You Save $3,768. Your Data Earns Them $6,565.

Introduction

The average American puts away about $3,768 a year. In that same twelve months, the platforms they use for "free" earn an estimated $6,565 off their data. The number that leaves your account is the one you see. The number you generate is the one nobody is required to show you.

That second figure comes from a report the Web3 Foundation published May 25. It's the first attempt I've seen to price the full machine, not just ad revenue, but AI subscriptions, enterprise licensing, API access, data brokerage, and the recommendation engines that decide what you buy next. The headline: Amazon, Alphabet, Meta, Microsoft, and Anthropic together extract $6,565 a year from each US internet user, $393,878 over a digital lifetime. Roughly the price of a house you'll never be paid for.

How the "free" bargain actually clears

The pitch has always been the same. Search is free, email is free, the map in your pocket is free. You don't pay because you're not the customer. Your attention and your behavior are the product being sold to someone who does pay. Everyone sort of knows this. What nobody could quantify, until now, was the rate.

The report calls its metric Personal Data Annual Value, or PDAV. The analysts screened 150 companies and kept 129 that qualified as "Big Data" firms under a seven-criterion test: 65 AI-first subscription and API companies, 31 ad-funded platforms, 14 hardware-bundled companies, 19 data brokers. The point of casting that wide was to capture revenue streams that older estimates left out. When people talk about what Google "makes off you," they usually mean ads. The PDAV framework adds the Copilot license your employer pays for, the API calls a startup makes against a model trained on your writing, the resale of your profile through a broker you've never heard of. Stack all of it and you get $6,565 a year for the central US scenario.

The report also runs conservative and expansive versions, landing at $4,816 and $8,531 a year, so the $6,565 sits in the middle of a documented range rather than at the top of one. And the five companies named (Amazon, Alphabet, Meta, Microsoft, and Anthropic) aren't splitting that figure evenly. Each one, the Foundation found, earns up to $1,000 a year off a single American. The total isn't the sum of one company's take. It's what happens when all of them run on the same person at the same time, every hour you're awake and using a screen.

So here's the thesis, plainly. The free internet is a payment system. You pay in data, the company captures the value, and you have no visibility into the transaction, no bargaining power over the price, and no share of the upside. That asymmetry is the design, not a side effect of it. You can't negotiate a rate you were never shown.

The receipts the companies filed themselves

The Web3 Foundation has a stake in its own conclusion (more on that below), so I went looking for the same order of magnitude in documents the companies can't walk back. They're there.

Start with Meta's own 2023 annual report. The 10-K reports average revenue per user, and the geographic split is the tell: US and Canada ARPU runs "more than 11 times higher" than Asia-Pacific. The final quarterly figure Meta disclosed for US and Canada (Facebook and Messenger only, not Instagram or WhatsApp) was $68.44 for Q4 2023. Sum the four quarters and a single American Facebook user generated somewhere around $213 to $220 that year, from one product line. Then, starting with its Q1 2024 filing, Meta stopped reporting geographic ARPU altogether and switched to a harder-to-parse company-wide metric. The number that let analysts price US users disappeared from the filings right as it was climbing.

Alphabet's scale is in its Q4 2025 earnings release: $402.836 billion in full-year revenue, up 15%, the first year past $400 billion. The company doesn't break that out per US user, but Proton's independent analysis does the arithmetic from public filings and lands near $460 a year per American from Google alone. Combine Google and Meta on advertising by itself and you're already at roughly $700 a year per US user, before AI, before enterprise licensing, before brokerage. That's the floor the PDAV number builds on top of. The gap between that $700 and the full $6,565 is the part the report is really arguing about: the value of your data once it stops being just an ad-targeting input and becomes training material, enterprise product, and resold profile. That's also the part hardest to verify independently.

Then there's the government's own finding. In September 2024 the FTC voted 5-0 to release a staff report concluding that major platforms "engaged in vast surveillance of consumers in order to monetize their personal information." Chair Lina Khan's summary: these companies "harvest an enormous amount of Americans' personal data and monetize it to the tune of billions of dollars a year." Bipartisan, unanimous. The same inquiry found that when the FTC asked the nine firms to name every advertiser, broker, and entity they shared user data with, not one could produce a complete list. They monetize data they can't fully account for, and they're under no obligation to tell you your share of it.

Who benefits, and how the money moves

The beneficiaries are the five companies, and the incentive is exactly what it looks like: money. Alphabet pulled $402.836 billion in 2025. Meta reported $200.97 billion, with $196.175 billion of it from advertising. Amazon's ad business alone hit $56 billion in 2024, up 18% in a year, and it now runs ads inside Prime Video for 200-million-plus subscribers. None of that is hidden. It's in the filings.

What's hidden is the per-person rate, and that's the part worth sitting with. The information asymmetry is the business, not a byproduct of it. A user who knew their data was worth $6,565 a year might ask for a cut, or shop the rate, or simply use the service less. A user who's been told the service is "free" does none of those things. Meta dropping geographic ARPU from its SEC filings in 2024 isn't a footnote. It removed the single public number that let outsiders estimate what a US user was worth, right as regulators and journalists were starting to do that math. Keeping the rate opaque is how the rate keeps working.

There's a second beneficiary worth naming, and it's the report's own publisher. The Web3 Foundation gains from every regulator and reader who walks away convinced that centralized platforms are extracting a fortune, because the Foundation sells the alternative. That doesn't make the extraction figure wrong. It does mean two parties profit from you believing it: the platforms who'd rather you never saw the number, and the crypto ecosystem that wants you angry about it.

What this misses, and where it overreaches

The $6,565 figure comes with real caveats, and they're worth walking through.

First, the source. The Web3 Foundation is not a neutral research institution. Its founder, Gavin Wood, holds an estimated $400–900 million in DOT, the token behind Polkadot, the blockchain network his company Parity Technologies built. The report's recommended fix is Web3-style decentralized infrastructure, the exact category Wood's fortune is tied to. And it landed 24 days before the Web3 Summit in Berlin, the Foundation's own marquee event. Read the report as an advocacy document with a financial stake in its conclusion, because that's what it is. The PDAV model is the Foundation's proprietary framework, not peer-reviewed, and the precise method for attributing diffuse revenue (enterprise licenses, AI training value) to a single person is where the number is softest. The reason it survives the conflict is that the FTC's bipartisan finding and Meta's own SEC filings corroborate the order of magnitude independently.

Second, the lifetime figure deserves a flag. The report's $393,878 lifetime number is roughly the median US home sale price of $403,200, a clean and real comparison. The Foundation also headlines an $831,497 version, but that one applies inflation-linking across a 61-year digital lifetime, which inflates the figure substantially. The annual $6,565 and the $393,878 lifetime number are the ones to anchor on.

Third, and this is the strongest objection: you get a lot back. MIT economist Erik Brynjolfsson's survey research, cited by Fed Chair Jerome Powell in 2019, found the median American would demand $17,530 a year to give up search engines, plus $1,173 for video streaming and $576 to give up Facebook. If those numbers are anywhere close, the consumer surplus dwarfs the extraction, and for plenty of people the trade is net-positive. The argument here was never that Big Tech delivers nothing. It's that you have zero visibility into what your end of the deal generates. And the surplus isn't evenly distributed. A lower-income user runs fewer premium services, collects less of that $17,530 in value, and generates roughly the same extraction. The asymmetry bites hardest on the people least able to negotiate it.

The Bottom Line

The companies can tell you, to the dollar, what they spent on safety. Mark Zuckerberg told the Senate in January 2024 that Meta had invested "over $20 billion since 2016" protecting users. No executive has ever stated, under oath or in a filing, what a single user's data is worth to the company collecting it. The number exists. It's just never pointed at you.

And no law makes them point it at you. The American Privacy Rights Act died in Congress in 2024. The SECURE Data Act, released in April 2026, would let you opt out of targeted advertising and data sales, which is progress. It still doesn't require any company to disclose the commercial value it generates from you. So the open question isn't whether your data is worth thousands a year. The filings and the FTC settled that. It's whether you'll ever get to see the invoice for a transaction you've been part of your whole life.