He Dropped Out Making $750K/Yr… Now He’s an AI Billionaire on Paper | David Stout

David Stout said WebAI is publicly valued at about $2.5 billion and that he owns roughly 50% of the company. That implies a paper stake of about $1.25 billion, which is why the hosts described him as a billionaire on paper.

Stout grew up on a ranch in Michigan with little exposure to the technology industry. His family heated its home with an outdoor wood stove, and he described cutting and storing enough wood each summer to survive Michigan winters as an early lesson in planning, persistence and problem-solving.

Starting a technology company seemed almost unimaginable where Stout grew up because he knew no one who had done it. He nevertheless sees farming itself as a business and believes the discipline of farm work prepared him for entrepreneurship.

Brockway Academy was a turning point because it gave Stout access to computers and introduced him to his future co-founder Ethan. The school let students learn at their own pace, while college was not initially a major priority for him.

Stout attended Davenport, a business school in Grand Rapids, Michigan, and launched a web-services business as a freshman with Ethan. They reached about 60 customers in their first year, doing backend web work before modern drag-and-drop website builders became common.

The college web-services business was generating more than $750,000 a year, according to Stout. Revenue was irregular, but he recalled the two college students occasionally receiving checks of roughly $60,000 in a single month.

Stout did not finish college because the business was already producing substantial income. His view was that continuing school made little sense while he was already making about $750,000 annually.

Entrepreneurship had been modeled inside his family even though technology had not. His father farmed, held sales jobs and bought a gas station after a motorcycle accident and coma prevented him from graduating high school; Stout’s maternal grandfather was an engineer who became a successful entrepreneur in his 60s and 70s.

Stout is the youngest in a split family, with one full brother and two half-brothers, and believes birth order influenced his personality. He described himself as attracted to difficult problems and unusually motivated to work on things other people were overlooking.

Around 2016, Stout became bored with building websites and began studying machine learning and machine intelligence. He and Ethan started adapting what they knew about software and web development to runtime AI systems, at a time when useful information on the subject was still scarce.

Stout described Ethan as an exceptionally capable engineer who could code almost anything. Their technical partnership grew naturally out of the earlier web-services company rather than beginning with WebAI.

Stout and Ethan had effectively decided in high school that they would build companies together. Stout said difficult family circumstances, poverty and shared hardship made their friendship more like a brotherhood, and he never seriously envisioned them not working together.

Stout’s parents supported his decision to leave formal education rather than insisting that he finish college. He believes their trust gave him an unusual advantage and says they assumed he could return to school if the venture failed.

His parents provided practical financial help despite not being affluent. When Stout overdrew his bank account while trying to acquire computing equipment, his father wired him money so he could keep working.

WebAI began with a $400 loan from Stout’s grandfather, used to buy a single computer that Stout and Ethan shared. The machine had an i3 processor, and they used it to experiment with running computer-vision models locally.

A central technical insight came from trying to process camera footage through distant data centers and finding the latency impractical. Moving huge quantities of video to the cloud and back convinced them that many AI applications would need computation close to where the data was generated.

That experience produced WebAI’s core early strategy: make AI run directly on phones, laptops and other devices. Stout said they deliberately reduced distractions and concentrated on on-device or “edge” AI, which he also saw as useful in places with weak connectivity such as rural areas.

Some investors initially considered edge AI too small a market. Stout recalled a meeting with Andreessen Horowitz in which someone in the room—not Marc Andreessen himself—said they did not think AI running on devices represented a sufficiently large opportunity.

Stout believes founder rejection can mean either that the problem is genuinely unimportant or that the market does not understand it yet. In retrospect, he puts WebAI’s original on-device thesis in the second category.

In 2017–2018, Stout and Ethan deliberately tried to prove the technology before raising institutional capital. At the time, he said even AI itself was still regarded by many people as speculative, while they were making the additional claim that it could work without cloud compute.

Getting their model working successfully on an i3 processor was the point at which Stout believed they had something sellable. He described it as finally having a real technological “SKU,” after which raising money to turn it into a product made sense.

Stout strongly prefers raising money after reducing technical and commercial risk rather than raising against an idea alone. His reasoning is that investors demand more ownership when they are taking nearly all the risk, whereas demonstrated execution lets founders preserve more equity and command a higher valuation.

Instead of relying primarily on a pitch deck, WebAI could show investors AI actually running on a laptop. Stout regards execution and delivery as much more valuable than the original idea because ideas themselves are abundant.

WebAI encountered extensive rejection while fundraising. Stout cited a rough rule of about 20 investor rejections for every successful funding outcome even for a venture-backable startup, and said founders therefore need unusually high tolerance for rejection.

He compared an admired investor rejecting a founder’s company to a parent tearing up a child’s drawing. At the same time, he said rejection can be useful because investors are not necessarily wrong given the evidence available at that moment, and their objections can improve a company’s direction.

Marc Benioff eventually became an investor in WebAI, according to Stout. He used that example to emphasize that WebAI was not rejected universally and ultimately attracted prominent backers.

Before he had institutional credibility, Stout pursued investors through cold email, repeated visits and even waiting in lobbies hoping to secure meetings. He portrayed access to major investors as something initially won through persistence rather than existing status.

Investor David Schuman became one of WebAI’s pivotal early supporters. Stout said Schuman discovered his work through a U.S. government/NSA report that identified him as a scientist doing interesting AI work and then traveled from New York to Michigan to meet him.

Stout described Schuman as a major early-stage investor whose investments included Oura Ring. Schuman reportedly told Bay Area investor friends that he was traveling to meet an AI scientist in Michigan, only to be told there were no good AI scientists in Michigan.

Stout said some of the investors who dismissed Michigan’s AI talent later invested in WebAI. He views Schuman’s early willingness to take him seriously as the kind of break that can change a founder’s trajectory.

Customer results became an important route to investment. Stout said successful customer deployments generated word of mouth among capital providers, and that once one strong investor joined, additional investor relationships became easier to compound.

His sales philosophy is therefore to lead with results rather than ideas. A founder cannot merely sell the concept; delivery itself has to become part of the sale.

Stout believes much of the long-term economic value in AI will move from general foundation models toward specialized models. His argument is that if every competitor can access the same general intelligence, that intelligence stops being a meaningful competitive advantage.

He argued that proprietary, specialized knowledge is what can differentiate companies in an AI economy. WebAI therefore works with information that may never appear on the public internet, such as specific techniques for packing chicken or operating a steel foundry.

Stout wants specialized AI to be excellent at its assigned domain rather than competent at everything. He illustrated this by saying an industrial model might understand a steel foundry extremely well while being unable to write a poem about frogs.

He distinguished WebAI’s approach from applications that place prompts or software layers on top of somebody else’s foundation model. The intended differentiation is the underlying specialized intelligence itself rather than simply different instructions to a common model.

Stout rejects the idea that the desirable end state is one universal intelligence shared by everyone. He believes markets and organizations create value through comparative advantage, specialization and different kinds of expertise rather than making everybody’s capabilities identical.

He outlined two broad AI futures: extremely capable general AI removes much of the need for employees, or individuals and companies increasingly own their own AI models instead of continuously renting intelligence. He considers the ownership model more likely.

Stout compared personal AI ownership with workers owning their own tools. His argument is that ownership creates independence because a person can carry that capability from one employer or situation to another rather than remaining tied to whoever supplies the tool.

He claimed Meta was spending about $50,000 per employee per year on AI usage or “tokens.” Stout doubts employers will indefinitely absorb costs at that level and expects pressure toward cheaper, locally owned intelligence.

Stout imagines personal AI models becoming durable assets that learn alongside their owners and can eventually be passed to their children. He compared the concept to an “iCloud of AI,” except the intelligence would reside under the user’s ownership rather than being switchable at the discretion of a service provider.

Stout believes centralized AI providers have a structural incentive to keep users’ data and intelligence flowing through their infrastructure. His alternative is AI that lives primarily on users’ own devices and collaborates with other specialized models when needed.

For people trying to become skilled in AI now, Stout’s advice is to use the available tools heavily while prices remain subsidized. He argued that roughly $20 consumer subscriptions currently provide far more compute and capability than their subscription price would normally justify.

At the same time, he recommends preserving and organizing the tacit knowledge that makes an individual or company distinctive. His expectation is that creating personalized models will become easier, making that accumulated expertise valuable training material rather than merely private know-how.

WebAI was preparing a consumer product designed to run locally on a MacBook or other personal machine. Its “persona” system would allow different models to access different knowledge bases and files, then automatically combine the appropriate models for the user’s particular task.

The product is built around selecting the right specialist intelligence for each problem instead of sending everything to one large general model. Stout said this should lower costs because computation happens locally while also producing more expert-level answers.

Stout expects personally owned AI models eventually to become income-producing assets. He described a future in which people or organizations create models based on their expertise and sell subscriptions or access to those models.

What worries Stout about AI is not the technology itself but the people and companies deciding how it is developed and controlled. He said technology is morally neutral compared with the decisions of the people steering it and argued that morality itself cannot simply be regulated.

He believes the economics of the current centralized AI industry will impose their own limits because leading companies are spending too much money. Stout sees capital markets and customers as mechanisms that may force accountability if those economics remain unsustainable.

Stout compared the right to own AI with a Second Amendment-style right to possess one’s own intelligence. He is especially concerned about intelligence becoming disproportionately controlled by one person or institution with enough political influence to determine how others may use it.

WebAI has grown from a two-person operation to roughly 200 employees. Despite the multibillion-dollar valuation cited earlier, the company remains relatively small in headcount.

Stout’s leadership philosophy centers on coachability, humility and servant leadership. He believes executives have to recognize their own egos and view leadership primarily as serving the people on their team.

He does not generally use individual one-on-one meetings at WebAI’s current scale and instead prefers group leadership. One key weekly meeting brings company leaders together for several hours during what he considers peak morning thinking time, roughly 9 a.m. through 11 a.m. or noon.

Stout relies heavily on people with specialist expertise rather than expecting himself to master every leadership discipline. He pointed to board member Dave Burn, described in the conversation as a founding Benchmark Capital partner with deep talent expertise, as the person he calls for questions about people and recruiting.

Stout said Dave Burn’s experience with Apple and Steve Jobs helped teach him how to identify exceptional talent. He credits that mentorship with improving his own ability to recruit senior people.

WebAI recruited Jason Rathg, described as the founder of the Pentagon’s Office of Strategic Capital, along with his team to lead WebAI’s defense operation. Stout credited Jason with rapidly expanding the company’s government business.

Stout said WebAI entered the government market less than a year earlier and had already become a prime contractor. He cautiously claimed its public-sector software team might be among the fastest-growing ever, explicitly acknowledging that the comparison could be wrong.

WebAI also recruited Dr. PJ Makish from Eric Schmidt’s SCSP to run its laboratory. Stout presented both that hire and the defense-team recruitment as evidence that WebAI can attract highly specialized people.

His recruiting approach relies heavily on consistency of belief and behavior over time. Stout believes talented people notice when a founder continues saying and doing the same things a year later, and that this credibility helps attract employees and board members.

Henry Ker has also acted as a mentor to Stout, particularly around servant leadership and faith. They were introduced through connections involving Stout’s Washington, D.C. PR firm, although Stout said they had not spent as much time together recently as he would like.

Christian faith became more central to Stout through his brother. He said his family cared about faith but had not consistently practiced it, while his brother made it tangible in his own life.

Stout said faith is uncommon in the technology industry and that some of his worst business experiences were with people who publicly advertised their faith. That has made him place more weight on living consistently and being “above reproach” than on publicly signaling belief.

He has deliberately sought mentors who will hold him accountable not only as a businessman but as a father and person. David Schuman is another mentor he named, while Henry Ker provides an example of how Stout would like to handle success.

WebAI’s upcoming consumer release was the product Stout said he was most excited about. He described the larger goal as making AI globally accessible while allowing the user—not WebAI or a cloud platform—to own the intelligence.

The consumer product was expected to launch in August and run entirely on the user’s device, including in airplane mode. Stout said users would not face conventional cloud-service usage limits or prompts telling them to wait until later or pay more to continue.

The WebAI application downloads a model directly onto the user’s machine. The model resides on disk, activates when the application opens and can operate without depending on a remote data center.

Users are intended to be able to build additional models inside the application for purposes such as content or marketing. Stout said those models could be shaped around expertise or ways of thinking that users value.

Stout described the product as fundamentally different from configurable GPT-style assistants because it creates actual models rather than merely prompting a shared underlying model to behave differently. He said users would also be able to share the models they create with others.

The consumer launch represents an entirely new market for WebAI. Stout said all of the company’s revenue up to that point had come from commercial customers or the public sector.

After moving to Austin, Stout initially needed office space for a team of roughly 20 people. He toured a rundown warehouse south of Congress that looked inexpensive but cost about $1 million a year, which he rejected as financially unjustifiable.

He instead found office space abandoned by a bankrupt company, complete with furniture, and negotiated a ten-year lease after discussing it with chairman David Schuman. The abandoned furniture itself was valuable because outfitting an office would otherwise have been expensive.

Stout negotiated an unusual provision giving WebAI naming rights and a building sign if it eventually occupied enough square footage. He suspects the landlord agreed because it seemed unlikely that the company would ever grow enough to trigger the clause—but it ultimately did.

He uses the building-sign story as an argument for making unusually ambitious asks. His reasoning is that people sometimes agree precisely because they believe the requested outcome will never occur, creating upside if the founder later makes it happen.

For ordinary businesses adopting AI, Stout recommends buying relatively inexpensive local computing hardware and moving some AI workloads onto equipment they own. His main reasons are avoiding token costs, rate limits and dependence on cloud infrastructure.

He specifically suggested equipment such as Apple Studio machines that can remain in an office and run AI tasks around the clock. For a small business, he sees this as a comparatively low-risk experiment because the main costs are the machine and electricity, while the hardware remains useful even if the AI experiment ends.

Stout predicted a major near-term shift toward local AI infrastructure and warned that suitable hardware could become difficult to obtain as enterprise demand rises. He said he would secure hardware early rather than wait for that migration to accelerate.

He defined a data center as essentially a computer the size of a building: centralized compute requiring enormous amounts of cooling, electricity and power. When someone asks a cloud AI system a question, that processing occurs on machines inside infrastructure of this kind.

One host described how his mother’s roughly 40-acre workplace at a garden center was acquired for about $160 million so the land could be redeveloped as a data center. The discussion used Northern Virginia as an example of how aggressively such facilities are consuming land.

Stout views the current data-center buildout as a potentially poor long-term infrastructure bet. He objected particularly to facilities consuming farms and other valuable land when the underlying computing technology may become dramatically smaller and more efficient.

He compared today’s giant data centers with early computers that occupied whole rooms. His argument is that the present physical scale of computation may be temporary rather than an enduring requirement of AI.

Stout also compared the data-center boom with railroad companies investing heavily in rail before aviation changed long-distance transportation. His broader point was that society can overbuild infrastructure optimized for one technological era immediately before a different architecture reduces its importance.

His preferred solution is not necessarily to relocate data centers but eventually to make AI efficient enough that far fewer are required. He therefore sees massive centralized infrastructure investment as a significant economic risk.

Asked what single message he would leave a younger generation, Stout said people should learn to listen to God and recognize that guidance. He described it not as an audible voice but as learning to recognize moments when one understands clearly what is right or wrong.

Stout explicitly identified himself as Christian and described Christianity as a relationship rather than merely a religion. He said it became convincing to him through experience, study and continued examination rather than through comparison-shopping among religions.

When asked how he wants to be remembered, Stout focused more on character than wealth or WebAI’s valuation. He said he wants to be remembered as someone who cared deeply about people, treated them with respect, loved his country and worked for a better future for the next generation.

Stout said a major personal motivation is wanting the United States and individual people to succeed over the long term. He believes the country is not currently on the right path and wants his work to contribute to a future he considers more compelling for generations beyond his own lifetime.