NVIDIA Explained Like You're 5 — Transcript
Full transcript
- 0:00This dishwasher boy built a $5 empire
- 0:03from the table of a Denny's diner.
- 0:05He pitched the business plan over cheap
- 0:06coffee in 1992, and 34 years later, his
- 0:10company makes $20
- 0:11every single hour.
- 0:13Not from the safe bet, but from a gamble
- 0:15nobody believed in that crushed entire
- 0:17industries and took over the AI space,
- 0:20the gaming industry, and even all of
- 0:22your app algorithms.
- 0:23And just as he started winning, the
- 0:25government moved to [music] bring him
- 0:27down.
- 0:29Jensen Huang was born in Taiwan in 1963.
- 0:33When he was nine, [music] his parents
- 0:35sent him and his brother to America for
- 0:36a better education.
- 0:38The plan was simple. Live with relatives
- 0:40in Kentucky, study hard, build a future.
- 0:45This doesn't look like a school. It
- 0:47looks more like somewhere people get
- 0:48sent after they've done something very,
- 0:50very wrong. That's because
- 0:52>> [music]
- 0:52>> it kind of is. Welcome to Oneida Baptist
- 0:54Institute. You'll learn English,
- 0:56discipline, and how not to complain
- 0:58while cleaning bathrooms.
- 1:00This is not how most tech CEO origin
- 1:03stories start, but it didn't break him.
- 1:05It taught him something no lecture hall
- 1:07could.
- 1:08When the situation is terrible, you
- 1:10adapt [music]
- 1:10or you disappear.
- 1:12He eventually made it to Oregon State,
- 1:14then Stanford, spent years at chip
- 1:16companies,
- 1:17AMD, then LSI Logic, studying
- 1:21semiconductors with a level of obsession
- 1:22that was, frankly, a little alarming.
- 1:25By 1992, Jensen had an idea.
- 1:28Graphics are going to change computing.
- 1:31Games are getting bigger. 3D is getting
- 1:33harder. CPUs can't carry this forever.
- 1:37Someone is going to build a chip for
- 1:38this new world.
- 1:40I think it should be us.
- 1:41Jensen, I have a mortgage, a stable job,
- 1:44and a family that currently believes I
- 1:46make responsible decisions.
- 1:48Good.
- 1:49Then this will be exciting. So, the plan
- 1:51is, quit our jobs, start from zero, and
- 1:54build a chip for a market that barely
- 1:56exists?
- 1:57Correct. I hate that this sounds insane
- 2:00and also kind of logical.
- 2:02They called the company Nvidia.
- 2:04The idea was ambitious. Build one
- 2:06powerful [music] graphics chip that
- 2:07could handle everything. A clean
- 2:10universal chip for the future of gaming
- 2:12hardware.
- 2:13For 2 years they worked. They spent
- 2:15every dollar they had. And in 1995,
- 2:18Nvidia launched its first product, the
- 2:20NV1.
- 2:22Jensen, we have a problem. Microsoft
- 2:25just [music] published the Direct X
- 2:26spec, full triangle based rendering
- 2:28pipeline. Holy crap, the NV1 runs
- 2:31quadratic mapping?
- 2:33Yes.
- 2:34We are the only ones going in the
- 2:35opposite direction. Games are about to
- 2:37be built from tiny flat triangles?
- 2:40Our chip is built for curved surfaces.
- 2:42We built the wrong chip with tremendous
- 2:44confidence. 2 years, every dollar. A
- 2:47chip built for a world that no longer
- 2:48existed. The NV1 [music] flopped.
- 2:51But Nvidia still had a lifeline. Sega
- 2:54had contracted them to build the
- 2:55graphics processor for the company's
- 2:57next generation console,
- 2:59the Dreamcast.
- 3:01Real money, real chance to start over.
- 3:04There was only one problem.
- 3:06Sega's chip used the same dead-end
- 3:08technology. [music] You're doing the
- 3:09Sega math again, aren't you?
- 3:11We'll spend the next year building
- 3:12something we already know is dead. And
- 3:14if we cancel, we lose the only customer
- 3:17still paying us.
- 3:18Uh, Pretty much.
- 3:21I miss normal employment.
- 3:23If we're going to die, let's die
- 3:25building the right chip.
- 3:26>> [music]
- 3:26>> So Jensen flew to Japan.
- 3:29The chip is a dead end. You should
- 3:31cancel our contract and find someone
- 3:32else.
- 3:33We shouldn't finish the contract. It
- 3:35would be a waste of your money.
- 3:36>> [music]
- 3:38>> There's one more thing. I still need the
- 3:40money, the $5 million remaining on our
- 3:42contract. Please put it into Nvidia as
- 3:44an investment instead. Otherwise, we'll
- 3:46vaporize overnight.
- 3:47>> [music]
- 3:48>> I have nothing to offer you. This money
- 3:50will most likely be lost, but I'm asking
- 3:52anyway.
- 3:55I'll need a few days.
- 3:56Jensen [music] bowed, walked out, and
- 3:58got on a plane. Nothing to do now but
- 4:01sit, and wait, and watch the Pacific
- 4:03stretch out beneath him.
- 4:05Jensen, you might want to see this.
- 4:08He opens [music] it, eyes wide. It's a
- 4:10massive 10% offer on the value of
- 4:13BetterHelp.
- 4:15Just kidding.
- 4:16Let's be real for a second. When the
- 4:18outcome is completely out of your hands,
- 4:20the anxiety Did I make the right call?
- 4:23That noise won't switch off. It doesn't
- 4:25go away on its own.
- 4:27The hardest part for me was noticing the
- 4:29pattern.
- 4:31I was pushing through instead of
- 4:33actually dealing with anything.
- 4:35My therapist asked me one question.
- 4:37Are you solving something, or are you
- 4:39rehearsing a disaster?
- 4:41I didn't have an answer. Just having a
- 4:43name for what I was doing,
- 4:45that single reframe broke the loop.
- 4:48BetterHelp, the paid [music] partner of
- 4:49today's video, makes starting therapy
- 4:51easier.
- 4:52Take a short quiz, get matched with a
- 4:54credential therapist, and meet by phone,
- 4:56[music] video, or text. Whatever works
- 4:59for your schedule.
- 5:01If it's not the right fit, switching
- 5:02costs nothing extra.
- 5:05And it's working.
- 5:06Over 6 million people have gotten help
- 5:08[music] through BetterHelp, rated 4.8
- 5:10out of five on the App Store.
- 5:13Click the link in the description,
- 5:13[music]
- 5:14or go to betterhelp.com/crowncapital
- 5:16for 10% off your first month of therapy.
- 5:19Every great ship [music] needs
- 5:20maintenance. So do you.
- 5:25Hello.
- 5:26The money is coming.
- 5:28The reason the Sega CEO said yes had
- 5:31nothing to do with business logic. He
- 5:33just liked Jensen.
- 5:35That $5 million bought Nvidia 6 months.
- 5:38And now they had one last shot to build
- 5:40the right chip, or
- 5:41>> [music]
- 5:41>> go bankrupt.
- 5:44They called it the Riva 128,
- 5:46triangle-based, [music]
- 5:47DirectX compatible, built from scratch.
- 5:51Weeks of late nights, bad coffee, and
- 5:53absolutely zero plan B. They had exactly
- 5:561 month of payroll [music] left in the
- 5:58bank to find out if it worked.
- 6:01The chip launched in 1997, and this time
- 6:04Nvidia did not miss. Game developers
- 6:06could actually use it. Customers
- 6:08actually bought it. From the brink of
- 6:09bankruptcy to a million units sold in 4
- 6:13months.
- 6:14Most CEOs would have taken a long
- 6:16vacation and slow down a bit.
- 6:18He walked back into the office and wrote
- 6:20one word on the whiteboard.
- 6:22More.
- 6:24So, by 1999, Nvidia launched the GeForce
- 6:27256 and gave the world a term it didn't
- 6:29really have yet, GPU.
- 6:32Graphics processing unit.
- 6:34It was a genuine leap. A chip that
- 6:36handled complex 3D calculations that
- 6:38previously required an entirely separate
- 6:40processor. For gamers,
- 6:42>> [music]
- 6:43>> this meant smoother worlds, better
- 6:45lighting, and monsters that looked
- 6:46slightly less like wet cardboard. Gamers
- 6:49loved it. Developers wept [music] with
- 6:51joy.
- 6:52Revenue exploded. And by the early
- 6:542000s, Nvidia had become the king of PC
- 6:57graphics. Every serious gamer, every
- 7:00visual effects studio, every game
- 7:02developer on the planet. The obvious
- 7:04move was simple. Keep making better
- 7:06gaming chips, sell them to gamers,
- 7:08become rich, buy a very large leather
- 7:10jacket museum.
- 7:11But Jensen started looking at something
- 7:13other people missed.
- 7:15A CPU is like one very smart employee
- 7:18doing tasks one by one extremely fast.
- 7:21Meanwhile, a GPU is like 10,000 interns
- 7:24doing [music] tiny calculations at once.
- 7:26Are the interns smart?
- 7:28No.
- 7:29But there are a lot of them, and if your
- 7:31problem can be broken into thousands of
- 7:32small calculations, physics, chemistry,
- 7:35weather, biology, artificial
- 7:37intelligence,
- 7:39suddenly, the chip built for video games
- 7:41[music] starts looking like something
- 7:42else, a cheap supercomputer.
- 7:45Jensen looked at this and thought, "What
- 7:47if the same chip that rendered
- 7:48explosions in video games could help
- 7:50simulate molecules, predict weather,
- 7:53model fluids, and maybe one day train
- 7:55machines to think?"
- 7:57So, in 2006, Nvidia launched CUDA,
- 8:00Compute Unified Device Architecture, a
- 8:03platform that let scientists,
- 8:04researchers, and engineers use Nvidia
- 8:07GPUs [music] for general-purpose
- 8:08computing.
- 8:10And then, he committed roughly $500
- 8:12million to build it.
- 8:14Jensen,
- 8:16the gaming business is printing money.
- 8:18Walk me through who CUDA is actually
- 8:20for.
- 8:21Researchers, scientists, people running
- 8:23fluid dynamics simulations, climate
- 8:26models, drug discovery?
- 8:28So, not our customers.
- 8:30Not yet.
- 8:32I'm starting to hate that phrase. This
- 8:34is our CUDA revenue after 2 years. Do
- 8:36you see a number?
- 8:37Not today.
- 8:39Jensen, this is not a business. We are
- 8:42subsidizing a hobby for PhD students.
- 8:45PhD students are where the future
- 8:47starts. For the first time, they can use
- 8:49a gaming chip like a supercomputer.
- 8:51We don't know what they'll build with
- 8:53it. But when they build [music]
- 8:54something important, it will run on us.
- 8:56If they build nothing,
- 8:58we still have the best gaming cards in
- 9:00the world and a $500 million science
- 9:03project. We've survived worse.
- 9:06>> [music]
- 9:06>> Six years passed. The gaming business
- 9:08kept printing money. CUDA kept printing
- 9:11almost nothing.
- 9:12And Jensen kept waiting for the problem
- 9:14big enough to prove him right.
- 9:16Then, in Toronto, someone found it.
- 9:20It's 2012. Every year, the world's top
- 9:23AI researchers compete in a challenge
- 9:24[music] called ImageNet. Tens of
- 9:27millions of images, and your software
- 9:29has to identify what's in them.
- 9:31You see the Toronto submission? Some PhD
- 9:34student [music] who running a a network
- 9:35on gaming cards. We've been hand-coding
- 9:37classifiers for a decade. Gaming cards
- 9:40aren't going to crack this.
- 9:41Ladies and gentlemen, thank you for
- 9:43waiting.
- 9:44The ImageNet [music] 2012 results are
- 9:46final. Second place, University of Tokyo
- 9:50with a 26.2% error rate. And this year's
- 9:53winner,
- 9:54University of Toronto, AlexNet, 15.3%.
- 9:59I'm sorry, 15?
- 10:00That's not a win. That's a murder scene.
- 10:03For years, researchers have been trying
- 10:05to teach computers how to see by
- 10:07manually designing the features they
- 10:09should look for.
- 10:10Edges, textures, shapes, patterns.
- 10:14AlexNet took a different path. Give the
- 10:16neural network enough data, enough
- 10:18layers, and enough computing power,
- 10:21and let it learn the features itself.
- 10:24Training a neural network requires
- 10:26millions of small calculations happening
- 10:28at the same time.
- 10:29Exactly the kind of work that Nvidia had
- 10:32been quietly perfecting for 6 years.
- 10:35The $500 million bet
- 10:36>> [music]
- 10:36>> had finally found its customer.
- 10:38Suddenly, every major AI lab started
- 10:40paying attention. Stanford, MIT, Google,
- 10:45Meta, Microsoft.
- 10:47They weren't just buying graphics cards
- 10:49[music] anymore.
- 10:50They were buying computing power for
- 10:51intelligence itself.
- 10:53And Nvidia had the hardware, the
- 10:55software, and the scale to deliver it.
- 10:58Great news for Nvidia, until in 2022,
- 11:00[music]
- 11:01Washington made a phone call.
- 11:05Before Washington called, Jensen had
- 11:07already tried to fix Nvidia's biggest
- 11:08weakness. Because Nvidia designed the
- 11:10chips,
- 11:12but it didn't make them.
- 11:13That job belonged mostly to TSMC in
- 11:16Taiwan.
- 11:17And he didn't own the underlying
- 11:19architecture, the fundamental
- 11:20instruction set that tells processors
- 11:22how to think.
- 11:24That belonged to a British company
- 11:25called ARM.
- 11:27A company whose technology sat inside
- 11:29virtually every smartphone on Earth.
- 11:31Apple, Samsung, Qualcomm, your pocket.
- 11:35Their DNA, not Jensen's.
- 11:38So, for all of Nvidia's power, [music]
- 11:39Jensen was still renting the building.
- 11:42He owned the furniture.
- 11:43So, in 2020,
- 11:44>> [music]
- 11:44>> Jensen made his move. Nvidia offered $40
- 11:47to buy ARM.
- 11:49Own the GPU, own the software, own the
- 11:52architecture. Simple.
- 11:54Unless you were literally everyone else
- 11:55in tech.
- 11:57Nvidia cannot own ARM. Agreed. Agreed.
- 11:59Also agreed.
- 12:01Wow, I hate how united we are. If Jensen
- 12:04owns ARM, then every chip we build runs
- 12:06through our most dangerous competitor.
- 12:08Phones,
- 12:09>> [music]
- 12:09>> servers, AI chips, everything. So, what
- 12:12do we do?
- 12:13We complain to every regulator with an
- 12:15email address.
- 12:17And they did. The FTC sued. The European
- 12:20Commission investigated. The UK blocked
- 12:22it on national security grounds.
- 12:2418 months later, Nvidia walked away.
- 12:27Then Washington called. Mr. Huang,
- 12:30effective immediately, advanced AI
- 12:32chips, the H100,
- 12:34>> [music]
- 12:34>> can no longer be exported to China.
- 12:37How much revenue are we talking about?
- 12:39That sounds like a you problem.
- 12:41Just like that, some of Nvidia's biggest
- 12:43customers were cut off.
- 12:45Alibaba, Baidu, ByteDance, China's AI
- 12:49companies,
- 12:50gone.
- 12:51So, Nvidia did what Nvidia always did.
- 12:53It engineered around the problem.
- 12:56They built a weaker version, the H800,
- 12:58designed to fit inside the rules.
- 13:01Legal enough to ship. Powerful enough
- 13:03[music] to sell.
- 13:04For about 5 minutes, because by late
- 13:062023, Washington tightened the rules
- 13:08again.
- 13:09The H800 was restricted to
- 13:11>> [music]
- 13:12>> The message was clear. Nvidia was now a
- 13:14strategic weapon in America's technology
- 13:16war with China.
- 13:18Jensen wasn't just a CEO anymore. He was
- 13:21a piece on someone else's chessboard,
- 13:23moved by people he couldn't negotiate
- 13:24with.
- 13:25But while politics got messier, demand
- 13:27got insane. OpenAI, Google, [music]
- 13:30Anthropic, xAI, Microsoft. Everyone
- 13:34building large AI models needed the same
- 13:36thing, more Nvidia chips, more data
- 13:39centers, more power, more cooling, more
- 13:42everything.
- 13:44Blackwell, Nvidia's next [music]
- 13:45generation chip, was being ordered
- 13:47before factories could even finish
- 13:48making it.
- 13:50By October 2025, Nvidia [music] crossed
- 13:52$5 trillion in market value.
- 13:54The most valuable company on Earth,
- 13:56ahead of Google, ahead of Apple.
- 13:59Not bad for a company that once had 30
- 14:01days of cash left and one very wrong
- 14:03chip.
- 14:04None of it was supposed to work.
- 14:06Every single [music] bet looked insane
- 14:08from the outside, and every single one
- 14:10paid off.
- 14:11As what he himself [music] said,
- 14:13"My will to survive exceeds everybody
- 14:15else's will to kill me."
About this transcript
This page contains the full transcript of NVIDIA Explained Like You're 5 by Crayon Capital, generated from the public captions YouTube serves with the video. The transcript has 2,199 words across 421 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.