You'll Own Nothing and You'll Be Miserable
How the Promise of AI Abundance Became a License Economy — and What to Do About It
A sentence that gave away the plot
On March 31, 2024, millions of people tried to launch a video game they had bought — some of them ten years earlier, at full price, on physical discs in shrink-wrapped boxes — and discovered it no longer existed.
The game was Ubisoft’s The Crew. It was a commercial hit; over two million copies sold in its first month, successful enough to spawn two sequels. Ubisoft stopped selling it in late 2023, then switched off the servers it required on March 31, 2024. Because the game checked in with those servers to run, every copy — online, offline, purchased, preordered, collector’s edition — became a small, polite error message.
When two Californians sued in late 2024, arguing that “bought” is supposed to mean bought, Ubisoft’s legal response contained a sentence that belongs carved above the entrance of every digital storefront on Earth. According to reporting on the filings, Ubisoft’s lawyers argued that the plaintiffs never actually purchased the game. What they had acquired was a limited license to access it. The game’s packaging, they pointed out, said so — in capital letters.
When France’s largest consumer protection organization, UFC-Que Choisir, filed its own suit in March 2026, its position was that revoking “licenses” this way violates consumer rights. In the interim, a grassroots movement called Stop Killing Games had collected over 1.2 million verified EU signatures and reached the European Parliament. California passed a law requiring digital storefronts to state clearly, before checkout, that “buy” means “license.” Shortly afterward, Steam began displaying exactly that disclaimer.
One small game. One sentence in a legal filing. And suddenly a decade’s worth of quiet inversion was visible in broad daylight.
We are, slowly and then all at once, being evicted from the ownership economy into the license economy. And the most remarkable thing about this transition is how neatly it inverts the most-repeated promise of the AI era — that abundant machine intelligence would liberate us from toil, not rent us back the conditions of our own lives.
The pattern you’ve been trained not to see
Once you notice it, the pattern is everywhere:
Your tractor. John Deere, the single largest maker of farm equipment in North America, has spent years locking farmers out of their own machines through software. The full diagnostic tool needed to repair a modern Deere tractor is available only to authorized dealers. A stripped-down version for farmers costs about $3,000 per year and still redacts the key functions. In January 2025, the FTC and several states sued Deere over these practices. In April 2026, Deere settled a parallel class action for $99 million and agreed to make diagnostic tools available for at least a decade. Public Interest Research Group estimates that repair restrictions across all manufacturers cost U.S. farmers about $4.2 billion a year.
Your car. In 2022, BMW began charging owners roughly $18 a month to activate heated seats — hardware physically installed in every car the moment it left the factory. The backlash was severe enough that BMW dropped that particular subscription in 2023. But the broader model didn’t go away; it retreated upmarket. Tesla now sells Full Self-Driving only as a monthly subscription. Mercedes charges annual fees to unlock acceleration in EQ electric cars. Volkswagen sells horsepower by the month. The car you paid $60,000 for is, increasingly, a device that has to phone home to deliver the performance you thought you bought.
Your printer. HP pushes firmware updates that disable third-party ink cartridges and, by some accounts, even expired first-party ones — a practice so extensively litigated it has its own Wikipedia entry.
Your books. In 2009, in a moment that should be taught in every business school as an omen, Amazon remotely deleted copies of George Orwell’s 1984 from Kindle devices over a licensing dispute. Customers woke up to find pages they’d highlighted the night before no longer existed.
Your software. Adobe killed perpetual Photoshop licenses in 2013. Microsoft Office became Microsoft 365. Autodesk, Jetbrains, 1Password — all of it moved from one-time purchase to perpetual rent, often alongside genuine product improvements that made the rental feel briefly reasonable.
Your music and movies. The “Buy” button on most streaming platforms is doing heroic work of imagination. Songs and films vanish from personal libraries when label contracts change. Movies you “own” on iTunes disappear when regional rights expire.
Your smart home. Revolv, a popular home-automation hub Google acquired through Nest, was simply bricked in 2016 when Google decided to stop running the servers. Every device turned into a paperweight. This happens quietly and routinely — it is now a risk category for any internet-connected appliance.
This isn’t a conspiracy. It’s a business model, compounded by network effects, venture-scale return expectations, and the quiet architectural fact that almost everything now needs a server somewhere to work. The software industry learned that recurring revenue is worth roughly five to ten times transactional revenue in enterprise valuations. Then every other industry noticed.
What we are watching is the financialization of possession. Things that used to be capital — tractors, cars, books, tools — are being refactored into services. This is wonderful for the balance sheets of a small number of platforms. It is, for everyone else, the slow disappearance of a form of wealth they may not realize they had.
What we traded away
The old ownership economy was never perfect, and it certainly wasn’t equally distributed. But it had a load-bearing property that is easy to miss when you have it and brutal to notice when it’s gone: owned things compound into stability.
A house you own is a hedge against rent increases. A car you own is a hedge against subscription fees. A library of books, records, tools, and skills you own is a hedge against a vendor you will never meet deciding the server isn’t worth running anymore. Ownership is, among other things, a form of insurance against the future’s moods.
The twentieth-century political project in most democracies — whatever you think of its execution — was broadly about spreading ownership. Homesteading. Public libraries that lent genuine copies of genuine books. GI Bills and mortgage guarantees. Pensions that were actual funds, not just promises. Small business formation as an ordinary, expected life path. The cultural default was that your work produced accumulating artifacts: a house, tools, savings, a trade, a business, a deed.
The twenty-first-century version is different. You don’t own the music you listen to, the shows you watch, the software you write in, the car you drive, increasingly the tractor you farm with, and almost none of the platforms you rely on to make a living. Much of this is fine in isolation — who really wants to rack their own servers for Netflix. It becomes strange when you add it up. Most of the financial flows that used to build household balance sheets now flow laterally into platform balance sheets as rent. The median American household pays more per month in subscriptions than at any point in history, and that figure continues climbing.
This is the decor of the room. Now look at the wallpaper.
The inversion: what AI was supposed to be, and what it is becoming
The pitch for artificial intelligence — the real pitch, the one in the blog posts and keynotes, not the carefully hedged ones in the 10-Ks — was abundance.
In 2016, Sam Altman wrote on the Y Combinator blog about funding a basic-income experiment because he thought technology might eliminate enough “traditional jobs” that societies would need to rethink how people receive income at all. Elon Musk has said repeatedly that in a benign AI future, “probably none of us will have a job” and that there would be “universal high income.” Even corporate communications from the major labs lean on the same frame: AI will produce so much economic output, so cheaply, that humanity’s relationship to scarcity itself will change.
The implicit deal was this: yes, the transition will be disruptive, possibly brutal for particular workers and industries, but on the other side lies something like a utopia of near-free cognition. UBI, in this framing, is a bridge to that utopia — a cash transfer to keep people whole while the economy reorganizes around machines.
The OpenAI-backed basic-income study, run by OpenResearch, released its findings in July 2024. Three thousand low-income participants received $1,000 per month for three years. The results were interesting but not euphoric: recipients worked about 1.3 hours less per week, spent more on food, rent, transportation, and helping family members, used a bit more healthcare, and made more entrepreneurial moves. They did not, contrary to fears, stop working. They also did not, contrary to hopes, find dramatically better jobs. Short-term mental-health gains faded after the first year.
UBI, it turns out, helps. It is not magic.
Meanwhile the labor market numbers have started to turn. Anthropic’s CEO Dario Amodei predicted in 2025 that AI could eliminate roughly half of entry-level white-collar positions within five years. Cornell University research found U.S. companies adopting AI have reduced junior hiring by about 13 percent. A Stanford study in 2025 provided early large-scale evidence that the effect lands hardest on workers in early-career cognitive roles. Employment of software developers aged 22 to 25 has fallen roughly 20 percent from its late-pandemic peak. In the first six months of 2025 alone, roughly 78,000 tech layoffs were attributed to AI; modeling estimates for the full year suggest the real, unreported number is several times higher.
Now sit with the pieces next to each other:
- The productive capital of the coming economy — training compute, model weights, inference capacity, proprietary data — is being concentrated into a handful of companies, more concentrated than any industrial revolution that preceded it.
- That same handful of companies owns, or is rapidly acquiring, the platforms on which the rest of us rent access to everything: documents, storage, software, media, communication, search, and now thinking.
- The compensation framework being proposed for the disrupted majority is a subsidy — potentially funded by taxing the AI companies themselves — that would flow through to them again anyway, because we rent our lives from them.
A sharp 2025 paper in Frontiers in Artificial Intelligence by Jean-Christophe Bélisle-Pipon frames this as a form of “symbolic violence” — a UBI advocated by AI elites risks legitimizing, rather than challenging, a world divided between AI owners, a narrow class of AI-capable workers, and a vast population of recipients. Whatever you think of the framing, the structural observation is hard to argue with. A basic income that keeps you housed and fed while the tools of production are owned elsewhere is not emancipation. It is a subscription to existence, paid by the very platforms that collect rent on everything else you do.
This is the inversion. The promise was that abundant intelligence would free us from toil. What is actually being built is a world in which even the refusal to pay rent — on your software, your car, your data, your tractor, your attention — becomes progressively more difficult.
The good news is that this is not a law of physics. It’s a set of choices. Which means there are counter-choices.
Why the arithmetic gets worse the longer you wait
Three dynamics make the license economy compound against you over time.
Rents are multiplicative, not additive. Every subscription is forever, and every forever is priced into the NPV of every platform that adds one. A $15 monthly subscription is a $1,800 ten-year commitment at zero inflation — and none of these prices stay flat. Netflix, Adobe, Microsoft 365, YouTube Premium, Spotify, ChatGPT Plus: every one of them has raised prices at or above inflation since launch. Meanwhile, the transaction costs of canceling are engineered upward.
Skills and knowledge atrophy in a rented world. When nobody owns, nobody maintains. When nobody maintains, nobody learns. Repair knowledge, sysadmin knowledge, the quiet competence of running your own things — these were dense commons fifty years ago. They are thinning now. The license economy is, among other things, a deskilling engine, because the incentive structure rewards platforms that keep complexity on their side of the API. This is especially true with AI tools: a generation of developers who only know how to call cloud APIs are, in an important sense, not portable across the end of a vendor relationship.
Political leverage follows ownership. Renters, historically, have less political voice than owners. The patient constituency for interoperability, repair, portability, and competition is built out of people who have something concrete to defend. Every subscription is a small vote for the current arrangement. Every piece of local infrastructure — a laptop you can repair, a server in your closet, a backup in a drawer, a skill in your head — is a small vote for the next one.
Fighting back: the personal level
The goal at the personal level isn’t to exit the modern economy. It’s to be strategically ownership-heavy on the things that matter to you, so that no single vendor decision can wipe out a meaningful part of your life.
Own your data. Pick a local-first architecture and stick with it. Practical stack: a NAS or small server running Nextcloud for files, Immich for photos (a genuinely compelling replacement for Google Photos), Paperless-ngx for documents, Jellyfin for media. Syncthing for device-to-device sync without a central server. Back up with the 3-2-1 rule: three copies, two media, one offsite. None of this is as convenient as the cloud. All of it is permanent in a way the cloud is not.
Own your devices. Prefer repairable hardware. Framework laptops are the cleanest expression of this ethic. Phones are harder, but a Pixel running GrapheneOS remains the best combination of privacy, security, and longevity. Run Linux where you can — Fedora, NixOS, or Ubuntu — not because Windows is bad but because the day Microsoft decides Windows is a service, you want other options.
Own your compute. This is newly important. Local LLMs have crossed the threshold from curiosity to workhorse. Ollama and LM Studio make it trivial to run capable models — Llama 3, Qwen, Mistral, Gemma — on an M-series Mac or a modest GPU. For the majority of day-to-day tasks (drafting, summarization, code assistance, research synthesis), a good local model is sufficient, free, private, and immune to service outages or policy changes. Pay for frontier models when the task genuinely demands it, but stop letting every keystroke flow through a third party’s context window.
Own your media. Buy DRM-free where possible: ebooks from Kobo or Standard Ebooks, audiobooks from Libro.fm, music from Bandcamp or as actual FLAC files. Keep a local library. Streaming is fine for discovery; the things you’ll still want in a decade deserve to live on your drive.
Own your information diet. RSS readers (try Miniflux or FreshRSS) instead of algorithmic feeds. Download the PDF, don’t just bookmark the article. Snapshot what matters with a tool like ArchiveBox.
Own your skills. The single highest-leverage personal investment in an ownership-hostile world is literacy in the tools that let you route around vendor decisions: shell, git, Docker, a scripting language, basic networking, basic electronics. You don’t have to become an engineer. You have to become hard to hold hostage.
Use AI as your leverage, not your landlord. The irony worth naming out loud: the same technology reshaping the labor market is also the single biggest boost to the ownership path that has ever existed. Every obstacle that kept self-hosting niche — cryptic config files, opaque error messages, the sysadmin tax of running your own things — is exactly the kind of friction LLMs dissolve. A local model on your laptop is a patient, always-available tutor that will write your Nextcloud docker-compose.yml, debug a failing Jellyfin transcode, explain what that nginx log line means, and script your Google Takeout into Immich. The historical asymmetry — “proprietary is easier because somebody else absorbs the complexity” — is inverting. With AI on your side of the fence, the complexity gets absorbed there instead. The license economy handed you a weapon; pick it up.
Fighting back: the SMB level
Small and medium businesses are where the license economy extracts the most, per capita, and where the freedom to leave is most valuable — and most neglected.
Own your customer relationships. The email list is the single most important business asset most founders under-invest in. Not your Instagram followers, not your TikTok reach, not your Shopify traffic — an owned list of customers and their consent, in a format you can export and re-import anywhere. Platforms are weather; an email list is climate.
Own your operational core. Self-host the handful of systems that would hurt most to lose. Practical picks: PostgreSQL (the most undervalued piece of software of the last two decades), Supabase self-hosted as a Firebase replacement, Mailcow or Mail-in-a-Box for email, Plausible or Umami instead of Google Analytics, n8n or Activepieces instead of Zapier for automation. An open-source CRM like EspoCRM or SuiteCRM will outlast any startup.
Use open formats religiously. CSV for tabular data, Parquet for analytics, Markdown for documents, plain SQL for queries, open API standards for everything. The test is simple: if this vendor disappeared tomorrow, could you read your data? If not, you don’t own it.
Contract for portability. Write data-export clauses and reasonable exit-assistance clauses into every vendor relationship. For critical software, ask about source code escrow. This sounds like enterprise paranoia; it is practical hygiene.
Host your own AI. For internal workflows — customer support triage, document processing, code assistance, internal search — a self-hosted model on your own hardware is frequently sufficient and keeps your operational data out of someone else’s training set. vLLM or Ollama on a single well-specced GPU server handles enormous volume. The break-even against API costs usually arrives faster than people expect.
Consider a homelab. A used 1U server from an off-lease liquidator, Proxmox on top, and a handful of containers can replace a surprising amount of SaaS for a small business. The skill involved is a moat, not a cost.
Let AI finally make “build” competitive with “buy.” For a generation, the default answer for a small business was rent the SaaS — because the developer needed to build a custom equivalent was too expensive, and the open-source alternatives lost on polish. AI collapses both sides of that equation. A founder or operations lead with a capable coding assistant can stand up a focused internal tool over a weekend that replaces two or three per-seat subscriptions, and maintain it without a full-time engineer. Open-source projects that used to lose on documentation and UX become usable because an LLM fills the gap in real time. Migration projects that used to take a quarter — exporting from one vendor, transforming the data, loading into a self-hosted replacement — take a week. The break-even between renting and owning has moved significantly in favor of ownership, if you use AI as a force multiplier for your own infrastructure rather than as another monthly bill.
Fighting back: the corporate level
Large enterprises have the buying power to bend markets, and the organizational depth to build their own alternatives. Most squander both.
Procurement as policy. Write real terms into SaaS contracts: data portability on demand, source-code escrow for anything mission-critical, SLA-backed API stability, open-standard interoperability (OIDC, SCIM, OpenTelemetry, OCI, ODBC, standard SQL dialects). Refuse contracts that penalize leaving. A hundred enterprise buyers doing this changes vendor behavior in a way a hundred consumer complaints never will.
Build what differentiates; rent what doesn’t. The classic framing, now more important than ever. If a capability is core to your strategic advantage — your models, your pricing engine, your customer intelligence — owning it outright is worth substantial overhead. If it isn’t, fine, rent it, but rent interchangeably. Multi-cloud by architecture, not by slogan.
On-prem and hybrid AI for anything sensitive. The large open-weight model families — Llama, Qwen, Mistral, DeepSeek — are good enough for the overwhelming majority of enterprise tasks when deployed with proper retrieval and guardrails. The capex to run them on your own iron is a rounding error compared to the annual contract spend most large firms now allocate to AI APIs. The returning benefit — your data does not leave your premises, ever — is increasingly a regulatory and strategic necessity.
Fund the commons. Large enterprises are the largest beneficiaries of open-source infrastructure (Linux, Postgres, Kubernetes, Python, Go, the HTTP stack) and the least proportionate contributors. Allocate engineers and dollars, systematically, to the projects you depend on. This is not philanthropy; it is securing the supply chain for your own foundation.
Cultivate depth, not just breadth. Specialists who understand how systems actually work — who can read Postgres query plans, debug network stacks, run their own Kubernetes, tune their own models — are the people who make an organization ownership-capable. A staff of vendor-certified specialists is an organization that has outsourced its capacity to leave.
Back the policy fights. Right to repair, digital markets regulation, interoperability mandates, open-model-weights policy, data-portability standards — these fights determine the shape of the next decade’s commercial terrain. Corporate policy teams that quietly support them buy their employers optionality; corporate teams that quietly oppose them are engineering their employers’ dependence.
Treat AI as an asset, not a service. The difference between AI that compounds your position and AI that erodes it comes down to a single question: at the end of the contract term, do you own more, or less? A frontier API call produces an answer and a line item. A fine-tuned open-weight model, trained on years of your proprietary data, retrieval-grounded in your own documents, deployed on infrastructure you control, produces a durable capability your competitors cannot rent from the same shelf. AI-assisted code modernization has quietly transformed the economics of paying down technical debt — the legacy systems that used to lock you into a vendor can actually get refactored now, at a tenth of the historical cost. Internal agents running against your own data, on your own infrastructure, with your own models, are strategic leverage. The same agents running against a third-party API are a forecast line item with usage-based pricing and a renegotiation date. The question for any serious enterprise is no longer whether to use AI, but whether the AI they use compounds into assets they own or costs they pay forever.
Ownership, deliberately
Here is the uncomfortable summary. Ownership used to be the default outcome of working, saving, and buying. In 2026, ownership is a deliberate, slightly effortful act. You have to notice it. You have to choose it. You have to maintain it. The frictionless default is to rent everything from a platform, including, eventually, your own cognition.
But the arithmetic still works. Every subscription canceled is a rent not paid forever. Every self-hosted service is a middleman subtracted. Every owned device, skill, model, dataset, relationship, and habit is a piece of the next economy that somebody — you, your business, your community — actually controls.
The abundance promise of AI could still be real; the only live question is whose balance sheet it lands on. Abundance built on top of a stack owned by five trillion-dollar companies, paid for by a universal subsidy routed through their balance sheets, is not abundance. It is a company town with a better marketing budget.
But abundance can also go the other way. AI is, at the same moment the license economy is using it to dismantle ownership, the single largest force multiplier for ownership that any individual, small business, or enterprise has ever had access to. The friction that used to keep self-hosting a hobby dissolves in seconds under a local model. The custom tool a small business couldn’t afford to build now takes a weekend. The vendor-locked legacy system a corporation resigned itself to paying rent on can actually be refactored, at a fraction of the historical cost. The same technology being used to dismantle the ownership economy is, on the other side of the fence, the most effective tool yet invented for reclaiming it.
The alternative to the license economy is not austerity or nostalgia. It is a deliberate ownership posture — layered into personal habits, small-business architecture, and corporate procurement — with an occasional, well-placed lawsuit at the top, for tone. Ubisoft’s engineers flipping off a switch in March 2024 was a gift. For the first time in a long time, millions of people saw, plainly, that the thing they thought they owned had been on loan the whole time.
That clarity is the beginning of the counter-move. The license economy is legible now. And the tools to build something different — cheap hardware, capable open-weight models, mature open-source stacks, a whole generation fed up enough to start caring, and, improbably, an AI that will patiently explain any of it to you at two in the morning — are sitting right there, on your desk, waiting for someone to take them seriously.
You’ll own nothing and you’ll be miserable — if you let someone else decide that for you. You don’t need permission to start owning things again. You just need to notice that it’s a choice, and that for the first time in a long time, the asymmetry runs in your favor.