Diamandis Awards $2M to Five Teams That Built Revenue in 90 Days

XPRIZE founder Peter Diamandis handed five teams a combined $2 million on September 28 after each built a revenue-generating company in 90 days from a blank slate. The payouts at his Moonshots Live event ranged from $100,000 for an AI tutoring startup to $500,000 for Polyform, an AI-native 3D asset platform.

The exercise wasn't a typical startup competition. Diamandis called it a proof of concept. "The mission here is not in these specific companies," he said. "The mission here is to show that it is possible. To teach people, to inspire people to do it themselves."

What changed: the cost of starting just collapsed

Cathie Wood, CEO of ARK Invest, framed the contest as a response to a labor market that has stopped absorbing graduates the way it once did. Average unemployment for new grads now runs six months; the median sits at three. Her pitch to that cohort: don't wait. Build something.

"I think we're going to have an entrepreneurial explosion," Wood said. Her math is simple. The 90% startup failure rate hasn't moved, but the cost of failing has. Three months of solo founder time with AI tools looks nothing like three years and a burned seed round.

She called the unemployment spell "a blessing in disguise." That reframing matters more than the slogan. If a 22-year-old can ship a working product without hiring anyone, the downside risk of trying collapses.

What does this change for a founder starting today?

The binding constraint has shifted. Diamandis and four judges — Wood, Logan Kilpatrick of Google DeepMind, Palmer Luckey, and Zynga's Mark Pincus — converged on a single thesis: capital and headcount no longer gate company formation. Finding a real problem and owning a proprietary input does.

Practically, that means founders should hunt for domain expertise or unique data first, then build the company around it. Kilpatrick was most explicit. Compute has been de-risked over the past three years at a cost of hundreds of billions of dollars. Data hasn't.

"The last part of this, which is currently the real bottleneck for continued progress, is data," Kilpatrick said. The investment signal he cited: the current Y Combinator cohort runs overwhelmingly data-driven, and even the best AI data companies trade around $10 billion today. He expects them to get "much, very much more valuable very, very soon."

The hardest number in the room: drug discovery

Wood saved her sharpest forecast for healthcare. Bringing a new drug to market currently costs about $2.4 billion and runs 13 years. Her estimate with AI: $600 million to $700 million — roughly a 70% cut — and eight years or less.

The mechanism isn't mysterious. AI lowers the failure rate across clinical trials. Every sunk trial compounds through the pipeline. If startup success rates climb from the 10% baseline, Wood suggested, the implication extends past pharma. It means AI doesn't just cut costs. It improves the odds of winning.

What the five winners actually built

The finalists used AI against specific, mostly workflow-heavy problems:

• Polyform ($500,000): AI-generated 3D assets that anyone can recolor and remix
• Tichen Sierra ($200,000): automates VAT tax credit recovery for Mexican businesses
• MyFixit: matches Nigeria's informal craft workers with paying customers
• Launch Bridge: takes an entrepreneur from idea to registered LLC in under 72 hours
• Dodo Prep ($100,000): turns any learning material into a personalized AI tutor

Diamandis noted 20 more teams won outside the final five. Mark Pincus, whose Zynga once built FarmVille into a billion-dollar social game business, warned dryly that those checks "will be diluted by taxes."

The outlier bets

Two judges answered "what would you build from zero" with non-software plays. Luckey pitched unregulated oxytocin dosing for marriage counseling, citing Sigma Aldrich's retail supply and zero regulatory friction. Pincus, contrarian about his own legacy, proposed reinventing social media to reward users for spending less time on it. "I could have used that 8 years ago," he said when Luckey pitched the marriage-counseling route.

The so-what

Diamandis's stated goal is replicability, not backing five startups. If his next cohort pulls from unemployed graduates and domain experts with proprietary data — not from the traditional founder pipeline — the "AI entrepreneurship explosion" thesis earns its evidence. For now, $2 million, 90 days, and five working companies is a small proof point with a very large claim riding on it.