The Loneliness Economy: What It Costs to Buy Back Connection
Company has become a line item. A look at what the loneliness economy sells, why the spending can be rational, and one test that separates it from waste.
There is a line item in a lot of budgets now that did not exist a generation ago, and most people have never added it up. It hides across four or five categories — a subscription here, a membership there, a class, a club fee, a standing Saturday expense — and every one of them is really buying the same thing. Company.
I do not think that is shameful. I think it is worth pricing.
What is actually being sold
The market that has grown up around isolation is broader than the headlines suggest, and it is not all apps. Sorted roughly by what you are paying for:
Access to a room with people in it. Coworking memberships marketed on community rather than desks. Members' clubs. Gyms whose actual product is the 6am class where the same twelve people show up. In most US cities, community-forward coworking runs somewhere in the low hundreds a month; the members'-club tier runs into the low thousands a year. You are renting a third place — the pub, the church hall, the union local — that used to be free or nearly free.
Structured activity with strangers. Run clubs, supper clubs, hobby leagues, class-based fitness, group travel. Prices run from nothing to a few hundred a month. The product here is not the activity. It is the recurrence and the low social cost of showing up alone.
A person, hired. Rent-a-friend platforms billed hourly. Professional cuddling, typically advertised somewhere near a hundred dollars an hour. Coaches and therapists, who often deliver something in this category alongside their actual clinical or professional work. This is the tier that makes people flinch, and it is worth noticing that the flinch is mostly about the transaction being visible.
A companion that is not a person. AI companion subscriptions, usually priced like any other consumer app — ten to twenty dollars a month, sometimes more for the tiers that remember you. This is the fastest-moving corner and the cheapest by a wide margin, which is exactly what makes it worth thinking about carefully.
Two things stand out when you lay them side by side. The price range is enormous — three orders of magnitude between an AI subscription and a members' club. And the pricing has almost nothing to do with how much actual connection each one produces.
The infrastructure we stopped paying for another way
The honest version of this story is not that people got lonely and companies pounced. It is that a set of institutions that used to produce connection as a by-product stopped producing it, and something had to fill the gap.
Robert Putnam documented the mechanics of that in Bowling Alone — the long decline in membership of the unions, lodges, congregations, bowling leagues, and civic associations that once organised American social life. Ray Oldenburg had already named what was disappearing: the third place, somewhere that is neither home nor work, where showing up alone is normal and nobody is selling you anything.
Those institutions were never free. They cost dues, tithes, volunteer hours, and a lot of tolerance for people you did not choose. But the cost was bundled into something you were doing anyway, which meant nobody experienced it as a connection expense. What is new is not the cost. It is the unbundling — connection now arrives as a discrete purchase with its own line item, which is the same thing that happened to news, to music, and to transport.
Meanwhile the case for taking isolation seriously as a health matter got a lot stronger. The 2023 US Surgeon General's advisory on loneliness pulled the research together and put the mortality risk of social disconnection in the range of smoking around fifteen cigarettes a day. Whatever you think of that comparison as a piece of communication, the underlying finding is robust: isolation is a physical health problem, not just an unpleasant mood.
Put those two together and the spending starts to look less like weakness. If the free infrastructure has thinned, and the cost of going without is measured in years, then buying some of it back is a defensible allocation of money. It belongs in the same mental category as a gym membership or a dental cleaning — a purchase against a known long-run risk.
Where it goes wrong
The trap is not that paid connection is fake. It is that paid connection is easy, and the free kind is not, and easy things reliably crowd out hard things when both are on the table.
Building an unpaid friendship in your late thirties is genuinely difficult work. It requires initiating without being asked, tolerating ambiguity about whether the other person wants this, remembering things about their life, and — hardest — showing up repeatedly with no structure forcing you to. Nobody sends you a calendar invite for a friendship.
Every paid option removes exactly those frictions. The club has a schedule. The class has an instructor. The app never has a bad day and never needs anything from you. That removal is the product, and it is genuinely valuable — for someone who has moved to a new city and knows no one, a structured paid thing may be the only realistic on-ramp there is.
The failure mode is when the on-ramp becomes the destination. You can spend meaningful money every month, feel socially active, be around people constantly, and end a year with no more actual friends than you started with. The spending was not wasted, exactly — the room was pleasant, the class was good. It just did not compound.
That word is the whole thing. In every other part of a financial life we distinguish between consumption and investment. Money spent on connection deserves the same distinction, and almost nobody applies it.
The connection spending audit
Here is the test I would use, and it is deliberately narrow: did this spending produce repeat contact with the same specific, named human being?
Not "was it enjoyable." Not "did I meet people." Repeat contact, same person, whose name you know. That is the only thing that turns an expense into infrastructure, because friendship is a function of repetition with a fixed set of people, and nothing else in the audit correlates with it.
Pull three months of statements and build a table. Every line where you would honestly say part of the motivation was company:
| Line item | Monthly cost | Names I now know | Contact outside the paid setting? | Verdict |
|---|---|---|---|---|
| Coworking membership | $260 | 4 | Yes — lunch with 2 of them | Infrastructure |
| Climbing gym | $85 | 2 | No | Consumption (promising) |
| AI companion app | $15 | 0 | N/A | Consumption |
| Supper club, 4x | $70 | 0 | No | Consumption |
Three columns do the work. Names I now know is brutal and clarifying — most people discover they have been attending something for a year and can name nobody. Contact outside the paid setting is the compounding test: if the relationship cannot survive outside the container you are renting, you are renting the relationship. And the verdict column forces the distinction that the statement itself hides.
Then total the two verdicts separately. The number that matters is not what you spend. It is the ratio. If your consumption column is four times your infrastructure column, you have a portfolio problem, not a budget problem.
One caution, because the audit can turn mean if you let it. Consumption is not a sin. Some spending on company is meant to be consumed and that is entirely fine — a concert, a good meal out, a class you took for the class. The audit is not there to shame the consumption line. It is there to stop you from mistaking it for the other line.
Actually budgeting for it
Most budget templates have no category for this, which is why it hides. Groceries, transport, entertainment, subscriptions, health — connection spending gets sprayed across all five and never gets looked at as a whole. The first move is simply to give it a name and a number.
I would set the number as a percentage of take-home rather than an absolute, because the right amount genuinely differs. Somewhere in the low single digits of take-home pay is a reasonable starting point for most people, higher in the first year in a new city, lower once you have a settled group and the spending has done its job.
Then three rules that make the category behave.
Front-load the year, do not spread it. Connection spending has a threshold effect. Showing up to the same thing weekly for two months does something that showing up monthly for a year does not, even at identical cost. If the budget only supports one, buy frequency and buy it in a concentrated block.
Set a review date on every recurring line. Six months, calendared. At the review, run the names test. A membership that has produced zero names in six months is not failing to be fun — it is failing at the thing you bought it for, and it should either change form or be cancelled. Almost nobody cancels these, because cancelling feels like admitting you have no friends. It is not that. It is reallocating.
Budget the unpaid version too — in hours. This is the one people skip and it is the important one. Put a number on hours per month spent initiating unpaid contact: the text you send first, the invitation with no event attached, the drive to someone's house. If the dollar line is going up and the hour line is flat, the substitution has happened, whatever the audit table says.
The last one is where I have caught myself. With a young child and a full working week, the paid option is not tempting because it is better — it is tempting because it fits in a slot. You can schedule a class. You cannot schedule the ninety unstructured minutes in which a friendship actually gets built, and the slot-shaped thing will win every single time unless you deliberately protect the other.
That is really the whole argument. Spend the money — the infrastructure is genuinely thinner than it was, and buying some of it back is a reasonable response to a real problem. Just do not let the ease of the purchase substitute for the difficulty of the thing, and check once a year whether the line item is buying you people or only buying you rooms.
Questions people ask
Is it pathetic to pay for social connection?
No, and the instinct that it is comes from a bundling accident. Our grandparents paid too — dues, tithes, and a great deal of obligation — the cost was just wrapped inside institutions they joined for other reasons. The transaction being visible now is what makes it feel shameful. Visibility is not the same as newness.
How much should I actually budget?
There is no researched figure, so treat any number you see, including mine, as a starting heuristic rather than a finding. Low single-digit percentages of take-home works for most people. The more useful discipline is the ratio between spending that produces repeat contact and spending that does not, which is measurable in a way that a target percentage is not.
Do AI companions belong in a connection budget at all?
They belong in the audit, because the money is real and the motivation is company. They will almost always land in the consumption column, since by construction they cannot produce a named human you see outside the app. That does not automatically make them a bad purchase — it makes them a bad substitute for the other column, which is a different claim.
What if I genuinely cannot afford any of this?
Then the hours line is the whole budget, and it is not a lesser version. The single highest-return connection activity available to anyone is initiating unpaid contact with people you already sort-of know, repeatedly, over months. It costs nothing and it is harder than any purchase on the list, which is exactly why the paid market exists.
My paid activity is fun but has produced no friendships. Should I quit it?
Only if you bought it for friendship. If you bought it for the activity and it delivers the activity, it is working and the names test does not apply. The audit fails when you use it to judge every expense by one metric. It works when you use it to check whether a specific stated intention is being met.
What strikes me, looking at the whole category, is how much of it is really selling permission — a structure that makes it acceptable to show up alone and want company. That permission used to be free, distributed by institutions most people no longer belong to. It seems worth noticing that we are now buying back, at retail, a thing we used to get as a by-product of belonging to something.