Sam Altman discusses with Stripe CEO: The era where ideas are more valuable than code has arrived!

null

Source: Stripe

On April 30, 2026, OpenAI CEO Sam Altman appeared on stage at Stripe’s annual conference, engaging in an in-depth fireside chat with Stripe CEO Patrick Collison.

The two have known each other for nearly twenty years, and the conversation covered pivotal moments in AI development, OpenAI’s management philosophy, changes in the entrepreneurial ecosystem, and AI’s profound impact on science and humanity’s future.

During the dialogue, Altman put forward a series of heavyweight viewpoints:

We are indeed in some kind of takeoff phase. AI development is extremely rapid, changing slightly every week compared to the last.

OpenAI has undergone three evolutions: from a research institution, to a product company, to a large-scale Token factory.

“The revenge of the ‘idea-driven’ types” has arrived: I now want to invest in people who deeply understand user needs, have product insights, but can’t code at all.

What excites me most about AI isn’t the product or the business model, but the potential to accelerate scientific discovery.

We are in a takeoff phase

When did the singularity begin?

Patrick Collison offered an interesting perspective at the start: viewing this year as the inaugural year of the singularity.

In response, Sam Altman said, “We are indeed in some kind of takeoff phase.” From the second half of last year to the beginning of this year, the capabilities of AI models crossed a critical threshold—especially in code generation.

“Every week is a little different from the last; things are developing very quickly.”

A shift in perception

Currently, Codex (OpenAI’s programming model branch) is in its “highlight moment.”

While the most dedicated users remain programmers, a large number of non-programming users are also rushing in, trying to use it for all their daily computer tasks.

Altman believes people will experience a more widespread cognitive shift: realizing how much time they waste on trivial computer tasks.

Switching messaging apps, copying and pasting content, handling repetitive emails that could obviously be automated—these small tasks are quietly eroding people’s focus and work experience. When most people truly realize that AI can help them eliminate these “drudgery jobs,” the feeling will be revolutionary.

Who is truly leveraging AI well?

CEO involvement is essential

After observing numerous enterprise clients, Altman concluded that the most successful AI applications share a common trait—CEOs personally get involved.

It’s not just about symbolically declaring “We embrace AI,” but actually building automation workflows themselves and demanding their teams to keep up. He cited Shopify’s CEO as an example: one of the earliest CEOs Altman has seen personally involved, directly pushing the entire company to incorporate AI into all aspects.

OpenAI is now experimenting with a new approach: sending an engineer to work directly alongside enterprise CEOs, helping them automate as many workflows as possible.

If a company’s leadership can truly feel the power of AI, that feeling will permeate the entire organization like a fractal.

OpenAI’s three evolutions

Sam Altman candidly shared the management evolution within OpenAI, which also reflects the path of AI industrialization.

Three phases of transformation

First phase: a pure research institution aiming to figure out how to build AGI when everyone thought it was crazy.

Second phase: continuing research while learning how to operate as a product company.

Third phase—currently entering: building a large-scale Token factory on top of the first two. Altman compares it to a new kind of utility—like electricity—where the world needs massive, cheap, accessible intelligence.

A vision of low-profit infrastructure

In response to questions about whether “AI giants will monopolize everything,” Altman referenced Stripe: Stripe and its customers are highly aligned—Stripe earns more when its customers do better, creating a healthy infrastructure relationship.

Altman hopes OpenAI can eventually serve as such a role: a provider of intelligent infrastructure, even if it remains low-profit forever, as long as it is large enough, fast enough, and deeply integrated with the world’s distributed economy.

He also admits that the cost of AI transformation is inherently low, making high profits difficult to sustain. The recent mass migration of users from competitors to Codex shows that switching friction in the AI era will continue to decrease.

Computing power investment: the most expensive infrastructure in history

Regarding large-scale computing investments, Altman said, “This will be the most expensive infrastructure project in human history.”

Each GPU’s efficiency has exceeded his expectations, but demand is growing even faster. As for how much computing capacity is needed? “I don’t have a good answer… to some extent, the demand is almost unlimited.”

OpenAI’s management philosophy

OpenAI gathers the smartest and most individualistic people in the world. Altman revealed that his secret lies in extreme belief games:

Resource concentration: During GPT-3 training, OpenAI invested nearly all its computing resources into one project. At the time, DeepMind warned that this could create a toxic competitive culture; OpenAI’s response was: We believe this is the right direction.

Shared vision: Altman believes that even if team members have personal conflicts or disagreements, their common belief in “scaling” allows them to sit together and solve problems.

“Direct management of hundreds of people”

When asked if he has any unusual management habits, Altman mentioned: he communicates directly with hundreds of people in the company via Slack—either through assistants or directly himself, with each message brief and to the point.

This decentralized approach sometimes provides him with very valuable information.

A new paradigm for startups

“The revenge of the ‘idea-driven’ types”

During YC days, Altman developed a deep-seated bias: he looked down on entrepreneurs who only have ideas and need a programmer to implement them, thinking it’s as absurd as saying “I have a good song idea, just need a guitarist to help me produce it.”

But now, “the revenge of the idea-driven types” has arrived.

People who deeply understand user needs and have product insights but can’t code at all can now quickly build products using AI tools. Altman says he is now very willing to invest in such people.

How to invest before the singularity?

AGI may arrive within three to five years—does the traditional ten-year venture capital investment horizon still make sense?

Altman’s answer: On this timescale, do everything with a “suspension of disbelief.” You can’t just do nothing because “the singularity will arrive in three years and we can’t see through it.” You still have to act as if life will continue.

OpenAI has signed twenty-year power and land agreements but only has a clear grasp of its product roadmap for the next two years—this is his approach: make long-term infrastructure investments while staying clear-headed about the near term.

AI is reshaping scientific discovery

Altman is most excited about AI’s potential to accelerate scientific discovery, not the product or business model.

He believes this will be the most profound contribution of this technology to human life quality.

Tackling complex diseases

Through collaboration with the Arc Institute, OpenAI is supporting research on complex diseases involving multiple genes, such as cancer and Alzheimer’s, using large biological foundational models like Evo 2.

AI is shortening research cycles from ten years to one year.

Breakthroughs in energy and materials

He specifically highlights a severely underestimated field: materials science.

AI excels at searching for optimal solutions within vast combinatorial spaces, which will lead to breakthroughs in catalyst development, energy efficiency, and more. He expects rapid progress here that will profoundly change all our lives.

👉 on energy, Altman boldly predicts: Driven by AI compute demands, the first profitable fusion reactor could appear within five years.

Democratization: Sam Altman’s final stance

At the end of the conversation, Altman discussed one of the most controversial decisions in OpenAI’s history: iterative deployment.

He recalled that many safety experts advocated locking AI in an “ivory tower,” controlled by a few elites, before distributing results to the world.

“I find that idea very uncomfortable,” Altman said. “Avoiding power concentration and truly making this technology belong to everyone is extremely important.”

“People will use AI in all sorts of ways, but I believe most are good, and most will use tools to do remarkable things. I think my most important contribution is to push this technology to be accessible and buildable by everyone.”