This message comes from NPR SponsorShipBob, Power Your Warehouse with ShipBob's warehouse management system to improve order for fulfillmentefficiency, ship ordersfaster, and reduce operatingcosts, all through one WMSplatform.
Go to ShipBob.com for a freequote.
This is Planet Money fromNPR.
Alright, we have a riddle foryou.
See if you can guess what we're talking abouttoday.
I was born at a Denny'srestaurant.
My name is Latin forEnvy.
I played a lot of video games in my youngeryears.
Today I'm more into poetry andtranslation.
I'm 31 yearsold, and I've been down and out so many times that my unofficial motto is that I'm 30 days from going out ofbusiness.
Still don't haveit? What if wesaid, for amoment, I was the most valuable company on theplanet.
Yeah, that'sright, the answer is the computer chip designer invideo.
And Envy's chips are the leading choice for artificial intelligencedevelopers.
In the last couple ofyears, Envy's sales have shot up like a jetplane.
It's stockprice, more like arocket.
But there are several chip companies outthere, like AMD andothers.
So why is Envy the one dominating the AIscene?
Hello and welcome to PlanetMoney.
I'm DarrenWoods. And I'm Weel andWong.
In video went from selling niche graphics cards for gaming to becoming one of the biggest companies in theworld, with help from another techgiant.
Today on theshow, two stories about innovators who went up against the statusquo.
And in theprocess, helps transform an industry with all come to relyon.
This message comes from in PR sponsorBetterHelp.
Social media makes it easy to compare your life to everyoneelse's.
But therapy can help you focus on yourselfinstead, so you can live your life to thefullest.
If you're consideringtherapy, give BetterHelp atry.
It's entirelyonline, designed to beconvenient,flexible, and tailored to yourschedule.
Visit BetterHelp at BetterHelp.com slash NPR to get10% off your firstmonth.
This message comes from in PRsponsor, CapitalOne.
Capital One offers checking accounts with no fees or minimums and no overdraftfees.
That's bankingreimagined.
What's in yourwallet? Termsapply.
See capital1.com slashbank.
Capital OneNA, memberFDIC.
In video is often called a chipmaker.
That's a bit of anoversimplification.
First ofall, in video doesn't actually manufacture the semiconductorchips.
That's done by companies like TSMC andSamsung.
In video does thedesigning, like where the circuits all go on thechip.
It also makes the software so that developers can work with thosechips.
This combination has its roots in1992.
Jensen Huang and two other engineer friends met at a Denny's restaurant in SanJose.
Their idea was to improve video games by building specialized chips for rendering 3Dgraphics.
This decision would become incrediblylucrative.
One person who spent a lot of time unirthing Nvidia's history is DavidRosenthol.
David is one half of the tech podcastacquired, which made this epic series of podcasts on the company'shistory.
Nvidia corporate communications got in touch andsaid, who were yoursources?
Who told you all this in thecompany?
And we'relike,well,nobody.
Wejust, youknow, we watched a lot of YouTube videos withJensen.
And we read a lot ofpapers.
This led to an in-person interview with Jensen Huang towards the end of lastyear.
When did you realize thebelief?
I didn't learn about it until I was toolate.
By all accounts working for Jensen is super intense and not foreverybody.
Failures are sharedpublicly, forexample.
But David says those who stay areloyal.
Jensen is truly singular in SiliconValley.
And I think the man is 60 yearsold.
Nobody from his generation is still running their company in the same way that heis.
He is probably more engaged in every detail in Nvidia than any of those other founders ever were at theircompanies.
David says the way that Jensen is involved in the details is one key to why Nvidia rode the AI wave soastutely.
He started moving the Nvidia ship in the direction that it is today with AI and machinelearning, starting in like 20 yearsago.
It's incredibleforesight.
It's obviously way before ChatcheeBT, but it's even before driverless cars or even voicerecognition.
How did he see thiscoming?
Well, I think this is the key toJensen.
If you were to say that sentence tohim, I think he would respond andsay,no, it was not before all thosethings.
All of those things werehappening.
They were just happening in the deep scientific computer science researchcommunity.
He was so deeply plugged into all of this that he knew the principal actors who now are some of the true leaders in AI research and open AI and they're out of like an elsewhere thesedays.
He knew thempersonally.
He was reading the whitepapers.
He was visiting them at universities all back in the mid2000s.
At thistime, Nvidia was mainly known by 15 year olds upgrading their gamingcomputers.
But in the early2000s, Jensen essentially made a huge bet on selling super computing power to a wider range ofpeople.
And as part ofthis, the company launched a software development framework calledCUDA.
CUDA acts kind of like a middleman between the software developer and thechip.
And the system would be crucial for AI after a turning point in2012.
This was the big bang moment that kicked off research and investment at a commercial scale into artificialintelligence.
It was at this annual competition for researchers to submit AI systems that could recognize images from a massivedatabase.
The database was calledImageNet.
And progress had beengradual.
Entries could correctly identify what a majority of imageswere.
But would typically have an error rate of around 25 or30%.
Thatyear, a team from the University of Toronto submits this entry that they callAlexNet.
It blows away the rest of thefield.
So the percentage of images it got wrong was around15%.
1.5%. This was a quantumleap.
The team used an existing method of AI called a neuralnetwork, which waspowerful.
But it had always been limited by how much computing power it required totrain.
To overcomethis, AlexNet did somethingdifferent.
The standard approach had been to train these models using central processing units orCPUs.
The generalist brains in yourcomputer.
But these calculate their instructionssequentially.
You can think of it like one calculation thananother.
Soinstead, this team ran these training calculations using graphics processing units GPUs what you would play video gameswith.
A graphics card handles tens of thousands of instructions at a time that I've really widened the pipe of the amount of compute that I can stuff through this thing at any given point intime.
And two graphicscards?
Invideo, ofcourse. People at Google and people at Facebooksaid, holycrap, you could use these image classification systems to build way better social media feedrecommenders.
So for imagerecognition, programmers were training the software to notice patterns and say pictures ofcats,whiskers,fur, fourlegs.
Andsimilarly, the big social media companies realized they could train software to notice patterns in what kinds of pictures and posts and movies people like on theinternet.
That was billions and billions of billions of dollars ofprofits.
So there was a good 10 year run where like nothing elsemattered.
And that meant that places like YouTube and Facebook and Instagram were hiring a lot of AItalent.
And David says this was actually part of the motivation for Elon Musk and Sam Altman to foundOpenAI.
They were really worried that Google and Facebook had just become a duopoly of all the AI development and research talent because it was theonly, youknow, economically viable usecase.
OpenAI's launch of chat GBT in late 2022 was the next hinge moment for AI and forInvidia.
It woke investors and everyday people to advancements in AI that got them dreaming about thefuture.
It's where Invidia suddenly started its journey becoming one of the most valuable companies on theplanet.
In the parlance of techinvesting, Invidia has this giantmoat.
David reckons it's protected from competition in the foreseeable future because ofCUDA, that developmentsystem.
These systems build on top of themselves over time and get more and more complex andpowerful.
So it would be like somebody going and starting a new phone operating system from scratch andsaying,okay, what are all the list of things we need to build to make this phone operating system viable to compete with Android andiOS?
That's a tallorder. David says network effects are alsoimportant.
Now millions of other developers useCUDA.
So if you're a collegestudent, that's the language you're going tolearn.
It's a self-reinforcingcycle.
This moat is something that has gotten people speculating about governmentaction.
News outlets like The New York Times have reported that the Department of Justice has cleared the way for possible antitrust action againstInvidia.
Jensen though has said he doesn't like the word moat and prefers to think of the company as working alongside the entire AIecosystem.
Whichever words you use to describeit, Invidia is standing on top of the world rightnow, though even with all its chipsales, its stock price is very highly priced by conventionalmeasures.
One misstep and there is a long way tofall.
Are the way into the moat filled withbarricudas?
Barricudas! That's quitefunny.
If you want to hear more onInvidia, check out our next Planet Money Bonusepisode, which is out in a couple ofweeks.
It'll feature my extended interview withDavid, where we're going to more detail on thecompany, on CUDA and JensenQuang.
That's just for Planet Money Plussupporters.
It's one of the perks you get for signingup, including sponsor free listening and access to exclusivemerch.
Go to plus.npr.org slash Planet Money formore.
After thebreak, another unlikely origin story that made Invidia's risepossible, how America's advanced chipmaking ended up inTaiwan.
This message comes fromnprsponsorhomes.com.
When you're home shopping as aparent, you have lots of questions about localschools.
That's why each listing on homes.com includes extensive reports on localschools, includingphotos, parentreviews, student teacherratio, schoolrankings, andmore.
The information is from multiple trusted sources and curated by a dedicated in-house researchteam.
It's also you can make the right decision for yourfamily.
Homes.com, we've done yourhomework.
This message comes fromnprsponsorhomes.com.
Newm understands that not everyone is starting from the same place and takes that intoaccount.
With their first evercookbook, The NoomKitchen, you can find 100 healthy and delicious recipes to promote betterliving, available to buy now wherever books aresold.
This message comes fromnprsponsorhMintMobile.
From the gas pump to the grocerystore, inflation iseverywhere.
So MintMobile is offering premium wireless starting at just $15 amonth.
To get your new phone plan forjust $15, go to mintmobile.com slashswitch.
This episode sponsor isCarvana, which offers the followingmessage.
Whether you're looking to sell your car in a hurry or thinking about parting ways with your trustyride, Carvana is the most convenient way to sell yourcar.
Go to Carvana.com to get an offer for your vehicle inseconds.
I'm Adrian Ma and I'm DarianWoods.
If you have the latestiPhone, you'll find inside it a computer chip with parts so fine-grained that you can measure them inatoms.
That processing chip contains 16 billiontransistors.
The iPhone chip and most cutting edge chips for computers and cars are made in a handful of factories on a small island in the Pacific inTaiwan.
And the story of how Taiwan came to make the most advanced chips in the world can basically be told through the story of oneman, MorrisCheng.
He's like a founding father of the chip slash semiconductorindustry.
It's not an exaggeration to say thatMorris' personal career obstacles and what he chose to do about them help shape the balance of power between the US and theworld.
When Morris Cheng grew up in China and HongKong, he fled three wars and saw a lot of poverty aroundhim.
So when he got into Harvard at age18, he had one goal to makemoney.
And he looked around at what a middle-class kid from China could do to make a lot of money back in1950.
Engineering. So he left Harvard forMIT, which actually had an engineeringprogram.
I really wasn't all that interested in mechanical engineering or engineering atall.
I was in it for thebox,right?
That'sright. That is Morris Cheng speaking in an interview with Stanford's president in2014.
And Morris eventually gets a job at electronics company Texas Instruments and Morris is promoted reallyquickly.
By the early1970s, he's vicepresident.
But prettysoon, he finds his career stalling and he can't get promotedfurther.
Cheng Tai-shia is an economist at the University of Chicago who was born inTaiwan.
And at onepoint, he even worked at the same organization that Morris onceled.
Cheng Tai says in the early1970s, all chipdesign,manufacturing,testing, and packaging was done in theUS.
And the US was so dominant because semiconductors were inventedhere.
It was prohibitively expensive to get into thisindustry.
You had to come up not only with the expertise that put together thedesign, but then you also had to spend several billion dollars putting up the manufacturingcapacity.
Just a couple of details about how hard it is to manufacturesemiconductors.
Your building has to be constructed so that it doesn't vibrate down to justnanometers.
You cannot build a facility where there's any dust that can fall on yourchips.
It has to be a thousand times cleaner than the hospital operatingtheater.
But there were a few things other companies coulddo.
In the1970s, Morris Cheng realized he could save his company a lot of money by outsourcing simple tasks like packaging and testing toTaiwan.
InTaiwan, incomes were only about a tenth of Americanincomes.
And moresupercharged, the company's growth by pioneering new ways of pricing the computer chips thatwould, youknow, result in a lossnow, but capture the market share as the cost of making chips got cheaper andcheaper.
Oh, and he also ran the division that invented the speakingspell.
Chang Tai-shia says that despite Morris Cheng's achievements for the nextdecade, his career kind of had awall.
He wanted to be chiefexecutive.
And then this is something that he has saidpublicly.
He felt that given the nature of a culture of TexasInstruments, he would never become theCEO.
Whynot? Because he'sAsian.
Imean, that's what he hassaid.
That it was a very whitecompany.
It was a culture there in which people like him don't getpromoted.
Then in1983, after a couple of years ofagonizing, and after 25 years with TexasInstruments, Morris Cheng decided to leave thecompany.
And beforelong, the premiere of Taiwan recruits Morris to run this thing called the Industrial Technology ResearchInstitute.
Basically, the goal is to turn Taiwan into a techpowerhouse.
Think of it as an organization that's partresearch, part government investing in newcompanies.
The culture very much was that of astart.
Cheng Tai worked here in1990.
It wasn't a typical governmentbureaucracy.
They were working 12 hourdays.
It's too muchdepartments.
People get paidnothing.
But there's this dream that it's a combination of getting rich and that you're going to change theworld.
And beforelong, the government of Taiwan realizes we have the semiconductor superstar right in front ofus.
Someone who could have been CEO at one of the very few companies in the world who can make cutting-edgechips.
So theysay, how about instead of helping out othercompanies, you run yourown, a semiconductorcompany.
And I want to pause here for asecond, because this is a pretty bolddecision.
It's a possibly foolishidea, because history books are full of cases where a governmentsays, we want to makewhatever,cars,spaceship, cellphones, and then it just failsspectacularly.
It costs an immense amount ofmoney, it doesn't turn a profit most of thosetimes.
But Taiwan decides to do it with one of the absolute hottest types of businesses to start both then andnow.
So Morris looks around and he realizes there just are not enough scientists who know how to design chips to build essentially the Taiwanese version of TexasInstruments.
So Morrissays, let's try somethingdifferent.
Hethinks, what do wehave?
Well, we've got cheap labor that's getting more and more skilled at advancedmanufacturing.
We've got solid supply lines forfactories.
And then he thinks about how America almost has the reverseproblem.
It's got too many people who can designchips.
So many people who would start their own semiconductorcompany, but building the factory in the US is tooexpensive.
His basic idea was tosay, why don't we develop a company that would only serve othercustomers, in which we completely get out of the designbusiness.
We only do the manufacturing part of thebusiness.
And this would allow us to serve lots of customers at the time their needs were beingmet.
And this simpledecision, this decision to separate manufacturing from design completely changed the chipindustry.
In1987, Morris Chang founded Taiwan semiconductor manufacturingcompany,TSMC.
And over theyears, more and more companies started to useTSMC.
Some big chip makers even sold off the manufacturing parts of their businesses knowing that they could design the chips and TSMC could makethem.
And with the advent of TSMC and its Taiwanesecompetitor,UMC, some startup founder could have a great idea for a new chip like a new graphics card forgaming, but they didn't need hundreds of millions ofdollars.
This is exactly what happened invideo.
It designed semiconductors inAmerica, but TSMC makesthem.
The company like NVIDIA would not have existed withoutTSMC.
So on the onehand,yes, TSMC took a lot of the chip manufacturing business away fromAmerica, but it also allowed entrepreneurial dreams from Silicon Valley to cometrue.
Now, TSMC is the most important semiconductor manufacturing company and the eighth largest company in theworld.
It makes the vast majority of the world's most advancedchips.
It's so critical for global supply chains that US lawmakers arenervous.
In August of2022, President Joe Biden passed a bill called the ChipsAct, which means $39 billion for companies making computer chips inAmerica.
Which is so muchmoney,right?
It's not that much money relative to how much is needed tobuild.
Because each semiconductor plant is about $20billion.
Okay, so you might got to build twoplants,maybe.
Butthis, my prediction is that this is going to be another boomdoggo.
And so this money is supposed to light a fire under the collective butts of the chip manufacturingindustry.
Like what the US is trying is something that Taiwanese government tried about 35 yearsago.
And it worked for Taiwan because Morris had the skills and saw slice of the market he could take it just the righttime.
And the government funded it like a businessinvestment,right?
Not a handout to acompany.
But I'm saying that the only plausible case for something like the Chips Act is if at some point in the future there's war between Taiwan andChina.
And TSMC is either destroyed or TSMC is taken over by the Chinese CommunistParty.
After $6.6 billion of the Chips Act money has been allocated and direct funding for a TSMCfactory.
This one is being built inArizona.
The chip production is set to start in2028.
And the ironyis, in another world where Morris Chang found more open doors in theUS, a company like TSMC could have been American to beginwith.
On the next planetmoney, 40 years ago an airline CEO had anotheridea.
Our concept was everybody wants to get from A toB.
They don't go on an airplane to getfood.
That sounds a lot like my airplaneexperience.
Well, please don't compare us to thespirit.
The story of the first national budget airline and why flying has become the way it istoday.
That's next time on planetmoney.
These stories originally aired on the Indigator from Planet Money and were originally produced by Julia Richie and Viet Le with help from NikkiWillett.
They were edited by Kate Kin Cannon and engineered by Sena Lo Freito and JamesWillett.
Fact checking was done by Taylor Washington and SierraJuarez.
Our planet money edition was produced by WillaRubin, edited by JennyLawton, Fact check by Sophia Schuchena and engineered by NeilRouch.
Alex Skolmark is our executiveproducer.
I'm DarianWoods. This isAmpia.
