Astro Teller: The $1B Bet No CEO Will Back, Moonshots 3x Cheaper in 16 yrs, and Clean Water at 1¢/L
In a Nutshell
Astro Teller runs X's moonshot factory, where only 2% of 100-200 annual ideas become projects, but those that succeed deliver 10x returns versus guaranteed smaller bets—yet most CEOs reject this math due to risk aversion. The system succeeds by enforcing first-principles economics, killing projects early when technical costs exceed viable pricing, and maintaining small teams focused on rapid learning rather than efficiency. AI is reducing moonshot costs 3x in 16 years by accelerating risk reduction, though human oversight remains essential for problem selection and social acceptance for the next two decades.
These notes were generated by AI and may contain inaccuracies.
Astro Teller's maternal grandfather was a Nobel laureate in economics. His paternal grandfather was Edward Teller, the inventor of the hydrogen bomb.
Three basic components are required for a moonshot hypothesis:
- A huge, nameable problem in the world that you want to solve
- A product or service that looks like science fiction that would solve that problem
- A breakthrough technology that gives hope to create that product and solve the problem
This creates a testable hypothesis, not a guaranteed success.
Two equal qualities are required:
- High courage to suspend critical judgment and embark on unlikely journeys
- Equal humility to recognize from the outset that the project is improbable and most likely won't succeed
This enables quick learning and moving to the next attempt when needed.
Ideas can be stopped quickly when:
- The idea isn't big enough
- The idea is too logical or likely to succeed
- It's not good for the world or has high likelihood of secondary problems
- Technical economics don't work
Apply first-principles thinking by examining:
- Best-case scenario for materials list
- Raw cost by weight of materials
- What people will actually pay
- Whether it can be logically acceptable from the start
Many early-stage moonshots fail sustainable business criteria.
Purpose and profit can support each other. Projects that lose money are unlikely to change the world.
- Astro joined Google 16.5 years ago when X was co-founded
- Google was worth approximately $30 billion at that time
- Current value is almost half a trillion dollars
- Target: 100-200 ideas per year that receive codenames
- After 5-6 years, approximately 2 moonshot-level projects emerge
- Success rate: 2%
AI will shorten the time from crazy idea to risk reduction where definitive proof can be stated. AI will allow raising the bar for audacity higher and achieving more in less time.
- Started 15.5 years ago with Andrew Ng and Jeff Dean
- Andrew Ng and others believed scale might matter for neural networks when they were considered dead
- First to make neural networks tens of thousands of times bigger than previously attempted
- Outputs include Tensor Processing Units (TPUs) and the Transformer architecture (the "T" in ChatGPT)
Every innovation is an overnight success that took 15-20 years. The general public doesn't recognize success until the 18th year of struggle.
Clean water production: Need to reach 1 cent per liter total cost. Current attempts achieve 10 cents per liter. Nearly 3 billion people are water-stressed without clean drinking water access.
Energy transmission across time and space: Multiple attempts made without sufficient excitement about results.
Education: Doesn't work in developed or developing worlds. Will continue returning to this problem.
From approximately 1970-2020, there was a slowdown in humanity's ambitious project pace. Stopped landing humans on the moon and applying physics innovations. Around 2020, moonshots resumed with superintelligence, robots, physical intelligence, and medical nanotechnology like lipid nanoparticles.
Undertaking ambitious projects is ridiculously easy if you don't care about efficiency. Find incredibly energetic and optimistic people, shower them with money, and you'll get moonshots. However, they won't have good return on investment.
X is building a moonshot factory that achieves audacity efficiently enough to make it reasonable for investors to provide more funding. Large corporations and countries are requesting help creating their own venture factories.
Core organization responsible for generating 10% profit year after year. Venture factory should be on the edge of the organization where crazy thinkers are located. Focus on 10x improvements (1000%), not 10% improvements. Venture side organization should report directly to the CEO.
Option A: Guaranteed $1 million in value Option B: $1 billion with 1 in 100 chance Option B has 10x higher projected value. When CEOs are asked if their managers, CEO, or board would support choosing Option B, hands go down despite initial preference for Option B.
"B" options must be isolated in a place that tolerates messiness. Examples include Lockheed Skunk Works on the periphery and Steve Jobs taking the Mac team off campus.
For the first time in human history, advanced technologies are cheap: solar power, cheap sensors, open-source and free blockchain technology. This enables portfolio approaches to disruptive experiments on the sidelines rather than requiring massive bets.
Teams that graduate from X, like Google Brain, had about 18 people. Small teams are maintained to find "bunny" moments to cut the Gordian knot rather than solving things conventionally.
A 20-hour professional certification course on groundbreaking innovation is available, taking 3.5 hours if only watching videos. A 50-minute condensed version runs at 7 p.m. with Joe Sargent, focused on rapid prototyping.
Cost of accessing graduate projects has dropped about three times over 16 years. Cost of each individual project through the process decreases 10-20% per year. Factors include getting better at the work, finishing projects faster, and artificial intelligence.
The cost of accessing Waymo versus its value are separate considerations. Waymo's value is enormous compared to development costs and failed projects, representing good return on investment.
X produces 2 moonshot projects per year. The limitation is not ambition but efficiency concerns. Cultural rhythm and habits are difficult to practice at scale: long-term thinking, intellectual honesty, humility, teamwork, and avoiding arrogance.
These adaptive behaviors require insane determination to create a microcosm where they actually occur. The methodology works with a few hundred people but scaling to thousands remains unsolved.
The super-project is systematizing radical innovation and copying/pasting moonshot factories. The world requests this capability. X has been building a community layer for 10 years, now with 50,000 people globally working on rapid launch environments.
The current world is a transitional period where three artificial intelligences and three people work together, overwhelmed by computing and money. The term "artificial intelligence" should be replaced with "superintelligence" as the operational details are less important than the actual problems being solved.
Every ambitious project combines agents and people working together, but the obsession is with the problem itself—whether clean water, grid-scale energy storage, education, fixing the electrical grid, or creating a circular economy. The goal is integrating solutions into businesses that grow over time without excessive costs.
Teams are currently working on the clean water project aiming for one penny per liter. In this specific situation, salaries still exceed the AI bill, though the exact percentage is unknown. AI and humans are partners in the innovation process, with ongoing automation of factory parts for ambitious projects.
The AI bill is currently less than the human recruitment bill. The 50/50 ratio, 75% AI, and 100% AI milestones may not reach 100% within the next decade because identifying problems and ensuring socially acceptable solutions requires human involvement for at least the next two decades. Computing capabilities are used only when they provide actual benefit, not for their own sake.
Ambitious projects fall into two categories: pain points (solving hunger, education problems) and forward-looking projects (getting to Mars, solving aging). The approach seeks the most daring things that can turn from crazy ideas into sensible ideas in less than a decade, then into sustainable businesses in less than another decade.
Time constraints are imposed to determine whether problems can be worked on reliably over the next decade and produce baseline solutions. As the money recipient from Alphabet, there must be clear communication about creating growth value for investors.
Over 16 years, approximately 2,000 ideas have been given code names, with 35-50 projects graduating, representing roughly 2% success rate. The importance of killing ideas early is emphasized through the monkey and platform analogy: focus on training the monkey first before building the platform, as the platform can always be built later if the core challenge is solved.
The approach is not about wanting to lose but recognizing that nothing is learned when right—only when wrong. Ambitious projects are a process depending almost entirely on learning speed. The goal is removing the stigma associated with failure, as feeling bad about learning moments leads to unconsciously avoiding them.
The objective is preventing projects from becoming "living dead men" that continue for years despite being known as failures. Examples include a current project that poses a physics dilemma if true, working in the lab without understanding why, resembling cold fusion (which is neither cold nor fusion). Five months were spent trying to understand this phenomenon.
The cost of false positivity—believing a project is ambitious and continuing for years only to discover it isn't—can reach tens of millions of dollars. The cost of false negativity—rejecting a genuinely ambitious project—is zero, as there are infinite problems and creative solutions to try again.
New materials present exciting opportunities, but the challenge extends beyond discovery to mass production. Steel took centuries to refine and mass-produce before changing the world. New materials require not just creating single chips or Nobel Prize-winning discoveries, but scaling to thousands of tons daily and making materials flexible for applications like coiled wires—the remaining 95% of the work beyond initial discovery.
The most difficult aspect for other companies to imitate is the leadership team's obsessive focus on engineering a culture that enables radical innovation. This creates protected space for explorers. The parent company's discipline in leaving the project alone for extended periods prevents the corporate immune system from reclaiming and ruining projects, allowing the magic to happen at the periphery.
Keep Peter H. Diamandis in your library
Save the videos and channels worth coming back to, and find them again in one place.
