Do you want to give it all away, or keep enough to keep building?
I keep returning to this question because it exposes a mistake that appears in business, technology, philanthropy, and public life. We often treat generosity and sustainability as moral opposites. Giving everything away appears pure. Retaining money, control, or advantage appears compromised.
But service is not only an intention. It is a system.
A system that cannot fund itself cannot keep serving people.
That principle is not a defense of hoarding. It is a defense of continuity. Money kept without purpose becomes extraction. Capacity preserved for the next round of useful work becomes stewardship. The difficult task is to tell the difference.
Generosity can become liquidation
There is a form of generosity that looks noble because its costs arrive later.
A founder underprices a product until every customer loves the deal and the company disappears. A maintainer gives away essential software while support requests consume the time needed to improve it. A nonprofit expands a program without funding the people and infrastructure required to sustain it. A talented person says yes to everyone until there is no attention left for the work that made the help possible.
In each case, the system converts its productive base into a short burst of goodwill. It serves more people today by sacrificing its ability to serve them tomorrow.
The opposite error is just as real. Organizations can invoke sustainability to justify permanent accumulation, weak accountability, or prices detached from value. “We need the margin” can become a respectable way to avoid asking who benefits from the margin.
So the principle is not “keep more.” It is keep enough: enough to maintain the system, improve it, survive setbacks, preserve independence, and continue the mission without requiring endless sacrifice from the people inside it.
Properly designed, margin is stored service.
Tesla had the current. Westinghouse built the circuit.
The popular version of the war of the currents is emotionally satisfying. Thomas Edison becomes the commercial brute defending direct current. Nikola Tesla becomes the generous visionary whose alternating-current system deserved to win. History then becomes a contest between business and brilliance.
The real lesson is more interesting because it is less tidy.
Edison did not merely develop a better bulb. His team worked on a complete electrical system: generators, wiring, switches, meters, manufacturing, and a commercial station. The National Park Service describes the Menlo Park laboratory as an early model for modern research and development, and notes that the capital required to build the electric-lighting industry brought powerful financiers into the project. Edison understood something inventors often resist: a discovery reaches ordinary people only after it becomes an operating system.
His direct-current network had serious limitations. It could not economically transmit power as far as alternating current, and early central stations were expensive. The Smithsonian’s account of the Pearl Street station notes that it took roughly five years to become profitable. Still, Pearl Street mattered. It turned electric light from a laboratory achievement into a service with customers, meters, maintenance, and lessons that could inform the next installation.
Tesla supplied a crucial part of the better technical path. George Westinghouse supplied the commercial machine around it. In 1888, Westinghouse obtained rights to Tesla’s polyphase alternating-current patents and brought Tesla into his company, according to the Library of Congress. Westinghouse had manufacturing capability, engineers, capital, contracts, and an organization able to turn a superior approach to transmission into infrastructure.
Tesla’s ideas mattered enormously. So did the institution capable of deploying them.
The lesson is not that Edison was right, nor that commerce deserves the credit for invention. It is that civilization received the benefit only when insight entered a circuit that could finance equipment, coordinate labor, absorb failure, collect revenue, and build again.
An electrical system must close the circuit. An economic system must do the same.
Open AI still needs an economic engine
The same tension now appears in artificial intelligence.
Open systems offer real public value. They let researchers inspect components, developers adapt tools to local needs, organizations run models on infrastructure they control, and communities improve shared foundations. The Open Source Initiative’s definition of open-source AI centers four freedoms: to use, study, modify, and share a system. Those freedoms make knowledge more portable and power less concentrated.
But openness is an access model, not a funding model.
Training models, running evaluations, serving inference, responding to security problems, maintaining documentation, and supporting users all consume money and skilled labor. Stanford’s AI Index reports rapidly rising estimated training costs for prominent frontier models. Even when model weights can be downloaded at no charge, someone still pays for the hardware, energy, storage, engineering, and ongoing operation.
Closed systems solve part of that problem by charging for access. Revenue can support research, dependable APIs, security work, customer support, and the next generation of models. Yet closure creates its own risks: dependency on one provider, limited inspectability, weaker user control, and the possibility that a valuable layer of general-purpose knowledge becomes governed by a narrow commercial interest.
This is often presented as a choice between open and closed AI. I think that is the wrong unit of analysis. The better question is: which layer should be open, and which paid layer can keep the whole system healthy without enclosing the commons?
A laboratory can publish research, release selected models or weights, support open standards, and still charge for managed products that provide reliability, scale, integrations, and service. OpenAI’s release of gpt-oss open-weight models alongside its hosted products is one current example of that hybrid structure. It is not proof that any particular balance is correct. It shows that access and revenue do not have to be decided with a single switch.
Open what benefits from broad inspection, adaptation, and shared improvement. Charge for scarce resources and differentiated service. Make portability real enough that the paid layer earns continued trust rather than manufacturing dependence.
The objective is not to give away the entire power plant. It is to keep the grid generative.
Wikimedia charges for delivery, not knowledge
Wikimedia offers an especially clear third example because its public mission is explicit: knowledge should remain freely available.
Wikipedia cannot fulfill that mission without servers, staff, legal defense, security, fundraising, and support for the volunteer communities that create and maintain its content. Reader donations remain central, but the Wikimedia Foundation has also developed Wikimedia Enterprise, a paid service for organizations that need high-volume, dependable access to Wikimedia content.
The distinction is important. The knowledge remains freely licensed. Existing public access remains available. Enterprise customers pay for a service layer designed around scale, reliability, and support. Wikimedia’s 2026–2027 financial plan describes Enterprise as a complementary revenue stream through which high-volume commercial reusers contribute to the infrastructure they depend on. Its operating principles state that the service should be self-funding and that its revenue supports the Wikimedia mission.
That is a thoughtful boundary. Do not sell the public good back to the public. Do charge sophisticated commercial users for an enhanced delivery system, then return that revenue to the public good.
The free layer creates reach, trust, contribution, and shared value. The paid layer converts a portion of that value into durability. Neither layer makes sense for long without the other.
Design the return path
Every mission-driven system needs a return path: some mechanism through which value delivered today replenishes the capacity required to deliver value tomorrow.
That mechanism does not always have to be a price. It can be donations, subscriptions, memberships, paid support, enterprise hosting, licensing, grants, an endowment, or a portfolio in which profitable work funds public work. What matters is that it is reliable enough to support the obligation being created.
When designing that return path, I would ask five questions:
- What must remain broadly accessible? Protect the part of the system whose openness creates the public benefit.
- Where does a user receive distinct, scarce value? Reliability, speed, customization, support, and reduced operational burden are legitimate things to sell.
- Does revenue return to the source of value? The people, infrastructure, and knowledge commons producing the benefit should become stronger as usage grows.
- Can people leave? Sustainable revenue should come from continued usefulness, not artificial captivity.
- What does “enough” mean? Define the reserves, reinvestment, and resilience the mission actually requires so sustainability does not become an excuse for accumulation without end.
This is the economic version of guarding the root. Fruit is meant to be shared. But stripping the tree to create the appearance of abundance destroys the next harvest.
Keep the ability to return
The strongest systems are not the ones that give the most in a single moment. They are the ones that can return, improve, and give again.
Tesla’s inventions needed Westinghouse’s capacity to deploy them. AI openness needs institutions capable of paying for compute, maintenance, and safety. Wikipedia’s free knowledge needs a revenue architecture that can support the infrastructure beneath it.
This does not make profit sacred. It makes continuity a design responsibility.
The mature choice is not between giving everything away and keeping everything for yourself. It is to decide what should be shared, what must be preserved, and how the value moving outward will send enough energy back through the system to keep it alive.
Give generously. Charge honestly. Reinvest visibly. Keep enough to keep building.