How Great Tech Leaders Think and Decide | Ex-Meta CTO & Gigascale Founder, Mike Schroepfer
These notes were generated by AI and may contain inaccuracies.
Mike Schroepfer spent 25 years building, starting, and scaling technology companies. Worked at two small startups in the dotcom boom. Started his own company after the dotcom crash, sold it to a bigger company on microsystems. Joined Mozilla, helped ship Firefox versions 1.5, 2.0, 3.0 when version numbers were small. In 2008, joined Facebook, smaller than MySpace at the time. Over 14-15 years, led engineering, built data centers, entered consumer hardware with virtual reality headsets, managed multiple billion-dollar acquisitions, built an AI research lab, spanning hardware, software, enterprise, consumer, scaled team from 100 to 35,000.
Now helps entrepreneurs via Gigascale Capital, hunting founders with big ideas using technological advances to build products that improve lives, solve climate/environmental problems, and become great businesses. Example: electric vehicles—no pollution, cheaper to operate/maintain, faster, quieter; traditionally more expensive but batteries getting cheaper, now cost-competitive.
August/September 2008: Facebook had a loved product (facebook.com), growing signups, new features, but urgent scale problem—software/hardware backbone didn't exist. First years: scaling/rebuilding software architecture, building hardware infrastructure. Released space in data centers, put in servers, but 2008 financial/real estate crisis stopped data center builds, so built own with challenges, mistakes, fixes. Hired experts in data centers, network design. Lesson: hire team to understand new areas. Startups solve out of necessity—no choice.
There's no getting away from the hard problems. You just got to get to it.
Identify critical risks, highest technical risks, work on those first. Human instinct: solve tractable problems (e.g., clean office instead of hard pitch). Focus on right hard problems. Early Facebook: pants-on-fire scaling, eventually had energy to look forward.
2013: Facebook started AI research. Debate: broad research lab like Microsoft/IBM Research? Decided focus solely on AI—too big/impactful. Mark Zuckerberg pushed for AI-only. Visibility to ImageNet challenge: neural net vastly outperformed others—10% better gap, trained on data. Powered by neural net size, dataset size, compute (training/inference)—all scalable thousands of times even without inventions.
It's a technology with a lot of runway. It's not fully optimized.
Exciting technologies at infancy vs. optimized for 150 years. Hindsight bias: people doubt new tech, claim obvious later. Early AI: image analysis, translation worked; chatbots poor (e.g., demos terrible, basic questions like 'how many people in photo? cat?'). Not useful yet—hard to believe improvement. ChatGPT moment: consumers experienced useful AI chatbot, believed.
Self-driving cars: people scared conceptually; Waymo rides in San Francisco—5 minutes in, bored, better than distracted/tired humans.
Challenge of new technology: until obviously useful/personal experience, easy to doubt. Once there, captured all value.
Identify technologies with opportunity to reach that point.
Three core things:
- Lightspeed test: How far from theoretical maximum? (E.g., spacecraft at 99.9% light speed—no room; 0.001%—lots.)
- Tailwinds: Inputs improving without your work (e.g., AI compute: Nvidia/TSMC/ASML ecosystem doubled power every 18 months for same dollar—no chip team needed).
- Problem importance to customer: Amazing tech can fail without customer problem (e.g., 3D TVs—immersive but people hate glasses). Mock up proxies to test customer interest—someone must buy to fund R&D.
Problem: scaling clean energy for 8 billion people—terawatts needed. Energy upstream of everything (AI, AC, clean water, manufacturing). Industrial revolution: harnessed energy for productivity/health/happiness.
Sustainability: re-engineering tens of trillions in economy (energy multi-trillion). Needs businesses/startups, not government/philanthropy. Examples: fusion, next-gen micro reactors, offshore AI data centers, kitchen dehydrators.
Mike's experience uniquely valuable for helping skip mistakes (hiring execs, product dev). Exciting: clean power can 10x/5x global use. Fusion: sun's power source, works in universe/on Earth; need reliable power plant—no inputs, no emissions, safe, unlimited. Power all of Austin, TX with one pickup truck/year fuel (vs. trainloads coal/gas).
Example: Mill—trash can for food waste: dries/grinds to coffee grounds. Average family: month = shoebox size, no smell, empty monthly. Pitch: empty less, stinks less. Diverts food waste emissions (methane from landfills). Customers love it, great revenue—better/faster/cheaper, improves lives, solves climate problem, great business.
Built tens of millions sq ft data centers, shipped tens of millions consumer hardware, scaled to tens of thousands, multi-billion acquisitions. Applies: problem ID, tech with headroom/tailwinds/customer demand, people—right leaders.
In startups, bet on team/founders who scale from 10 to 300+ employees (rare adaptability). Meet 1000s founders/year, pick 10 best to scale to public company—pattern matching on scaling ability.
- Relentlessness/determination: Endless near-death disasters, nos from recruits/investors/customers. 30 nos = 1 yes wins. Mike's first startup: tons nos (one investor slept), but Sequoia yes—defining.
- Humility + curiosity: Learn new domains rapidly (CEO job changes daily: tech, customers, recruiting). Consumers of new info—learn/hire fast (e.g., run board, pitch investors).
No checklist (e.g., second-time founders)—meet multiple times, references.
Examples:
- Common Fusion Systems: Bob Mumgard (plasma physicist, first company). Operator: builds team, tells story, raises money. Excels at mission clarity, aligning people.
- Mill: Matt Rogers (second company, ex-Nest). Builds great team. Solves hard unknowns fast (e.g., US gov approval for food waste → chicken feed: promised, done in 2-3 months).
Magic of startup: team takes down big problems no one's solved before.
carry_over: ""
Keep EO in your library
Save the videos and channels worth coming back to, and find them again in one place.
