How X's Moonshot Factory Systematizes Radical Innovation

What if the secret to building a Waymo wasn't relentless belief — but ruthless intellectual honesty? Astro Teller, who has led X for 16 years, argues that systematizing radical innovation means training yourself to kill ideas faster, not fight harder for them.

Key Facts

• Ideas evaluated per year: 100–200 ideas receive a code name at X each year (Astro Teller)
• Target success rate: 1–2% of started ideas become once-in-a-generation opportunities (Astro Teller)
• Founding team equity: Graduating teams receive founders' shares with no personal financial risk taken during exploration (Astro Teller)
• Key metric: Learning per dollar, not progress per dollar, is what X optimizes for (Astro Teller)
• Board meeting format: Mature projects hold 'Quarterly Learning Reviews' — not progress reviews (Astro Teller)

What Is the Moonshot Factory and How Does It Work?

X (Alphabet's Moonshot Factory) is a 16-year-old innovation lab designed to try 1,000–1,500 ideas per decade, kill most of them cheaply, and graduate the 1–2% that could become transformative businesses like Waymo or Google Brain — all without founders risking personal capital.

Most founders pour everything into a single idea. X is built on the opposite principle: run many ideas in parallel, fail fast, and protect teams from the psychological and financial cost of betting wrong. Astro Teller, who co-founded X and has led it for 16 years, describes it as a '21st century Bell Labs' parked inside Alphabet — not to solve Google's problems, but to find entirely new ones.

Outputs come in three forms. Some projects, like Google Brain, become so central to Google they move back inside. Others, like Waymo (self-driving cars) and Wing (drone delivery), become independent Alphabet subsidiaries. Increasingly, the best ideas graduate as fully independent companies where Alphabet holds a minority stake without a board seat.

The efficiency of this model depends entirely on one discipline: stopping bad ideas early. According to Teller, 'our job is to try 1,000, maybe 1,500 ideas per decade... if we get a couple Google Brains or Waymos per decade, especially if we're really efficient at throwing most of those ideas away early, then this is really good for our investor Alphabet and really good for the world.'

"Taking moonshots is really easy, super easy if you don't care about efficiency. You find really high-energy people who think they're right and are pretty smart, you dump a bunch of money on them, and you'll get some moonshots. It's just not an efficient process." — Astro Teller

"What I've become obsessed with over the last 20 years is how do you systematize radical innovation?" — Astro Teller

The Card Counter Mindset: Detaching Ego from Ideas

X operates like card counters at a casino — not lucky bettors. Teams must rewire their sense of self-worth from 'winning' to 'how wisely we explore.' Killing a project is celebrated as intellectual honesty, not failure. No one tracks whose idea it was, because ideas are cheap; discovering their value is what matters.

The most counterintuitive thing about X is its culture around failure. Teller is explicit: if you believe deeply in a single idea, don't come to X. 'If you want to go to Vegas and you just believe that 17 is your lucky number and you want to bet on 17 on the roulette wheel... that's not what happens at X. At X, we're trying to be the card counters of innovation.'

This requires a fundamental psychological rewire. From childhood, most people are taught that winning equals worth. At X, that equation breaks things. Teams that cling to ideas become unable to kill them honestly. So X does something radical: it doesn't track whose idea it was. 'Ideas are trivially easy. We have a thousand ideas a year... our alpha comes from how efficiently we can discover which ideas are valuable.'

The structural protection matters too. Because no one bets their own money at X, the fear of being wrong shrinks. If a project dies, you try again. If it graduates, you get founders' shares. 'The good news is you haven't bet your kids' college fund on the teleporter,' Teller notes.

"You have to rewire your sense of self-worth to how wisely you can explore ideas, how viciously, in fact, we can attack our own ideas and pressure test them to see which ones survive." — Astro Teller

"If you want to have street cred here, it's not because you came up with the idea. Who cares? It's because you helped us kill the idea." — Astro Teller

• X does not track or credit whose idea originated a project.
• Street cred at X comes from killing ideas or finding the insight that makes a good idea great.
• Teams work on a 'major and minor' simultaneously to reduce existential attachment to any single project.
• Celebrating honest kills is built into X's culture as a concrete norm.

Train the Monkey First: Testing the Riskiest Assumption

X's core operating principle is to attack the riskiest part of any idea first. Don't build the pedestal before training the monkey. Every project must have a testable hypothesis that targets its biggest uncertainty — whether that's technoeconomics, physics, or user adoption — before any other work proceeds.

Teller tells a memorable story from a Wall Street Journal conference where he struggled to explain why teams should work on the hardest thing first. He landed on a metaphor that stuck: imagine your project is to get a monkey to stand on a 10-foot pedestal and recite Shakespeare. 'Which should we do first — build the pedestal or train the monkey? If you built the pedestal, you're half done, but you're really 0% done at burning down the risk.'

Now 'what's the monkey?' is shorthand throughout X. It means: what is the riskiest, least-proven assumption in this project, and are you actively attacking it right now? For Wing (drone delivery), the early monkey was safety. For the clean-water-from-air project, the monkey was technoeconomics — could they tell a credible story about reaching a penny per liter within a year? At 10 cents per liter, Teller says, it's 'glamping.' At a penny, it changes a billion lives.

Critically, the monkey test applies even before building anything. Teller's rule: try to kill the idea on a whiteboard first. If it survives, run a simulation. If that fails, find a university with the right lab. Only after exhausting cheaper tests do you invest in expensive infrastructure.

"If you built the pedestal, you're half done, but you're really 0% done at burning down the risk." — Astro Teller

"If the scary part is the technoeconomics, let's not wait. Let's get really crisp about what has to be true." — Astro Teller

Learning Per Dollar: Why Progress Is a Smoke Screen

X optimizes for information gain, not progress. Progress reports signal activity; learning reviews reveal whether an idea is worth continuing. When the goal is over the horizon, rowing harder doesn't help — you need the sextant. Spending 90% of your time on navigation and 10% on execution is the right ratio for moonshot exploration.

Teller draws a sharp contrast between startup mode and exploration mode. Traditional startups are built for speed when the goal is visible — 'the only really hard thing is how fast you run getting from here to there.' But moonshots operate in a fundamentally different environment. 'If your goal is over the horizon, you don't even know what direction it is, and it might not even exist.'

His rowboat analogy captures it perfectly: if you're lost at sea, rowing vigorously feels productive but achieves nothing. 'If we're lost in a rowboat, I'm going to spend 90% of my time with the sextant and 10% of my time rowing.' The sextant is the testable hypothesis, the kill criteria, the quarterly learning review. The rowing is everything else.

This philosophy shapes X's internal governance. Mature projects don't have progress reviews — they have Quarterly Learning Reviews. 'Don't show me your progress. Progress is like a smoke screen... What we want to know is what have you really learned?' This framing also explains why X is comfortable graduating companies with seed-stage product market fit after six years — they've run seven different tests of the right connection point. That's not failure; it's precision.

"If we're lost in a rowboat, I'm going to spend 90% of my time with the sextant and 10% of my time rowing." — Astro Teller

"Information gain, the learning per dollar, is what we're actually trying to maximize." — Astro Teller

Kill Criteria: How X Decides to Stop a Project

Kill criteria are written commitments made at project launch — a team's honest promise to their future selves about what must be true to continue. They are owned by the team, not imposed by leadership. When the criteria aren't met, the default is to kill the project, and the team must argue actively to continue.

One of X's most practical tools is pre-commitment. When a project moves from investigation to full-time pursuit, the team writes down specific, measurable conditions that must be met within a defined period. These aren't revenue targets — they're learning targets. For the clean-water project, the kill criterion was: within one year, can the team tell a 'fairly exciting and specific story' about reaching a penny per liter? Not achieve it. Just tell a credible story.

Teller insists the team writes these criteria themselves. 'I don't make up the kill criteria. I want you to make the kill criteria because I want you to own it.' If a leader cares more about keeping their project alive than finding the truth, they're misaligned with X's culture — and that misalignment alone is grounds to kill the project.

The incentive structure matters too. Teller acknowledges that in traditional venture-backed startups, entrepreneurs are rationally incentivized to keep going even when the signs say stop — they've burned their boats. X removes that trap by ensuring no one has personally risked capital. 'It's their life that's ticking away,' he says — and that should be enough motivation. The kill criteria make the decision point legible before emotions cloud it.

"This is like an honesty message from you to your future self." — Astro Teller

"Once we agree that it's more important to find the truth than for you to keep the teleporter alive — you tell me. I want the truth, and so do you." — Astro Teller

• Kill criteria are written at the moment a project goes full-time — not after problems appear.
• Criteria are learning milestones, not revenue or product milestones.
• Teams own the criteria; leadership holds them accountable to what they wrote.
• Failing to meet criteria makes killing the default; continuing requires active justification.

Moonshot Market Fit vs. Product Market Fit

Product market fit asks whether people pay for something and pull it toward them. Moonshot market fit asks whether you're positioned to change an entire system — not just sell into it. Building revenue with incumbents who don't want their industry disrupted is a ladder to the moon, not a path to it.

Teller respects product market fit but warns it can be dangerously narrow. The spectrum from 'are people paying for it' to 'are the right people pulling it from us at a price that covers costs' matters — but it's still only part of the picture for system-level change.

His electric grid example illustrates the difference. X couldn't fake a grid in a lab or demand data from operators protecting national security infrastructure. So they spent five years building rapport and solving adjacent problems — work that looked like product development but was actually trust-building. 'The first couple of things we did weren't even the moonshot. It was just getting a rapport going.' That's moonshot market fit: understanding what it takes to move a whole system, then working backwards to identify your first real partner.

The trap is selling to incumbents who don't want change. You can build a billion-dollar business. You will not change the industry. Teller is direct with potential partners: 'We're going to blow up your industry. Are you interested in doing that with us? If not, please do not sign up with us. You'll just be angry at us later.'

"If I ask you to get to the moon and you show me that you're a foot closer to the moon every day — not interesting. That's probably a ladder to the moon." — Astro Teller

"We don't want product market fit. We want this sort of moonshot market fit, which is much more holistic." — Astro Teller

Product Market Fit vs. Moonshot Market Fit