Transcriptly

How Great Tech Leaders Think and Decide | Ex-Meta CTO & Gigascale Founder, Mike Schroepfer

EO21:00Added Aug 31, 2026

Meet Mike Schroepfer, former CTO of Meta and partner at Gigascale Capital. He joined Facebook in 2008 and spent 17 years at the company, leading its engineering organization for over a decade and later serving as a Senior Fellow.

Watch on YouTube →
Contributed by 刘嘉琪

Transcript

Transcript format

  1. Intro

  2. 00:00Hi, I'm Mike Sheper. Um, I spent 25
  3. 00:02years building, starting and scaling
  4. 00:03technology companies. Worked at a small
  5. 00:05startup. Worked at two small startups
  6. 00:07actually in the dotcom boom. Then
  7. 00:08started my own company after the dot
  8. 00:10crash. Sold that off to a bigger company
  9. 00:12on micros systemystems. Then joined
  10. 00:14Mozilla which made the Firefox web
  11. 00:15browser helped ship 152030 back when
  12. 00:19version numbers were small. They weren't
  13. 00:20in the '7s. And then in 2008 I joined a
  14. 00:22little social network called Facebook
  15. 00:24which was you know smaller than MySpace
  16. 00:26at the time. And then over the next
  17. 00:2814-15 years led engineering, built data
  18. 00:30centers, uh took us into the consumer
  19. 00:32hardware business with virtual reality
  20. 00:34headsets, managed multiple
  21. 00:35billion-dollar acquisitions, built an AI
  22. 00:37research lab, sort of build hardware,
  23. 00:38software, enterprise, and consumer, and
  24. 00:40scaled the team from, you know, 100 to
  25. 00:4235,000ish. I'm now helping great
  26. 00:45entrepreneurs through starting a venture
  27. 00:47capital firm called Gigascale Capital
  28. 00:49where we are hunting for founders with
  29. 00:51big ideas using technological advances
  30. 00:54to build products that makes people's
  31. 00:56lives better. and it solves a climate
  32. 00:58environmental problem and it turns out
  33. 00:59to actually be a great business along
  34. 01:00the way. Easy example of this that is
  35. 01:02you know electric vehicles spew no
  36. 01:04pollution. They're cheaper to operate.
  37. 01:06They're cheaper to maintain. They're
  38. 01:07faster. They're quieter. They have been
  39. 01:09traditionally more expensive. But as
  40. 01:11batteries get cheaper and cheaper,
  41. 01:12they're getting more and more cost
  42. 01:14competitive. And so we're looking for
  43. 01:16products that are cost competitive and
  44. 01:18people love and then also are better for
  45. 01:20human health and for the planet. That's
  46. 01:22a version of what we do.

  47. Scaling Facebook: Building the Backbone Under Pressure

  48. 01:35It was August or September of 2008. We
  49. 01:38had a product that people loved, the
  50. 01:40website facebook.com. More people were
  51. 01:42signing up every day. Those are new
  52. 01:43features being added to the site all the
  53. 01:45time. At the time, the most urgent
  54. 01:47problem was scale. Literally the problem
  55. 01:50was you know every week or every month
  56. 01:51on keeping the site running and the
  57. 01:54software backbone and hardware backbone
  58. 01:56for building this sort of product didn't
  59. 01:58really exist. So a lot of the first many
  60. 02:00years I spent there was scaling and
  61. 02:03rebuilding the software architecture of
  62. 02:05the site and then building the hardware
  63. 02:07infrastructure to do this. You know when
  64. 02:08I first got there we were releasing
  65. 02:10space in those data centers and putting
  66. 02:11servers in them and and building it up.
  67. 02:13But because of the financial crisis of
  68. 02:15the real estate crisis in 2008, people
  69. 02:17had stopped building large data centers
  70. 02:19and so we couldn't get more space. So we
  71. 02:21had to build our own and there was a lot
  72. 02:23of challenges and you know we had to
  73. 02:24learn a lot. None of these things were
  74. 02:25perfect the first time. We made some
  75. 02:26mistakes and we had to fix them and we
  76. 02:28had enough humility to know that we had
  77. 02:31to learn a bunch of new things and so
  78. 02:33you know the first goal was hiring
  79. 02:34people who had done data center work who
  80. 02:36had done network design people with the
  81. 02:37actual background in this space. So the
  82. 02:39goal was go hire a team to understand
  83. 02:41this new area and then help us get good.
  84. 02:44Like many problems in a company, you
  85. 02:45know, if you're a founder or prospective
  86. 02:47founder watching this, most of the
  87. 02:49things a startup does, it does out of
  88. 02:51just complete necessity. It's like,
  89. 02:53well, we can't get space. We don't
  90. 02:54really have a choice here. We have to
  91. 02:56figure this out. One of the lessons that

  92. You Can’t Avoid the Hard Problems

  93. 02:58I took away was there's no getting away
  94. 03:01from the hard problems. You just got to
  95. 03:03get to it. So it's like okay if we need
  96. 03:04to solve this problem let's like figure
  97. 03:06out what are the critical risks what are
  98. 03:07the things we understand the least or
  99. 03:09the highest technical risks and let's
  100. 03:10work on those first. I think there's
  101. 03:12this like human instinct to solve
  102. 03:14tractable problems. So it's kind of like
  103. 03:16I've got this task I want to do but my
  104. 03:18room's kind of messy. I'm going to go
  105. 03:19clean my whole office. I don't want to
  106. 03:20go work on the hard thing. And like
  107. 03:21that's a very human sort of trait. So
  108. 03:24it's like important to work against it.
  109. 03:25Say like actually no it doesn't matter.
  110. 03:27My office is messy. like if I don't get
  111. 03:29this pitch done right and get raise
  112. 03:30money for this company, then nothing
  113. 03:32else matters. Doesn't matter if my
  114. 03:33office is clean or not. So, I think that
  115. 03:35trying to get people focused on the
  116. 03:36right hard problems and running at them
  117. 03:38is is really critical.

  118. Why Meta Went All-In on AI

  119. 03:43Early years of Facebook, it was sort of
  120. 03:45just pants on fire all the time. Wasn't
  121. 03:47totally solved, but it wasn't so much
  122. 03:49consuming all of our energy that we had
  123. 03:50a little bit of energy to look forward.
  124. 03:532013, Facebook started the AI research.
  125. 03:56One of the questions was, you know,
  126. 03:57should you establish a more
  127. 03:59broad-reaching research lab? There were
  128. 04:01things like Microsoft research or IBM
  129. 04:04research or others who were kind of like
  130. 04:05could do software theory, could do lots
  131. 04:07of different areas of technology. The
  132. 04:09most preient part of all of this was to
  133. 04:11say AI is such a big thing and so
  134. 04:14impactful that you wouldn't want to
  135. 04:16spend time on these other things. You'd
  136. 04:17want to take all the energy you had and
  137. 04:19focus it on AI. This is a debate Mark
  138. 04:21and I had and I you know I think he was
  139. 04:23the one actually who kind of pushed to
  140. 04:24say like let's make sure we just do AI.
  141. 04:26Part of it was having visibility to what
  142. 04:28was going on in the ecosystem. There was
  143. 04:30you know this the imageet challenge was
  144. 04:32this famous academic challenge of object
  145. 04:34identification that no one was paying
  146. 04:36that much attention to except when the
  147. 04:38first neural net entered the challenge
  148. 04:40and sort of was so much better than
  149. 04:42everything else. It is one of these rare
  150. 04:44moments, you know, there are these
  151. 04:46moments in tech where something shows up
  152. 04:47and it's like, "Oh my gosh, that thing
  153. 04:49is like 10% better than anything, a very
  154. 04:52large amount compared to any other gap,
  155. 04:54and it used neural net's training on
  156. 04:56data." So, it's a different approach.
  157. 04:58You then look at that thing and say
  158. 04:59like, okay, is that at the end of its
  159. 05:02runway in terms of capability or is it
  160. 05:04at the beginning? And you say, well,
  161. 05:05what's powering it? you say what's
  162. 05:07powering it is like the size of the
  163. 05:08neural net, the size of the data set,
  164. 05:11and the amount of computation you can
  165. 05:12give it both in training and in
  166. 05:14inference. And even at the time it was
  167. 05:16like wow, we we can scale all of those
  168. 05:18things by thousands of times easily. And
  169. 05:21so even if we invent nothing new, you
  170. 05:24could get a lot more out of this. And so
  171. 05:26this is what I'd like to say is like
  172. 05:27it's a technology with a lot of runway.
  173. 05:29It's not fully optimized. But but the
  174. 05:31point being is as a technology, it was
  175. 05:33at its infancy in terms of its ability
  176. 05:34to grow and scale. And those are the
  177. 05:36things that I think are really exciting.
  178. 05:38And I think when we've, you know, when
  179. 05:39Meta's invested in them and when I've
  180. 05:40invested in them, that's where you have
  181. 05:42a lot of opportunity versus things that
  182. 05:44have been optimized for 150 years. And
  183. 05:46there's just not a lot of room for for
  184. 05:48improvement.

  185. Take the Leap: Believe Before It Works

  186. 05:50Everything always feels obvious and
  187. 05:53afterthought. And I had definitely seen
  188. 05:55people who at the time something was
  189. 05:57coming out questioned and doubted it.
  190. 05:59And then 3 years later, it was obvious,
  191. 06:01oh, I knew all along this was going to
  192. 06:02be great. And I was like, no, no, you
  193. 06:04didn't. Um, and and I think it's fine.
  194. 06:06It's like human nature to doubt
  195. 06:07something until you can touch and feel
  196. 06:08it. And you had to take that leap. And
  197. 06:10not everyone, you know, beginnings of a
  198. 06:12new technology takes that leap and they
  199. 06:14say, "Well, I I just don't believe it's
  200. 06:15going to scale or work." And sometimes
  201. 06:16they're right. Like sometimes things
  202. 06:18don't work and they don't scale, you
  203. 06:19know, and then every time you get
  204. 06:21something to a useful point, there's
  205. 06:23always some other useful point it hasn't
  206. 06:24done yet. So we you know the very first
  207. 06:26part of AI that started working
  208. 06:27commercially really well was this sort
  209. 06:29of ability to analyze images and to say
  210. 06:31like okay we can start labeling things
  211. 06:33in these images and understand them um
  212. 06:35translation started to work reasonably
  213. 06:37well so you could do text you know from
  214. 06:39one language to another but this idea of
  215. 06:41a sort of chatbot that I could talk to
  216. 06:44that had any semblance of intelligence
  217. 06:47it was better than anything we had
  218. 06:49before but still not very good at the
  219. 06:52time the demos were terrible you know
  220. 06:53you could barely say how many people
  221. 06:55were in this photo? Is there is there a
  222. 06:56cat in this photo? And if you think of
  223. 06:58this from a like what can I do with it
  224. 07:00standpoint, it wasn't obvious like this
  225. 07:03it wasn't good enough to do anything
  226. 07:04with. And so you'd have to look at that
  227. 07:06and say like oh no no that'll get better
  228. 07:09which is hard for a lot of people to
  229. 07:12believe. And then it's not until you can
  230. 07:13touch and feel and it's really the kind
  231. 07:15of the chat GPT moment where most
  232. 07:16consumers actually had their first
  233. 07:18experience with an AI chatbot and you're
  234. 07:20like oh wait it actually can do some
  235. 07:21useful things for me now I believe. You
  236. 07:23know, you have a similar experience with
  237. 07:25self-driving cars. Most people like
  238. 07:27conceptually get scared of them and
  239. 07:28like, "Oh, it' make me super nervous."
  240. 07:30I've taken a lot of people on Whimo
  241. 07:31rides in San Francisco. It's one of my
  242. 07:33favorite things to do. You get in the
  243. 07:34back of the car, like 5 minutes in,
  244. 07:36you're bored. You're like, "Oh, it's
  245. 07:38like a better driver than than most
  246. 07:40distracted, tired humans." Um, and
  247. 07:42you're like now on your phone and you're
  248. 07:43like, "This is boring." And so this is
  249. 07:45the challenge of new technology is I
  250. 07:47have yet to encounter a new technology
  251. 07:49that until I could show it to you in a
  252. 07:51way that was obviously useful that you
  253. 07:53could personally experience. It is so
  254. 07:55easy to doubt it. Once you get to that
  255. 07:56point, you've like captured all the
  256. 07:58value. So a lot of the challenge is like
  257. 08:00how do you identify technologies that
  258. 08:02have an opportunity to get to that point
  259. 08:04but aren't there yet? Um because that's
  260. 08:06where the sort of place to have impact

  261. Three Core Questions for Spotting Breakout Technologies Early

  262. 08:08is. So if I'm looking at a new
  263. 08:10technology trying to decide whether this
  264. 08:11is something that might be a breakout,
  265. 08:14you know, transformative technology, I
  266. 08:15think there's three core things I'm
  267. 08:17looking for. The first and most
  268. 08:18important question is sort of
  269. 08:20understanding the light speed test. As
  270. 08:22far as we know in experimental physics,
  271. 08:24you can't go faster than the speed of
  272. 08:25light. If I was building spacecraft, you
  273. 08:27know, and I was at 99.9% the speed of
  274. 08:30light, there's not a lot of room for me
  275. 08:32to get faster, right? And you say, I'm
  276. 08:33going to make it twice as fast. Like
  277. 08:34that's really, really, really, really
  278. 08:36hard. You know, if I'm at 0.001 Oh, one
  279. 08:38the speed of light. I got a lot of room
  280. 08:40to go before I've hit any theoretical
  281. 08:41limit. And so for most technologies, my
  282. 08:44first question is how far away from
  283. 08:47theoretical maximum is the current
  284. 08:49version of the thing? Like how much
  285. 08:50headroom do you have to scale an
  286. 08:52improvement? Is it thousands of times?
  287. 08:54Is it 1%. Um that's question number one.
  288. 08:57Then question number two is are there
  289. 09:00tailwinds? Are there things that are
  290. 09:02happening that make this technology
  291. 09:04better that you are not working on? And
  292. 09:06usually this means that there's some
  293. 09:08input into the component that improves
  294. 09:10year-over-year without your work. In the
  295. 09:12AI world, for example, the idea of
  296. 09:14getting more compute power was happening
  297. 09:16without our effort because Nvidia, TSMC,
  298. 09:20ASML, the entire chip ecosystem was
  299. 09:22developing faster, more powerful chips
  300. 09:25year-over-year. So for the same dollar I
  301. 09:27could get every 18 months about double
  302. 09:28the compute power. So I didn't need to
  303. 09:30do any work. I didn't need to go build a
  304. 09:31chip team and go do design chips. It's
  305. 09:33just like every year I got more
  306. 09:34computation without me doing a single
  307. 09:36amount of work. So that's question
  308. 09:37number two is do I have some tailwind
  309. 09:39making my product better even when I'm
  310. 09:41asleep. And then number three which is
  311. 09:43the hardest is what problem are you
  312. 09:46solving and how important is it to your
  313. 09:48customer? Your customer could be a
  314. 09:49consumer could be a company but there
  315. 09:51are plenty of amazing technologies that
  316. 09:53have made great advances that don't
  317. 09:55actually solve a problem for people. The
  318. 09:57biggest example of this is 3D TVs. For a
  319. 09:59long time, everyone's like, "Tre is it's
  320. 10:01the next step forward from 2D. It's more
  321. 10:03immersive." It turns out people don't
  322. 10:05want to put like special glasses on, you
  323. 10:07know, and sit in their couch and do it.
  324. 10:09So 3D TVs were a technological leap that
  325. 10:12didn't solve a problem for consumers
  326. 10:14that they cared about. So it's been
  327. 10:15mostly a failure. What we do with
  328. 10:17advanced technologies is you try to mock
  329. 10:19them up and say like, "Okay, I haven't
  330. 10:21built the thing yet. what's the best
  331. 10:22sort of proxy of the thing that I can
  332. 10:24use to show people to get a sense of if
  333. 10:27I built this thing would you like it
  334. 10:29that is that customer exploration is is
  335. 10:31really important because ultimately
  336. 10:34someone has to buy that technology some
  337. 10:37business model have to pay for it
  338. 10:38someone has to buy it to fund all the
  339. 10:40R&D and if there isn't some loop of
  340. 10:42money there somewhere it'll eventually

  341. Why I Left Big Tech to Break the Bottleneck to Progress

  342. 10:53To me, the problem was obvious. I mean,
  343. 10:55it's something that I've been passionate
  344. 10:57about for a long time. I had the very
  345. 10:59first Nissan Leaf, which is the first
  346. 11:00electric vehicle I could buy. Early
  347. 11:02generation products. It was terrible.
  348. 11:03Had very bad range. Really, the question
  349. 11:05was, is there really anything I can do
  350. 11:07about it? It just it feels like a big
  351. 11:09overwhelming problem. And it takes
  352. 11:11either some hubris or some naive to
  353. 11:14believe that I can actually have a
  354. 11:16meaningful impact on it. And I think
  355. 11:18what I decided at some point was it
  356. 11:20didn't really matter whether I could. I
  357. 11:22had to try. And when I think about the
  358. 11:23problems humanity needs to go tackle,
  359. 11:26massive amounts of additional clean
  360. 11:27energy is upstream of everything we want
  361. 11:29to do. You know, if you sort of take a
  362. 11:31longer view of the industrial
  363. 11:32revolution, really what humanity has
  364. 11:34done is we've harnessed energy, whether
  365. 11:36it's animals, whether it's fuels or
  366. 11:37sometimes renewables do work for us. So
  367. 11:40instead of manually farming or manually
  368. 11:43digging, you know, we have machines that
  369. 11:44do it for us. And that has had a huge
  370. 11:46uplift in productivity and in human
  371. 11:49health and happiness. You know, the only
  372. 11:51way we're going to get AI progress is by
  373. 11:53massively increasing energy use. The
  374. 11:55only way we're going to get people in
  375. 11:57comfort and air conditioning with clean
  376. 11:58water. It's it's energy. That is the
  377. 12:01upstream problem to everything you look
  378. 12:02at. We know how to desalinate water and
  379. 12:04make clean water. We know how to keep
  380. 12:05people cool on a hot day. We know how to
  381. 12:07manufacture lots of different things. We
  382. 12:09know how to make super intelligent
  383. 12:10assistance. We don't know how to scale
  384. 12:12that to 8 billion people without
  385. 12:14terowatts of additional clean energy and
  386. 12:16there are lots of ways to go solve that
  387. 12:18problem and lots of entrepreneurs off
  388. 12:20tackling solutions that can that can
  389. 12:22have an impact in the world at that
  390. 12:23scale. And it took me a little bit of

  391. Startups Will Solve Sustainability

  392. 12:26time to figure out exactly the mechanism
  393. 12:28to do so. And and what I figured out was
  394. 12:30solving sustainability is is
  395. 12:32re-engineering tens of trillions of
  396. 12:34dollars in our economy. You know, energy
  397. 12:35alone is a multi-t trillion dollar
  398. 12:37business. And if you want to do that,
  399. 12:39you can't do it with government money.
  400. 12:40You can't do it with philanthropy. You
  401. 12:42need businesses to be investing. And the
  402. 12:45place to do that is usually through
  403. 12:47startups. Is you've got this cohort of
  404. 12:49really amazing entrepreneurs out there
  405. 12:51chasing ideas from fusion to next
  406. 12:53generation micro reactors to offshore AI
  407. 12:56data centers, dehydrators for your
  408. 12:58kitchen. And that my personal experience
  409. 13:01in building companies over 25 years was
  410. 13:04just uniquely valuable. As I spent time
  411. 13:06with entrepreneurs, I realized there was
  412. 13:08a lot I could do to help them skip over
  413. 13:10the common mistakes you make when
  414. 13:11building a team. How do I hire
  415. 13:12executives? How do I manage sort of
  416. 13:14product development? All of these sorts
  417. 13:15of things. And so, it's just a great
  418. 13:17coalignment with my skills and sort of
  419. 13:19the change I want to see in the world.
  420. 13:21And it's just awesome to see these
  421. 13:22products take off in the market. And so
  422. 13:24there's a lot of exciting work happening

  423. Better, Faster, Cheaper: Products People Actually Love

  424. 13:26in clean power that is effectively
  425. 13:28unlimited, meaning we can 10x, 5x, you
  426. 13:31know, multiple orders of magnitude above
  427. 13:33what we're using today worldwide in
  428. 13:34power and power the whole planet. The
  429. 13:36most exciting and most ambitious of this
  430. 13:38is fusion. The idea of basically the the
  431. 13:41power source of our sun. We know it
  432. 13:43works in the universe. We've actually
  433. 13:44made it happen on planet Earth before.
  434. 13:46We know how to make fusion work. We just
  435. 13:47haven't yet figured out how to turn it
  436. 13:49into a reliable power source. If I can
  437. 13:51make it work, I can build a power plant
  438. 13:53that requires almost no inputs and
  439. 13:55produces no emissions. It's completely
  440. 13:58safe and we can build kind of unlimited
  441. 14:00these power plant that could power all
  442. 14:02of Austin, Texas, would require one
  443. 14:04pickup truck a year of fuel, which is
  444. 14:06absolutely insane. That would take train
  445. 14:08loads of of coal cars or amazing amounts
  446. 14:10of gas to do the equivalent sort of
  447. 14:12generation. So, from an efficiency
  448. 14:13standpoint, it's sort of the endgame for
  449. 14:15power generation. So this is energy and
  450. 14:16I could talk a lot more about those but
  451. 14:18let's talk about you know just one other
  452. 14:20example of something that people
  453. 14:21watching this might experience at home
  454. 14:24which is um you know the idea of
  455. 14:25throwing away food in the garbage. When
  456. 14:27you throw food in the garbage it goes to
  457. 14:29a landfill. It rots, releases methane.
  458. 14:31This is a major source of near-term
  459. 14:33warming and waste. There's a company
  460. 14:34called Mill. It's a little trash can. It
  461. 14:36looks like a little pop-up trash can. Um
  462. 14:38but it's magic. You throw food into it.
  463. 14:41It dries it up, grinds it up, turns it
  464. 14:43into what looks like little coffee
  465. 14:44grounds. You can do this in an average
  466. 14:46family home for probably a month and at
  467. 14:48the end of that month you have a shoe
  468. 14:49box size of these grounds. Most
  469. 14:51importantly, it doesn't smell. You don't
  470. 14:53have to empty it for a month. And so the
  471. 14:54the pitch to consumers is empty your
  472. 14:57trash less and it stinks less, you know,
  473. 14:59and I kind of laugh. Nobody really
  474. 15:01enjoys emptying their trash. And so I
  475. 15:02would say you can do it less and it's
  476. 15:04smaller and it stinks less then
  477. 15:05everyone's excited. And it turns out
  478. 15:07this is a major diverter of food waste
  479. 15:10related emissions. And the thing about
  480. 15:12this product is if you meet someone who
  481. 15:13has it, you'll know it because they love
  482. 15:15it. It's a product that people buy
  483. 15:17because they love and it saves a food
  484. 15:19waste problem, saves a consumer problem.
  485. 15:21And guess what? This company is making
  486. 15:23great revenue and and money on us. This
  487. 15:25is an example of a better, faster,
  488. 15:27cheaper. It makes people's lives better,
  489. 15:29and it solves a climate environmental
  490. 15:30problem. And it turns out to actually be
  491. 15:32a great business along the way.

  492. Technology Matters — People Matter More

  493. 15:39People have operated big things. people
  494. 15:41started companies very few people have
  495. 15:42seen the sort of the scale you know I
  496. 15:45built tens of millions of square foots
  497. 15:47of data center space we've shipped tens
  498. 15:48of millions of consumer hardware
  499. 15:50products you know scale teams to tens of
  500. 15:52thousands managed lots of multi-billion
  501. 15:54dollar acquisitions so you know I've had
  502. 15:56the great fortune of working with an
  503. 15:58absolutely incredible cohort of people
  504. 15:59on an amazing set of technologies in
  505. 16:02hardware in software in deep research
  506. 16:04and and others all everything we've
  507. 16:05talked about here is is what we bring to
  508. 16:07bear was how do you find the right
  509. 16:08problem identification the technology
  510. 16:10that has headroom to scale and tailwinds
  511. 16:12and customer demand. And then what we
  512. 16:14haven't talked about is is people. You
  513. 16:15know, a lot of my job ended up being
  514. 16:18finding out who were the right leaders,
  515. 16:20technical, organizational, otherwise to
  516. 16:22take something forward. And in the
  517. 16:23startup realm, the team is ultimately
  518. 16:25what you're betting on. You know, these
  519. 16:27are the people who are going to build
  520. 16:28that company. And we're looking for
  521. 16:30founders who could take the company as
  522. 16:32far as possible. Meaning the challenge
  523. 16:34of a company is a company at 10
  524. 16:35employees at preede or seed is a very
  525. 16:38different company than a 300 person
  526. 16:40company with customers in the series C
  527. 16:42and that rate of change is unusual for
  528. 16:45humans like you don't usually encounter
  529. 16:47environments that change that much and
  530. 16:48so there is a rare set of people who can
  531. 16:51scale through those changes and I've had
  532. 16:53the great fortune of working with many
  533. 16:54of them I've had the ability to do it
  534. 16:56myself and so a lot of what I'm looking
  535. 16:58for is that people identification of
  536. 16:59like as we meet thousand founders a year
  537. 17:02these are the 10 for this year that we
  538. 17:04think have the best shot at scaling this
  539. 17:06company into a public company and that
  540. 17:08is a lot of the sort of that pattern
  541. 17:10matching on people's ability to scale is
  542. 17:13a lot of what we're doing. So coupled
  543. 17:14with technology and market, you know,
  544. 17:16it's it's really people.

  545. What Winning Founders Have in Common

  546. 17:18I think what we see in founders, you
  547. 17:21know, there's questions about how do you
  548. 17:22evaluate it? What we look for in
  549. 17:23founders is number one, you need a
  550. 17:26complete relentlessness, a
  551. 17:28determination. Building a company is a
  552. 17:30never-ending series of near-death
  553. 17:32disasters and a lot of people telling
  554. 17:34you what you're doing isn't going to
  555. 17:35work and a lot of people saying no. New
  556. 17:37recruits say no, investors say no,
  557. 17:39customers say no. And you need to get 30
  558. 17:41nos in a row. It doesn't really matter
  559. 17:43if 30 investors say no if one says yes.
  560. 17:45I mean, I remember from 25 years ago
  561. 17:47going out on Sand Hill Road trying to
  562. 17:49convince people to invest in my startup
  563. 17:50as a first-time founder. And I got a ton
  564. 17:52of nos. You know, I had someone fall
  565. 17:54asleep in one of our pitch meetings. But
  566. 17:55then we got the world's best venture
  567. 17:57capital firm, Sequoia Capital, to say
  568. 17:59yes. And that was the defining moment,
  569. 18:01you know, and that that helped us build
  570. 18:02a really successful company. And so, as
  571. 18:04a founder, you need to have this
  572. 18:05determination to just like keep going
  573. 18:07despite setbacks. That's number one.
  574. 18:09Number two is building a company as a
  575. 18:11CEO is a different job every single day.
  576. 18:14You might have to solve technical
  577. 18:16problems. You might then have to go talk
  578. 18:17to customers. You might have to go
  579. 18:18recruit people. You might have to get a
  580. 18:20lab space. You're going to be doing a
  581. 18:22different job every single day. And
  582. 18:24there's a category of people I call them
  583. 18:26like just consumers of new information.
  584. 18:28And they have this combination of
  585. 18:30humility that they don't know something
  586. 18:32and curiosity figure out how to learn
  587. 18:34it. And that combination allows them to
  588. 18:37do everything like how do I run a board
  589. 18:39me? How do I pitch an investor? Like
  590. 18:41these are all things our founders
  591. 18:42learned how to do successfully. And so
  592. 18:44you're looking for people who
  593. 18:46demonstrate this ability to decide they
  594. 18:48don't know something and then figure out
  595. 18:49how to learn it as quickly as possible
  596. 18:52um andor hire people who know how to do
  597. 18:54it, you know, for their company. So it's
  598. 18:55it's really those two things. this like
  599. 18:57unrelenting determination and this like
  600. 19:00ability to understand, identify and
  601. 19:02learn new domains on a very rapid clip.

  602. There’s No Perfect Founder Checklist

  603. 19:05Well, the challenge is, you know,
  604. 19:06everyone tries to distill it down into
  605. 19:08a, you know, if I could just like check
  606. 19:10off a couple of things in their
  607. 19:11background, you could find the founder.
  608. 19:12But that never works. Like the number of
  609. 19:14times that if you if I make a rule where
  610. 19:16we only do second time founders or this
  611. 19:17and that, I can give examples that
  612. 19:18violates that rule. We just have to meet
  613. 19:20founders and then do our own evaluation
  614. 19:22of it. That is the most important part
  615. 19:23of the job. And we evaluate them by
  616. 19:25meeting them multiple times. We evaluate
  617. 19:27them by calling references and people
  618. 19:28that worked with them. But I'll give you
  619. 19:29a couple of examples of people in clean
  620. 19:31techch that I think are phenomenal.
  621. 19:33You've got most wellunded companies
  622. 19:34working on fusion come fusion systems
  623. 19:37founded by Bob Mumgard who's a plasma
  624. 19:39physicist. This is his first company.
  625. 19:41He's never worked at a company before.
  626. 19:42You know, if I told you to give me a
  627. 19:44resume for a CEO of a thousand person
  628. 19:47company, a plasma physicist is probably
  629. 19:48not what you would search for. But when
  630. 19:50you meet him, he is an operator. He has
  631. 19:52learned very quickly how to build a
  632. 19:54team, how to rely on others, how to tell
  633. 19:55a story, how to raise money. He is
  634. 19:57absolutely phenomenal. What a CEO needs
  635. 19:59to do is like describe in deep clarity
  636. 20:02the mission of the company and what's
  637. 20:03important and get a large number of
  638. 20:05people on board and focused in that
  639. 20:06direction and he does an exceptional
  640. 20:08job. And then you have people like Matt
  641. 20:10Rogers at Mill. This is his second
  642. 20:11company. His first company was Nest and
  643. 20:13he was sort of great along the way but
  644. 20:14he didn't rest in his own laurels. He
  645. 20:16did what great founders do which is
  646. 20:18build a great team around him. What we
  647. 20:20look for in founders is, you know, the
  648. 20:21job of building a company is solving
  649. 20:24hard problems that no one's ever solved
  650. 20:25before that you don't actually totally
  651. 20:26know how to solve. And so when we met
  652. 20:28Mill, they said, "Oh, we're going to get
  653. 20:30approval from the US government to take
  654. 20:32this return food waste and turn it into
  655. 20:34chicken feed." And uh, you know, we're
  656. 20:36going to get it by X time. We're not
  657. 20:37exactly sure how to do it, but we're
  658. 20:38going to get it done. Like 2 or 3 months
  659. 20:40later, they're like, "Yep, we got it
  660. 20:40done. It's proofed. We're we're now
  661. 20:42doing it." And it's just like a series
  662. 20:43of like we're going to go after this
  663. 20:45problem and then we're going to go solve
  664. 20:46it. each individually are amazing, but
  665. 20:47as a team figure out how to go take down
  666. 20:50big problems. And that's the magic of a
  667. 20:51of a startup.