"What If You Succeed?" Patrick Collison on Startups & AI

Key Insights

• Ask "What if you succeed?" before raising money—think about whether you'll enjoy working on your company for decades, not just whether you'll avoid failure.
• Stripe data shows explosive growth: New business formation is up 2x year-over-year, the largest jump ever recorded. Median businesses are also performing better than a year ago.
• AI hasn't centralized opportunity—it's decentralized it. More companies are starting and scaling faster than ever, with thousands of potential winners ahead.
• Timing matters less than concrete customer problems. Stripe took two years to launch publicly because security, partnerships, and infrastructure required upfront investment—but production users from day one kept the company grounded.
• Don't fear big labs. History shows that even dominant companies (Google, Oracle) can't execute everything. The real risk today is obsolescence, not being crushed by incumbents.

The "What If You Succeed?" Question

Most founders obsess over failure. Patrick Collison argues that's backwards. Before raising significant capital, ask the inverse: What if you actually succeed?

Will you want to work on this company for 10, 17, or 30 years? For Stripe, the answer was yes. Collison loves working with the world's most innovative companies—from early startups using Stripe's Atlas incorporation service to scale-ups like Shopify and OpenAI. He's never met a Stripe customer he found boring.

This mindset shift matters because many startup domains involve unglamorous tasks: payroll, infrastructure, compliance. But if you choose the right market—one aligned with your interests—the schlep becomes a feature, not a bug.

Why Building Takes Time (Sometimes)

Stripe's journey challenges the "launch fast, iterate" orthodoxy. The company took nearly two years from first lines of code (fall 2009) to public launch (September 2011). Inside YC, this seemed heretical.

But Stripe's domain demanded it. Payment processing requires security, banking partnerships, money movement infrastructure, and regulatory compliance. You can't prototype these things away.

What saved Stripe from disappearing into the wilderness: production users from day one. By January 2010—just two months in—the first live customer, Ross Boucher, was using the platform. His requests drove development: a dashboard to view charges, refund functionality, the ability to receive payouts. Real customer feedback, not speculation, shaped the product.

This production-user approach meant the company could wait two years for a big public launch without spinning wheels. Every month, the private beta grew. By the time Stripe went public, it had genuine traction and proof of demand.

The AI Era Reshapes Startup Timing

The traditional lean startup playbook—identify a niche, buy Google ads, iterate—emerged during capital constraints. Today, AI changes the equation.

Coding agents make software cheaper and faster to build. This could make lean iteration more competitive, with many companies hunting the same small niches. Conversely, ambitious 1.0 launches—once expensive and risky—are now feasible. Some of the most successful recent companies (OpenAI, Anthropic, Anduril) ignored lean startup doctrine, launching with bold, differentiated starting points.

Collison suspects the future favors "aggressive decorrelation"—companies starting from truly divergent angles where others aren't already competing.

The Stripe Data: Never Been a Better Time

Stripe's real-world payment data reveals a boom:

• New business formation is up 2x year-over-year—the largest jump in Stripe's history (exceeding even 2019-2020's COVID spike).
• Median business performance improved compared to last year. Companies reaching $1M, $5M, and $10M revenue thresholds are hitting those milestones faster.
• Time to revenue for new companies is declining.

This isn't a story of frivolous startups; serious businesses are scaling at unprecedented speed. YC batches show the same trend: companies grow from zero to meaningful revenue within 90 days, driven partly by enterprises willing to buy from early-stage startups—a shift from the past.

The reason: fear of obsolescence. Businesses know the status quo is risky. Even if new startups seem unproven, sticking with legacy systems looks more dangerous. This "spring-loaded" readiness to adopt new solutions has never been stronger.

Will AI Centralize Power or Spread It?

A common fear: AI will concentrate wealth and opportunity among a handful of mega-labs like OpenAI, Google DeepMind, and Anthropic.

Collison sees evidence otherwise. Yes, model providers have done well and will continue to. But:

1. History says dominance is incomplete. Google seemed omnipotent 20 years ago but hasn't executed on every opportunity, even with vast talent and resources. Executing 100 different priorities is hard for any organization.
2. Stripe's data shows broad-based growth. Thousands of companies are getting started, retooling, and leveraging AI capabilities. This isn't a winner-take-most dynamic.
3. The future is decentralized. Based on current trends, Collison expects "many thousands of winners" and a more spread-out economy, not a hegemonic AI cartel.

Vertical-specific disruption will happen (some niches will be obviated by LLMs). But that's different from centralization; it's normal market change.

Conclusion

Patrick Collison's core advice: Think long-term about what you're building, ground yourself in real customer problems, and don't be paralyzed by fear. The data—both from Stripe and YC—suggests we're in an unprecedented window for starting companies, especially in an AI-driven world. The "what if you succeed?" question isn't pessimistic; it's the clearest way to ensure your early hustle becomes a fulfilling decades-long journey.

---

Original source: Patrick Collison: "What If You Succeed?"

powered by osmu.app