Why to trust your own judgment over the consensus

1 The bottom line

Bet against the consensus only when your own reasoning explains why the crowd is wrong, and then hold the bet through the doubt. The crowd reasons from the present, so the largest gains go to those who see the future first.

2 Key takeaways

• Back a contrarian conviction only when you can state, in plain words, why the crowd is wrong. A reason can be tested and defended, while a feeling collapses under the first hard question.
  • Gates and Allen concluded that demand for high quality software would be nearly unlimited, when everyone believed software should be free. Bill Gates, Microsoft27:19
  • Gmail bet that storing and searching email would keep getting cheaper, so it offered 1 GB when Hotmail gave 2 MB and Yahoo Mail 4 MB. Paul Buchheit, Google8:36
• Treat the crowd's objections as a description of the present, and test them against where the trend is going. Doubters usually defend what exists, and the bet is about what comes next.
  • Google Brain's DistBelief updated parameters on stale data, against the field's belief that training had to be synchronous. Jeff Dean and Andrew Ng, Google45:13
  • Schultz felt in his bones that LA was the next market, though his operator wanted San Diego, and the halo there beat every other city. Howard Schultz, Starbucks59:20
• Cut the cost of being wrong before you place the bet. Conviction survives the doubt better when a failure cannot end you.
  • Nvidia had about six months of money and one tape-out left in 1997, so Jensen verified the chip before building it. Jensen Huang, Nvidia8:39
• Once the reason holds, commit the whole organization and move early. A half-bet gets pulled back by the doubters before it can pay.
  • Nvidia built programmable shaders to escape the commodity graphics trap. The bet left it months from running out of cash, and revenue went from $158 million to about $1.4 billion in 2001. Jensen Huang, Nvidia1:11:50
  • Carnegie's way was to focus on the objective and cut through convention, even to the point of burning stalled cars. Andrew Carnegie7:51
• Expect the doubt to last, and keep going on your own evidence. Most great bets looked foolish for years before they looked obvious.
  • Knight's Stanford classmates met his Japanese running shoes idea with boredom and not one question. He spent weeks in the library anyway. Phil Knight, Nike6:54
  • Sonneborn told Dick McDonald the company would one day be bigger than F.W. Woolworth. Dick thought Kroc had a genuine nut on his hands. Harry Sonneborn, McDonald's32:57

Client server was dead and the people in the room would figure it out at the funeral.

Larry Ellison, Oracle

2.1 Find the reason the crowd is wrong

A contrarian bet earns capital only when you can name the specific flaw in the consensus.

• Ask why the new thing works and whether it scales. Jensen Huang, Nvidia13:54
  • When AlexNet leapfrogged 30 years of computer vision work, Jensen asked why it worked and whether it scaled. He concluded that deep learning was a universal function approximator that scales in data and depth. Jensen Huang, Nvidia13:54
  • Analysts said the database market was maturing and the internet would hurt Oracle. Ellison reasoned the internet would multiply database transactions and the number of people touching the databases. Larry Ellison, Oracle2:07
  • Thrun resisted self-driving cars until Larry Page asked for a technical reason it was impossible. Thrun finally admitted, "I'm just afraid." Larry Page, Google2:53:12
• Anchor the bet on a trend you can state with a number.
  • Gates doubted Moore's law at first, because exponentials are rare. Once he accepted it, he concluded that computing would be nearly free and sit on every desk. Bill Gates, Microsoft31:00
  • Others knew the idea of CDMA but did not believe it could work, so no one patented it. Qualcomm bet that one or two more turns of Moore's Law would make it possible, and filed in 1986. Irwin Jacobs, Qualcomm1:05:20
  • Gates later ranked the internet above the IBM PC and the graphical interface, saying exponential phenomena cannot be ignored. Bill Gates, Microsoft19:08
  • In 1914 Wilsdorf predicted pocket watches would almost completely disappear. Watchmakers had listed real objections, such as delicate movements, dust, water and poor accuracy. Hans Wilsdorf, Rolex14:55
• Be suspicious when everyone follows the same rule of good practice.
  • Singleton avoided dividends, never split the stock and ignored quarterly guidance. In 1997 he said that if everyone is doing buybacks, something must be wrong with them. Henry Singleton, Teledyne15:15
  • The Pennsylvania was divesting its western connections while executives prided themselves on low debt. Gould bought all four western legs, reasoning that owning the whole chain gave him more power. Jay Gould29:39
  • Peers told Ortega that nobody in fashion invested in technology and logistics this way. He said he would do it even if no one else did. Amancio Ortega, Inditex23:48
  • Consultants told Hermès to outsource and add cheaper products. Jean-Louis Dumas kept the craft, and revenue grew roughly 40x in under 30 years. Hermès1:19:20

2.2 Check why the doubters doubt

Many objections come from the doubter's position or old data, so sort them before you give in.

• Discount objections from people whose business the idea threatens.
  • Ray Ozzie's cloud project ran outside every division, because it threatened Windows Server, its partners and its sales incentives. Ballmer gave it air cover and later made the company all in. Steve Ballmer, Microsoft3:38:20
  • Everyone in the room but Eddie Cue worried that Jobs's online store would alienate retail partners. Cue pointed out the channel had lost two million dollars the year before, and Jobs sided with him. Steve Jobs, Apple1:10:10
  • Intelligent Cloud reached $88 billion in revenue in fiscal 2023, the largest segment at Microsoft. Microsoft3:38:20
  • Iger ended the Netflix deal and gave up licensing income worth around 2% of profits, to gain control of customer relationships. Bob Iger, Disney3:04:21
• Do not let old data decide a product that has no history.
  • Marketing said nobody would buy a player that could not record. Morita pointed to the millions who bought car stereos with no recording. Akio Morita, Sony43:45
  • Disney feared home video would cannibalize theaters. It was wrong, because people lose and break tapes and want more. Michael Eisner, Disney38:25
  • Amazon's elasticity data said price cuts did not pay for themselves in the short term. Bezos judged that lower prices over five or ten years would produce far more free cash flow. Jeff Bezos, Amazon47:30
  • Vendors asked whether Amazon understood its business when it allowed negative reviews. Bezos accepted lost sales, because better purchase decisions pay off over time. Jeff Bezos, Amazon46:00
• When incumbents see a threat, ask whether it is really an ally.
  • With only about 4,000 TVs in existence, Disney broke with United Artists rather than give up his television rights. He called television the ally of the motion picture, not its enemy. Walt Disney1:27:40
  • Every amusement park operator told Disney that Disneyland would fail. He left those meetings happier and refused to hire anyone with park experience. Walt Disney1:33:00

2.3 Lower the cost of being wrong before you bet

The winners did not bet blind. They simulated, tested small or took the loss themselves.

• Pull future risk forward by simulating it today.
  • Asked if the lesson is to bet on conviction, Jensen said no. He pushes in only when he knows it will work, and prefetches everything in the future that can be simulated today. Jensen Huang, Nvidia10:52
  • He bought a nearly defunct company's emulator. The software team ran the full stack on it, watching Windows paint at roughly an hour per frame. Jensen Huang, Nvidia8:39
  • Before CUDA there was Cg, and before the Riva 128 bet there was a full emulation of the chip, drivers, games and QA. Jensen Huang, Nvidia10:52
• Run a live test before the full commitment.
  • Baldwin and Bowker gave Schultz 100 square feet in the sixth store to test a coffee bar, after he banged on the door for two years. Howard Schultz, Starbucks20:24
  • Eisner started Pinocchio on home video with a limited run. The 1.7 million tapes sold out at once for about $50 million. Michael Eisner, Disney38:25
  • Other animators called synchronized sound unnatural. Disney previewed Steamboat Willie for an audience, and the reaction won him a distribution deal. Walt Disney42:00
• Put your own money and name behind the bet.
  • Partner Pratt wanted to wait on buying oil leases. Rockefeller offered to put up $3 million of his own and take the loss personally, and Pratt came in. John D. Rockefeller, Standard Oil54:30
  • Disney mortgaged everything he owned, including Mickey Mouse and Donald Duck, to make Snow White. Walt Disney1:08:58
  • Morita told doubters in his own company that he would take personal responsibility for the Walkman. Akio Morita, Sony43:45
  • When independent directors disliked a threefold capex increase, Chang said he was the one responsible for running the company, and they accepted. Morris Chang, TSMC50:40
• Place the bet on the steady part of a volatile market.
  • Rockefeller refused to speculate on wells because of boom and bust cycles. He put $4,000 into a refinery in 1863, the steady, strategic middle of the value chain. John D. Rockefeller, Standard Oil51:32

2.4 Commit fully and move before it is obvious

Once the reason holds, a half-bet loses to the pull of the doubters, so commit the whole company.

• Overrule your own lieutenants when the call is yours to make.
  • Several key Nvidia executives argued against more deep learning investment. Jensen overruled them and announced the shift at an all-hands meeting. Jensen Huang, Nvidia51:40
  • Finance wanted to kill a $130 million order for Nvidia GPUs in 2014. Larry Page approved it personally, saying the future of Google is deep learning. Larry Page, Google1:42:28
  • Chang's R&D team called 28nm the sweet spot of Moore's law. He took it as TSMC's tide, and capex jumped from roughly 2 to 2.5 billion dollars a year to almost 6 billion. Morris Chang, TSMC44:05
• Burn the boats so the company cannot drift back.
  • Ellison stopped all work on client server applications and put Oracle's whole engineering effort on the internet. He told staff they were mistaking the present for the future. Larry Ellison, Oracle40:59
  • Facebook stopped caring about right side desktop ads and moved engineering to mobile, knowing it would miss quarters. By Q4 2013 mobile was 53% of ad revenue. Mark Zuckerberg and Sheryl Sandberg, Meta3:46:30
  • By moving tools, database and applications early, Ellison forced competitors to become followers. Larry Ellison, Oracle40:59
• Force the decision when the company's default is the safe path.
  • Roy wanted cheap packages and live action to protect a company heavy with debt. Walt said to sell the company or make features again, and Roy raised $2 million for Cinderella. Walt Disney, Disney2:29:50
  • When the board rejected Disneyland, Walt formed WED Enterprises in 1952 and pulled artists off Disney projects. They became the first Imagineers. Walt Disney, Disney2:41:11
  • Cinderella earned about $8 million in first-year rentals on a $2.2 million cost. Walt Disney, Disney2:29:50
  • Paul Allen quoted Shakespeare, saying to take the current when it serves, because the PC wave would come with or without them. Paul Allen, Microsoft44:14
• Leave the safe job when it asks you to drop the idea.
  • Detroit's electric company offered Ford the general superintendency if he dropped his engine. He chose the automobile and left on August 15, 1899, with his wife's agreement. Henry Ford, Ford Motor Company35:03
  • Litton's president said the component business was too crowded. Singleton left at 43 and founded Teledyne with $450,000. Henry Singleton, Teledyne24:12
  • Bezos imagined himself at 80 and asked which path he would regret least. He left a likely succession at D. E. Shaw. Jeff Bezos, Amazon55:12
  • GE told Packard on his first day that electronics had no future there. HP later grew larger than all of GE was at that time. David Packard, HP13:53

2.5 Hold the bet while everyone doubts you

The price of being early is a long stretch of looking wrong, so decide in advance what you will watch.

• Judge the bet on your own metrics, not on the stock price.
  • After shares fell more than 80% in a year, Bezos wrote that Amazon was in a stronger position than at any time in its past. He watched internal metrics, which were improving. Jeff Bezos, Amazon32:30
  • Newsfeed drew 30,000 angry emails, and 10% of users joined a group against it. Engineers saw engagement soaring, so Zuckerberg added privacy controls and kept it. Mark Zuckerberg, Meta2:18:19
  • Google's stock fell 27% in January 2006 as investors wanted a pure play search company. It kept building Gmail, Maps, Docs and then YouTube. Google6:05
• Fund a long, ugly stretch before the payoff arrives.
  • CUDA spending cut Nvidia's gross margin from 45% to 35% in 2007 and 2008 while the financial crisis hit demand. Jensen held the course as investors demanded a change. Jensen Huang, Nvidia46:40
  • CUDA took more than six years to become usable, and Jensen's logic was that if you don't build it, they can't come. Jensen Huang, Nvidia22:40
  • When AlexNet won ImageNet in 2012 on CUDA, Nvidia was the platform already in place. Jensen Huang, Nvidia47:50
  • Commercial driverless service in Phoenix took until October 2020, 11 years after Waymo began. Larry Page, Google2:53:12
  • Jobs kept writing monthly checks to Pixar for years, putting in more than $50 million of his roughly $70 million. Steve Jobs, Apple37:32
• Take the doubt as fuel, and keep an inner judge.
  • Advisers and his own father called Arnault mad for going after Boussac. He said the consensus acted as a stimulus. Bernard Arnault, LVMH23:00
  • He kept buying LVMH shares as the price rose from about 1,250 to 4,720 francs, while Racamier sold. On January 5 and 6, 1989 he bought 8% to take control. Bernard Arnault, LVMH55:57
  • Carnegie's motto was that the judge within sits in the Supreme Court and can never be cheated. Andrew Carnegie7:07
• Build a small circle that keeps believing with you.
  • Jensen says belief must be very strong, because outsiders question you and a falling stock is embarrassing. He credits long-time colleagues, his family and early investors for never giving up. Jensen Huang, Nvidia1:24:15
  • Edison sought to instill supreme confidence in everyone around him, including the capitalists funding him. They stayed with him through the swings between hope and pessimism. Thomas Edison51:20
  • Knight's father kept asking how long he would keep jackassing around with shoes. His mother quietly bought a pair and wore them around the house. Phil Knight, Nike18:28

2.6 Where they differ

The founders disagree on how much proof a bet needs, and on who can afford to overrule the room.

• Some bets were made without proof, and some only after it.
  • Jensen insists he bets only on what he has simulated. Jensen Huang, Nvidia10:52
  • Walt said they did not know whether the public would go for a cartoon feature, only that audiences would not buy a bad one. Walt Disney, Disney1:24:03
  • Honda imported machine tools costing about 10x the company's assets. A recession hit right after and he could not pay. Soichiro Honda, Honda32:00
• Overruling the room is easier for a founder who cannot be fired.
  • Ben Thompson argues Zuckerberg could answer TikTok at once, despite the Apple tracking hit, only because he could not be fired. The stock was down 72% by Halloween 2022, then later rose about 5x. Mark Zuckerberg, Meta5:09:09
  • A hired CEO would likely have waited 6 to 12 months. Ben Thompson on Meta5:09:09
  • Even Zuckerberg's Newsfeed lesson, that users will adapt, did not always hold. Mark Zuckerberg, Meta2:18:19
• Outsider ignorance helps in some bets, and craft knowledge in others.
  • Sol Price and his partners knew nothing of retail, so they did not know what wouldn't work. They broke 1950s conventions, and Fedmart did three times its projected first year sales. Sol Price, Costco21:43
  • Émile Hermès bought out his brother in 1919 and asked his craftsmen what they could make by hand that clients would want today. The answer led to car-era bags. Émile Hermès, Hermès34:00