhyme with this page’s funding thesis: the adult side is undercounted in the market data for the same reason it’s absent from the funding data.

3. The segment closest to this site: the AI girlfriend app market is sized at $2.32 billion in 2025, reaching $2.91 billion in 2026, about 25.5% annual growth.

A rare girlfriend-specific estimate, methodology not public, so treat it as directional. Even at face value it implies the romantic segment alone out-earns the narrowest whole-category estimates from two years earlier, which matches what my testing wallet keeps discovering: this end of the market is where the paying happens. My deep numbers on it live in my AI girlfriend statistics.

4. 72% of US teens have used an AI companion, and 52% use one regularly.

The adoption statistic that reframes the entire category from niche to mainstream, and, for investors, the one that explains both the appetite (a generational behavior shift) and the anxiety (it’s minors, which is exactly why 2026’s funded startups all lead with safety positioning).

5. Roughly 1 in 5 US adults has chatted with an AI romantic partner, and “companionship and therapy” ranked as the #1 use of generative AI in 2025.

Companionship isn’t a side effect of the AI boom; by usage ranking, it’s the main event. Hold this against the funding map: the most common use of generative AI is the one tier-one capital funds only in de-romanced form.

6. The category’s engagement is what usage-based businesses dream of: Character.AI sessions average around half an hour, and Chai reports 2 million daily users at 90-minute average sessions.

Session lengths that embarrass social media are the reason this category keeps attracting capital despite its risks: whatever else these companions are, they are the stickiest consumer AI products yet measured.

7. Worldwide search interest for “ai girlfriend” is up +2,700% versus the preceding five years, hitting its all-time peak in 2026.

The shape tells the story better than the multiplier: the term barely existed before 2023, ignited in the months after ChatGPT’s launch, plateaued through 2024 at roughly a third of today’s interest, then doubled across 2025 and 2026 to touch the 100 mark, the all-time high, this year. Note the sequencing against the funding data above: search demand went vertical in 2023, while the institutional funding wave arrived in 2025-26, two years behind the users, and only for the de-romanced versions of what they were searching for.

8. Companion chatbot companies have disclosed ~$415.6 million in institutional funding across 30 tracked companies, 15 of which have raised at all.

The methodology sentence that most funding articles skip: this counts publicly disclosed rounds only, treats undisclosed as unknown rather than zero, and excludes one giant edge case covered below. Real capital in this category is likely higher, but this is what anyone can actually verify, and the pattern inside it is the story.

9. The average disclosed total is $121.3 million; the median is $9.25 million. That 13x spread is the concentration statistic.

When an average runs thirteen times the median, a handful of giants are carrying the entire number. The typical funded company in this category raised well under ten million dollars; the category’s headline figure is really three or four stories wearing one statistic.

10. The top three raises, Character.AI ($193M), Embodied ($70M+), and Chai ($55M), hold 76.5% of all non-edge-case funding.

Three companies, three-quarters of the money, and notably, three completely different fates: one sold its founders to Google, one shut down and bricked its robots, and one is a profitable business funded by its GPU vendor. Concentration says nothing about outcomes here, which is its own finding.

11. The edge case: Inflection AI raised $1.525 billion for its companion chatbot Pi, more than the rest of the category combined, times three.

Honesty ladder: I track Inflection but exclude it from headline sums, because Pi was a general companion chatbot from a frontier lab, not a character or relationship platform, and its capital story ended differently too: a ~$650M Microsoft licensing-and-acquihire deal in March 2024, after which Pi was effectively sunset. Include it and the category total balloons to $1.94 billion while describing the market worse.

12. Exactly 15 of the 30 companies tracked, half the dataset, have no disclosed funding of any kind.

And this isn’t the obscure half: it includes several of the highest-traffic platforms in the category. In most markets, no-funding correlates with no-traction. In this one, the unfunded half contains much of the traction, which is the anomaly the rest of this page unpacks.

13. The 2025-26 funded wave: five companion startups raised a combined $66.5M+: Tolan ($30M), Status AI ($17M), CODE27 ($10M+), Shapes ($8M), and Charms ($1.5M).

A genuine reopening of the category after the post-Character.AI chill, with marquee names attached: Khosla led Tolan’s Series A, Lightspeed led Shapes’ seed, and General Catalyst and Union Square Ventures backed Status. Companion chatbots are very much fundable in 2026. The question is which kind.

14. The pattern across all five: every newly funded companion startup is structurally distanced from romance.

Tolan’s aliens are programmed to refuse romantic and sexual interactions entirely. Shapes markets itself as the antidote to “AI psychosis” and one-on-one attachment. Status turns characters into a gamified social world. CODE27 sells licensed anime IP in hardware. Charms builds a creator economy. Five rounds, five different mechanisms for being a companion company that is definitely not an AI girlfriend company. Whatever the reason, that positioning is consistent across every newly funded company in this group.

15. Every tier-one venture round in the dataset sits in the SFW or wellness cohort: a16z, Khosla Ventures, Lightspeed, General Catalyst, Union Square Ventures, Y Combinator, Coatue.

I built a separate investor-level table for this page, and the cleanest way to state its finding: I could not find a single tier-one venture firm leading a disclosed round in an explicitly NSFW companion company. Not one, across the entire history of the category. The closest exceptions are tiny: an accelerator check into Dippy and roughly $1M of angel money into SpicyChat.

16. The transparency tell: SFW rounds come with proudly named leads; the category’s edgier raises hide their investors.

Tolan’s announcement names Khosla and Keith Rabois in the first paragraph. Shapes names Lightspeed. Meanwhile CODE27’s $10M arrives from unnamed “leading global venture funds,” and on the adult side, disclosure simply doesn’t happen at all. In this category, whether investors want their name attached is itself a data point about where the reputational line sits.

17. Disclosed funding by cohort: SFW and wellness platforms $412.5M; NSFW-friendly platforms $3.1M. That is 99.25% against 0.75%.

This is the headline test of the hypothesis, and the gap is not subtle. For every dollar of disclosed institutional capital that has reached an adult-friendly companion platform, about $133 went to the safe side of the category.

18. 13 SFW-cohort companies have disclosed at least one round. On the NSFW side: two, SpicyChat (~$1M of angel money) and Dippy ($2.1M from an accelerator).

And note what those two exceptions are: not institutional venture rounds with named funds, but the smallest checks in startup finance. The largest disclosed NSFW-cohort total ($2.1M) is 92 times smaller than the largest SFW one ($193M).

19. Median disclosed raise: $10.5M in the SFW cohort vs $1M in the NSFW cohort.

Medians strip out the giants, and the gap survives: even a typical funded SFW companion company raises ten times what the rare funded adult one does. Whatever explains this pattern, it isn’t Character.AI skewing the math.

20. The traffic-vs-capital disconnect: Candy AI alone draws an estimated 15-25 million monthly visits, top-five in the category, on 0% of its funding.

And Candy is not the outlier; it’s the pattern. JanitorAI operates what testers describe as the largest character library in the unfiltered space with zero funding disclosures; SEMrush measured roughly 138 million visits in December 2025, and Similarweb ranked it inside the top 250 websites globally by mid-2026. Add CrushOn, Muah, Nomi, and Kindroid, all undisclosed, and a very large share of real category demand runs on platforms that institutional capital has, on paper, never touched.

21. The revenue disconnect is starker: the NSFW side’s flagship claims more ARR than the funded side’s flagship earns.

Candy AI’s claimed $25M+ ARR (traffic-based estimates run $50-120M) was reached on zero disclosed dollars. Character.AI, with $193M raised, was reported around $32M in 2024 revenue, rising to roughly $50M in 2025 (Business of Apps; Sacra’s equity research tracked a $30M run rate mid-2025). Capital and cash flow are living on opposite sides of this market, which is the strangest single fact in the dataset.

22. Chai runs a $70M ARR business whose $55M in capital came from its GPU vendor, CoreWeave, not from venture firms.

The most instructive capital structure in the category: a $2M pre-seed in June 2023, then strategic rounds from CoreWeave at a $450M valuation, capital from the company selling Chai its compute. Chai proves you can build one of the category’s biggest revenue businesses while never taking a traditional VC dollar, and that alternatives to Sand Hill Road exist for companies that want growth capital without its conditions.

23. Of the 15 companies with no disclosed funding, the dataset finds three distinct models: revenue-funded, parent-funded, and deliberately opaque.

Half the dataset has no disclosed funding, but the reasons split three ways:

• Revenue-funded (apparent): Candy AI, JanitorAI, Kindroid, Nomi, and peers, where subscriptions appear to carry the business outright.
• Parent-funded: Talkie, whose capital flows through its Chinese AI-lab parent MiniMax (roughly $1.15B raised privately before its January 2026 Hong Kong IPO), and EVA AI, operated inside the private dating conglomerate Social Discovery Group.
• Opaque: CrushOn, Muah, and several girlfriend apps whose corporate structure resists mapping entirely.

“Unfunded” turns out to mean three different things, and only one of them means what the hypothesis assumed.

24. The ownership-opacity stat, from my own testing: adult companion platforms bill under names like “Dream AI” and “NDAI.CC”, and their corporate parents are barely traceable.

I pay for and test these apps myself, and the biller names on my statements are deliberately quiet: DreamGF charges as “Dream AI” and GirlfriendGPT as “NDAI.CC”, findings from my own subscriptions. Candy AI’s operator EverAI has been linked in reporting to a Malta-based structure and the Octavian Holding group. This matters for funding data: an adult platform could hold private capital invisibly, because nothing about these structures is built for disclosure. Which is exactly why this page counts disclosed funding only, and says so.

25. The economics explanation: category analyses put gross margins on paid adult-companion subscriptions near 90%, which makes venture capital optional.

A subscription business converting at adult-content prices with 90% gross margins funds its own growth. On those numbers, the absence of VC on the adult side stops looking like rejection and starts looking like founders who never needed to dilute, and who benefit from staying private in a category where investor scrutiny cuts both ways.

26. Character.AI: $43M seed (Dec 2021) → $150M Series A at $1B (Mar 2023, a16z) → Google licensing deal at $2.5B (Aug 2024) that bought out investors and took the founders back to Google.

The category’s defining capital story: the fastest unicorn run in consumer AI, then the “reverse acquihire” that sent Noam Shazeer, Daniel De Freitas, and ~30 researchers to Google while investors exited at 2.5x. Worth observing, carefully: Character.AI prohibited explicit content from its earliest days, before, during, and after every round. Its investor-friendly positioning didn’t follow the money; it preceded it, which is at least consistent with positioning being part of what made the money possible.

27. Replika, the category’s pioneer, raised just $11 million across three rounds, and its romantic features have whiplashed with its regulatory life, not its funding.

Founded in 2015 (as Luka), backed early by Y Combinator, and never venture-scaled: Character.AI raised 17 times more in a single round than Replika has in a decade. Its adult-content arc ran on a different axis entirely: erotic roleplay removed in February 2023 after Italy’s regulator intervened, partially restored for legacy users after user outcry. Replika is the nuance case: modest capital, shifting content policy, and proof the two don’t move in lockstep.

28. Candy AI is the bootstrapped giant: $0 disclosed funding, $25M+ claimed ARR, top-five category traffic, and a reported M&A offer in 2025.

The inverse of Character.AI on every axis: no named investors, no disclosed rounds, explicit content at the core of the product, and a business that apparently grew to eight figures on its own cash flow inside two years. If the funding gap has a poster child on each side, these two companies are it.

29. The biggest write-off: Embodied raised roughly $69M from Intel Capital, Amazon’s Alexa Fund, Sony, and Toyota, and shut down in December 2024 when a funding round collapsed, initially bricking children’s Moxie robots.

The category’s hardest lesson in what funded doesn’t mean: a corporate-VC-backed, safety-positioned, kids-wellness companion, the most fundable profile imaginable, whose $799 robots stopped working when the money did. A community rescue effort later restored partial functionality. Meanwhile several unfunded adult platforms from the same era are still running profitably.

30. The quieter failures: Dot shut down in October 2025, Soulmate was sold and closed in 2023, and Forever Voices collapsed after its founder’s 2023 arrest.

Three different failure modes: Dot’s founders split over vision with roughly $3.7M raised; Soulmate’s owner sold the company and gave users one week to say goodbye to companions some genuinely grieved; Forever Voices (maker of the CarynAI girlfriend bot) simply imploded. Shutdown risk, notably, runs across both cohorts, funded and not.

31. And the largest companion bet no longer exists as one: Inflection’s $1.525B-funded Pi was effectively sunset after Microsoft’s ~$650M licensing-and-acquihire in March 2024.

The edge case’s ending completes the outcomes picture: of the three biggest capital events in companionship-adjacent AI, two ended with Big Tech extracting the founders and models (Inflection to Microsoft, Character.AI to Google) rather than the products winning their markets. Investors did fine; the companions themselves became licensing collateral.

32. The verdict: institutional capital and adult revenue occupy remarkably separate halves of the companion chatbot market.

Three conclusions, each carried by its own numbers. First, the money is top-heavy: a few enormous rounds dominate a median raise under ten million, so the category’s headline total describes a handful of companies. Second, the institutional side has a hard content line: 99.25% of disclosed capital sits with SFW and wellness platforms, no tier-one firm appears in any adult-companion round, and 2026’s newly funded startups are aggressively de-romanced by design. Third, the adult side largely runs without disclosed capital: subscription businesses, quiet billing entities, and companies like Candy AI that report a claimed $25M+ ARR with $0 raised on the record. That division is what the dataset demonstrates. Why it exists, whether it is a wall one side built or a border both sides keep, is a question these numbers cannot answer by themselves; my interpretation lives in my research note on the adult-content funding line.