1990s
The hosts say 70 to 80 separate graphics card companies received venture funding in Nvidia's early era.
“Yeah, it was 70 or 80 separate companies making graphics cards all got funded.”
Topic
Nvidia
203 facts from 51 episodes, in date order. These are what speakers said on the shows, each linked to the moment it was said. They aren't independently verified.
1990s
Nvidia's original graphics approach used quadrilaterals as polygons instead of the triangles other companies adopted.
“They used quadrilaterals as polygons instead of triangles as polygons.”
1990s
Nvidia CEO Jensen Huang laid off 70% of the company during its early struggle for survival.
“Jensen laid off 70% of the company and they did two completely crazy things.”
1990s
Nvidia designed its chips in software emulation without making prototype chips first, unlike its competitors.
“they designed all of their chips in software emulation as opposed to what everybody else did”
1993
Nvidia's first funding round was about $2 million from Sequoia and Sutter Hill at a $6 million post-money valuation.
“I assume a million each, but two million dollar total round at a six million dollar post money valuation.”
1993
Nvidia was named after the Latin word invidia, meaning envy, after the founders had been labelling chip design files NV for 'next version'.
“And they find the Latin word in video, I N V I D I A, which means envy.”
1993
Nicolai Tangen states that Jensen Huang started Nvidia in 1993 at age 30.
“You started Nvidia in 93, you were 30 years old.”
1993
Huang says he did not know how to be a CEO and nobody at Nvidia knew how to build a company when it began.
“I didn't know how to be a CEO, and n no, nobody in the company knew how to build a company.”
1993
Huang says NVIDIA was founded in 1993 as the only consumer 3D graphics company at that time.
“at the time that we started the company in 1993, we were the only consumer 3D graphics company ever created.”
1995
Huang says NVIDIA's original 1993 architecture was called Direct NVIDIA, and DirectX arrived in 1995.
“In 1993, we had Direct NVIDIA. Yeah. Huh. And which in 1995 became, you know, well, DirectX came out.”
1996
In 1996 Sega backed away from Nvidia's quadrilateral-based graphics approach, which conflicted with Microsoft's triangle-based Direct3D.
“And then in 1996, Sega's like, yeah, we're not so sure about this quadrilaterals thing either.”
1997
Nvidia's Riva 128 sold 1 million units within four months of launch in 1997.
“So it works. They sell 1 million units of the Riva 128 within four months.”
1997
After the Riva 128, Nvidia shipped each hardware generation in about six months while the rest of the industry took 18 to 24 months.
“And they now have a process within the company to design and ship each next generation of their hardware in a six-month timeline, while the rest of the industry is on an 18 to 24-month timeline.”
1997
In 1997 NVIDIA had about six months of cash left and tested the Riva 128 entirely in simulation rather than on a physical prototype.
“And you guys have about six months of cash left. And so you decide to do the entire testing in simulation rather than ever receiving a physical prototype.”
1997
Huang says that by the time NVIDIA recognized its architecture was wrong, Microsoft had released DirectX, which was incompatible with it, and about 30 competitors had appeared.
“Microsoft had already rolled out DirectX. It was fundamentally incompatible with NVIDIA's architecture. 30 competitors have already shown up”
1997
Huang says NVIDIA bought emulation hardware from a company called Ikos that was shutting down because it had no customers.
“There was a company called Icos. And on the day that I called them, they were just shutting the company down because they had no customers.”
1997
Huang says the Riva 128 was NVIDIA's third chip, after NV1 and NV2.
“Remember, Riva 120 was MV3.”
1997
Morris Chang says his relationship with NVIDIA's Jensen Huang began with a letter Huang sent him around 1997 asking TSMC to take NVIDIA as a foundry customer.
“It started, my relationship with Jensen, started with a letter that he sent to me. I think it was 1997.”
1997
Morris Chang recalls NVIDIA in 1997 as facing bankruptcy with perhaps 50 or 60 employees.
“They were facing bankruptcy, I think. And they had maybe 50 or 60 employees.”
1997
Morris Chang says that for NVIDIA to be a major TSMC customer it would need to generate at least $50 million in revenue a year.
“and to be a major customer of ours, he would have to produce revenue for us of at least 50 million a year”
1999
Nvidia released the GeForce 256 in 1999, marketed as the first GPU and described by the hosts as five times faster than anything else on the market.
“the first GeForce in 1999, and it's the GeForce 256. It's so powerful. It has 5x better graphics performance than like anything else on the market.”
1999
Nvidia went public in 1999 at roughly a $600 million market capitalization.
“They go public in beginning in 1999 at a $600 million market cap.”
1999
Intel released its own standalone graphics cards around 1999 to compete directly with Nvidia, and the hosts say they were poorly reviewed.
“They come out with their own external cards right around this time, like 1999 to directly compete.”
1999
Nvidia shipped the first GeForce card, the GeForce 256, in fall 1999, followed by the GeForce 2 in spring 2000 and the GeForce 2 Ultra in fall 2000.
“So in the fall of 1999, they ship the first GeForce card, the GeForce 256. In the spring of 2000, GeForce 2. In the fall of 2000, GeForce 2 Ultra.”
1999
Morris Chang says NVIDIA became one of TSMC's five biggest customers within two or three years of the 1997 meeting.
“Within two or three years, they did become one of the biggest five customers of TSMC.”
2001
Nvidia reached about $1.4 billion in revenue in calendar 2001, becoming the fastest semiconductor company to reach a billion in revenue and joining the S&P 500, according to the hosts.
“Which makes them the fastest semiconductor ever to reach a billion in revenue and gets them added to the S&P 500.”
2002
The first reference to using Nvidia GPUs for non-graphics work was a 2002 research paper on simulating cloud movement.
“the first reference to using NVIDIA GPUs for non graphical purposes was in a 2002 research paper on using computers to simulate the movement of clouds.”
2004
Nvidia's gross margin in 2004 was only 29%, compared with about 66% today according to the hosts.
“Back in 2004, that gross margin was only 29% that they were able to command as a premium on their cards.”
2006
The hosts say AMD tried to buy Nvidia before buying ATI, but the deal failed.
“But AMD's first choice was actually NVIDIA. So AMD tried to buy NVIDIA to make that their graphics line”
2006
AMD acquired ATI, described as Nvidia's only lasting serious graphics competitor, with the deal announced in 2006 and closing in 2007.
“AMD acquires ATI. And ATI was a very legit competitor. It was the only standing legit competitor to NVIDIA through its whole life.”
2006
Huang says NVIDIA's move toward the data center began almost 17 years before the interview, starting with streaming from GPUs for cloud gaming.
“Our journey to the data center happened, I would say, almost 17 years ago.”
2007
Nvidia's market capitalization reached just under $20 billion by mid-2007.
“By mid 2007, the stock reaches just under $20 billion market cap.”
2007-2008
Nvidia's stock price fell by more than 80 percent between October 2007 and November 2008 during the global financial crisis.
“and NVIDIA's stock price fell by more than 80% from October 2007 to November 2008.”
2008
The first product shipped using an Nvidia Tegra chip was the Microsoft Zune HD media player.
“It was the Microsoft Zune HD media player.”
2009
Morris Chang offered NVIDIA more than $100 million to settle the 40-nanometer dispute, with the offer valid for 48 hours.
“More than $100 million, yeah. And I also said, our offer is effective 48 hours.”
2011
Nvidia bought the British mobile baseband company Icera in 2011, whose founders later started Graphcore.
“they buy a mobile baseband company called iSera, a British company called iSera in 2011.”
2013
In 2013, Huang overruled executives who thought deep learning was a passing fad and committed Nvidia to going all in on it.
“Deep learning is going to be really big, he said. He said this in 2013. We are going all in on it.”
2014
Nvidia's gross margin was about 30% in 1999, passed 50% in 2014, and reached 66% by 2022.
“back in 99, they had a gross margin of 30% on their graphics chips. And then in 2014, they broke the 50% mark.”
2014
Google's order of 40,000 NVIDIA GPUs, costing $130 million, was personally approved by Larry Page over finance's objections.
“That's a big enough price tag that the request gets elevated to Larry Page, who personally approves it, even though finance wanted to kill it.”
2016
Nvidia did not return to its 2007 market cap peak of about $20 billion until 2016.
“It's not until 2016 that NVIDIA gets back to the $20 billion market cap peak that they were in 2007 when they were just a gaming company.”
2017
Lee says he met Nvidia's Jensen Huang in 2017 while at Lone Pine.
“I remember meeting Jensen Huang in 2017. It's kind of crazy to think at the time, like today, that, like, it wasn't what it is today.”
2018
In 2018 Nvidia changed the terms for its GeForce consumer cards to prohibit their use in data centers.
“in 2018, they actually do change the terms of the user agreements of their consumer cards, of GeForce cards, that you cannot put them in data centers anymore.”
2019
NVIDIA announced in March 2019 that it would acquire the networking company Mellanox for $7 billion in cash.
“In March of 2019, NVIDIA announced they were acquiring Mellanox for seven billion dollars in cash.”
2019
NVIDIA released the Megatron language model in August 2019, with 8.3 billion parameters trained on 512 GPUs over nine days.
“in August of 2019, NVIDIA released what was at the time the largest transformer-based language model called Megatron. 8.3 billion parameters trained on 512 GPUs for nine days”
2020
Nvidia acquired the Israeli data center networking company Mellanox in 2020 for about $7 billion.
“in 2020, they acquire an Israeli data center compute company called Mellanox that I believe focuses on like a networking compute within the data center. Yep. For about $7 billion.”
2022
Nvidia's data center revenue was about $3 billion two years before the recording, roughly half of gaming, and then tripled to over $10 billion a year.
“So two years ago, they had about three billion of revenue and it was only about half of their gaming revenue segment.”
2022
Nvidia announced an Arm-based data center CPU named Grace to pair with its Hopper GPU architecture.
“indeed, this year, NVIDIA announced they are making a data center CPU, an ARM-based data center CPU called Grace to go with the new Hopper architecture for their latest GPU.”
2022
By GTC 2022, Nvidia reported 3 million registered CUDA developers and 450 SDKs and models, with 60 new ones announced that year.
“There are now 3 million registered CUDA developers, 450 separate SDKs and models.”
2022
Nvidia generated about $27 billion in revenue in its most recent fiscal year before this episode.
“They did 27 billion dollars last year in revenue.”
2022
The hacking group Lapsus stole access to Nvidia's source code and demanded it remove crypto mining limits and open source its drivers.
“they demanded two things. They said, one, get rid of the crypto governors, like, make it so that we can mine”
2022
From September 2022 NVIDIA split its data center GPUs (Hopper) and consumer gaming GPUs (Lovelace) into separate architectures.
“Starting in September of 2022, they broke out the two business lines into different architectures.”
2022
NVIDIA launched the H100 data center GPU in September 2022 at about $40,000 per chip.
“So they launched it in September 2022. It's the successor to the A100. One GPU, one H100 costs $40,000.”
2022
Sales to mainland China made up about 25 percent of NVIDIA's revenue in the prior year.
“Last year, China was 25%, or sales to mainland China, was 25% of NVIDIA's revenue.”
2022
After US export controls announced in September 2022, NVIDIA could not sell H100 or A100 chips in China and created reduced-speed A800 and H800 versions.
“NVIDIA can no longer sell their top-of-the-line H100s, or A100s, to anybody in China. So, they created a nerfed SKU, essentially, that meets the regulations, the performance regulations, the A800 and H800s.”
2022
Amit Joshi says OpenAI built its early generative AI on existing infrastructure such as Microsoft's Azure platform and NVIDIA chips, which were not designed specifically for generative AI.
“they used the Azure platform, they used the NVIDIA chips, obviously, which were not, let's face it, they were not designed specifically for generative AI.”
2022
After Acquired's NVIDIA episodes, the company reportedly told the hosts that Jensen Huang had listened and asked who their inside sources were.
“Jensen has listened and wants to know who your inside sources are because it's the most correct telling of NVIDIA's story ever.”
2023
Kevin Scott said Microsoft was deploying NVIDIA's Hopper GPUs at scale, including features such as eight-bit floating point arithmetic.
“we're deploying, uh, hopper now at scale, and you know, like a bunch of the features of hopper, like, you know, eight bit floating point”
2023
The host says CoWoS advanced packaging accounts for roughly 10 to 15 percent of TSMC's capacity, which limits how many H100s NVIDIA can get.
“So CoWoS represents right now about 10 to 15% of TSMC's capacities, and many of the facilities are custom built for exactly these types of chips that they're producing.”
2023
NVIDIA's DGX H100 system, which bundles eight H100 GPUs, starts at $500,000 per box.
“So if you want to buy a new top-of-the-line DGX H100 system, the price starts at $500,000 for one box.”
2023
NVIDIA's DGX Cloud service starts at $37,000 per month for an A100-based system, hosted in partner clouds such as Azure, Oracle and Google.
“So starting price for DGX Cloud is $37,000 a month, which will get you an A100-based system, not an H100-based system.”
2023
NVIDIA forecast second-quarter revenue of $11 billion in May 2023, citing unprecedented demand for generative AI compute.
“due to unprecedented demand for generative AI compute in data centers, NVIDIA forecasts Q2 revenue of $11 billion, which would be up another 53% quarter over quarter over Q1 and 65% year over year.”
2023
NVIDIA reported quarterly revenue of $13.5 billion for its second quarter of fiscal 2024, up 88 percent from the prior quarter.
“total company revenue of $13.5 billion, up 88% from the previous quarter and over 100% from a year ago.”
2023
NVIDIA's data center segment generated $10.3 billion of its $13.5 billion quarterly revenue in the quarter reported in August 2023.
“revenue of $10.3 billion. So $10.3 out of $13.5 for a segment that basically didn't exist five years ago for the company.”
2023
CUDA reached 4 million registered developers in May 2023, up from 3 million in 2022.
“And then just one year later, in May of 2023, CUDA has 4 million registered developers.”
2023
NVIDIA's gross margin was 70 percent in the quarter discussed, with a forecast of 72 percent for the next, versus about 24 percent in its commodity graphics card era.
“This last quarter, they had a gross margin of 70%. And they forecasted for next quarter to have a gross margin of 72%.”
2023
NVIDIA had about 26,000 employees in 2023, compared with about 220,000 at Microsoft.
“They have 26,000 employees, and that sounds like a big number, but for comparison, Microsoft, whose market cap is only twice as big, has 220,000.”
2023
Jensen Huang, as relayed by the host, framed NVIDIA's opportunity as $1 trillion of installed data center assets growing by $250 billion a year.
“There is $1 trillion worth of hard assets sitting in data centers around the world right now. Growing at $250 billion a year.”
2023
The hosts of Acquired say they spent about 500 hours researching NVIDIA over two years before interviewing Jensen Huang.
“After researching NVIDIA for something like 500 hours over the last two years, we flew down to NVIDIA headquarters to sit down with Jensen himself.”
2023
At the time of recording, NVIDIA had a market value of about $1.1 trillion, the sixth largest in the world.
“At the time of recording, NVIDIA is worth $1.1 trillion and is the sixth most valuable company in the entire world.”
2023
Huang says NVIDIA has about 3,200 employees on its Israel team from the Mellanox acquisition and calls the deal one of his best strategic decisions.
“We have some 3,200 people there now, and it was one of the best strategic decisions I'd ever made.”
2023
Huang describes the Hopper system as weighing 70 pounds with 35,000 parts.
“when you see a hopper, it's 70 pounds, 35,000 parts, 10,000 amps.”
2023
Huang estimates 250 to 300 million active CUDA GPUs are in use.
“There are literally a couple of hundred million, right? 250 million, 300 million installed base of active CUDA GPUs being used in the world today.”
2023
Huang says NVIDIA partnered with Getty Images and Shutterstock rather than scraping the internet for image-generation training data.
“we partnered with Getty and Shutterstock to create commercially a fair way of applying artificial intelligence”
2023
NVIDIA's two founding venture investors, Sutter Hill's Tench Coxe and Sequoia's Mark Stevens, are still on its board.
“The two founding VCs are still on the board. Sutter Hill and Sequoia? Yeah, Tinch Cox and Mark Stevens.”
2023
Alison Beard says Nvidia earns about 26 billion in annual revenue and employs about 26,000 people worldwide.
“It now earns 26 billion in annual revenue and employs 26, zero people around the world.”
2023
Jensen Huang says Nvidia uses AI to design its chips, including finding optimal placements and connecting transistors into circuits.
“we're using AI to design our chips, to find optimal placements to, uh, help us, um, connect transistors”
2023
Jensen Huang says Nvidia is starting to use AI to trace a failure back to the line of code that likely caused a bug.
“we're starting to use AI to help us determine from a failure where the bug likely, uh, started from”
2023
Jensen Huang says US restrictions on chip sales to China limit the performance level of Nvidia's GPUs rather than banning sales outright.
“The limitations are related to the performance levels of our GPUs. This is like restrictions of of, uh, CPUs and and computers”
2023
Jensen Huang says Nvidia faces many local competitors in China while it complies with the export restrictions.
“the challenge there is that the local markets, especially in China, has a lot of competitors, and so we have to compete in the marketplace”
2023
Acquired first asked NVIDIA about interviewing Jensen Huang almost two years before the 2023 interview, before it began its NVIDIA research.
“I think we originally reached out to NVIDIA before we started our NVIDIA part one research almost two years ago”
2024
Nicolai Tangen noted that Nvidia's roughly $3 trillion market capitalization was nearly twice the size of Norway's sovereign wealth fund, the world's largest.
“Which is nearly twice the size of, uh, this fund, which is the largest sovereign wealth fund in the world.”
2024
Garman says most AI-related value currently goes to NVIDIA, but that this reflects how early the cycle is and there is enough for everybody.
“There's probably enough for everybody. Uh, today, most of it does go to NVIDIA, I think that's that's a lot of it.”
2024
Nilay Patel said Amazon's AI push includes more than $8 billion in funding for Anthropic, its own AI chips to compete with Nvidia, and nuclear power investments.
“That includes more than 8 billion in funding for Anthropic, a huge push to build its own AI chips to compete with Nvidia, and even nuclear power investments”
2024
Matt Garman said the newest generation of AI hardware, including AWS's Trainium2 and Nvidia's Blackwell GB200s, should give customers a two to four times boost in compute per dollar.
“we just announced with Trainium two, and eventually, uh, with, uh, Blackwell and and GB200s, I think will give customers a two to four x boost in compute capability per dollar.”
2025
The New York Times reported that Nvidia had reached a valuation of about $5 trillion.
“the New York Times, which, uh, which is reporting that Nvidia is now worth $5 1 trillion as it.”
2025
The New York Times article described Nvidia as a linchpin in the Trump administration's trade negotiations with Asia.
“The AI chipmaker has become a linchpin in the Trump administration's trade negotiations with Asia.”
2025
Carvalho says DeepSeek reportedly trained its models without advanced NVIDIA GPUs, whose export was restricted under the Chips Act.
“supposedly without the advanced NVIDIA GPUs whose export was restricted under the Chips Act.”
2026
Bussmann says JP Morgan announced it is building its own AI infrastructure with a chip provider other than Nvidia, which he sees as an exception among banks.
“It's only one bank, JP Morgan announced that they are also building their own infrastructure with, uh, chip provider, then not the typical Nvidia, et cetera.”
2026
Daniel Whitenack says Nvidia and others had recently made public appeals about the importance of open-weight and open-source models.
“Nvidia and others making pleas to the community around the importance of open weight models or open source models.”
2026
The Washington Post reported that Nvidia gave $25 million toward the White House ballroom, three months after Trump allowed it to sell advanced chips to China.
“Biggest donation, $25 million from Nvidia, made three months after Trump decided to let the AI giant sell advanced chips to China over the objections of security experts.”
Undated
John Bathgate says Nvidia has close to 70 percent gross margins and operating margins in the low 40s, which he attributes to TSMC absorbing the manufacturing cyclicality.
“nvidia's got close to 70 gross margins and like you know low 40s operating margins”
Undated
Early on, Nvidia landed a deal with Sega to supply 3D graphics for its arcade machines and next-generation home console.
“Like they land a big deal with Sega to power their arcade consoles and their next generation home console to be the 3D graphics engine and would ultimately become the Sega Saturn.”
Undated
Nvidia's first chips were designed around scarce memory costing about $200 per chip, while rivals later used more memory costing about $50.
“NVIDIA's chips were designed to be super, super tight on memory, and the memory cost about $200 in component parts to go into their chips.”
Undated
Nvidia laid off 70% of its staff, leaving about 35 people, to extend its runway while it rebuilt around Microsoft's standard.
“They have to lay off 70% of the company, which they do. They go down to about 35 people.”
Undated
Nvidia spent about a million dollars, roughly a third of its cash according to the hosts, on a chip emulator so it could skip physical prototypes.
“They had to spend a million dollars just to get the emulator, you know, hardware and software to do this.”
Undated
Nvidia and TSMC signed a multi-year deal making TSMC Nvidia's primary chip foundry, a relationship that continues today.
“The next year, the two companies sign a huge multi-year deal for TSMC to become the primary foundry for NVIDIA and still are today.”
Undated
Microsoft signed Nvidia to supply the original Xbox's graphics in a deal worth $500 million a year, with a $200 million advance.
“And they do a huge, huge deal, $500 million a year deal for NVIDIA to supply the graphics for the Xbox with a $200 million advance.”
Undated
The GeForce 3 introduced programmable shaders and lighting to Nvidia's GPUs, letting developers write custom graphics effects.
“The GeForce 3, which introduces for the first time programmable shaders and lighting on the GPU.”
Undated
Nvidia created Cg, an extension of the C programming language for programming graphics, lighting and shaders on its GPUs.
“They call it CG. Literally, like they extend the C programming language with graphics libraries and capabilities to directly program graphics and lighting and shaders for the GPU.”
Undated
Nvidia has powered only three game consoles: the original Xbox, the PlayStation 3 and the Nintendo Switch.
“So you'll note there are three game consoles in the history of game consoles that NVIDIA has powered. The original Xbox, the PlayStation 3, which we'll talk about next time, and the Nintendo Switch.”
Undated
Nvidia was an early investor in Keyhole, which Google acquired and which became Google Earth.
“Google acquired it. It became Google Earth, and NVIDIA was one of the early investors.”
Undated
A speaker says NVIDIA's early backers included Sequoia and Sutter Hill.
“They had such a great team and so much funding from Sequoia and Sutter Hill.”
Undated
Hamilton Helmer says that NVIDIA, the first computer graphics chip company in Silicon Valley, found within months that about 90 other funded companies were doing the same thing.
“they wake up 6, 12 months later and there's 90 other companies that have all been funded doing the same thing.”
Undated
Unlike most graphics and peripheral companies of the time, Nvidia wrote its own drivers instead of leaving them to downstream partners.
“But all the other graphics cards companies at the time and most peripheral companies, they let the further downstream partners write the drivers for what they were doing.”
Undated
According to a LinkedIn search by the host, about 1,100 Nvidia employees have CUDA in their job titles.
“There are 1100 employees dedicated specifically to the CUDA platform.”
Undated
In a Stratechery interview, Jensen Huang said Nvidia offered over 150 SDKs serving industries from gaming to robotics.
“We have over 150 SDKs that serve industries from gaming and design to life and earth sciences, quantum computing, AI, cybersecurity, 5G, and robotics.”
Undated
Nvidia's Tegra platform powers the Nintendo Switch and earlier powered the original Tesla Model S touchscreen.
“The Tegra platform still to this day is the main processor of the Nintendo Switch.”
Undated
Nvidia developed DLSS, Deep Learning Super Sampling, which renders games at lower resolution and uses deep learning to upscale for higher frame rates.
“Where they basically have faced this problem of, well, we either could render stuff at low resolution with less frames.”
Undated
Nvidia deliberately limited the crypto mining performance of some gaming cards and introduced dedicated mining cards.
“they artificially crippled the card to make it worse at crypto mining.”
Undated
One of NVIDIA's company-risking bets, per the host, was a roughly seven-year investment in AI and CUDA and its own developer ecosystem before the market existed.
“And then a third one around the like seven year investment before the market was there in AI and CUDA and building their own developer ecosystem.”
Undated
A host says NVIDIA designed an early graphics card entirely in emulation, shipped it without testing it, and it worked.
“they designed the whole thing in emulation never tested it and shipped it to customers and it worked.”
Undated
Jensen Huang says he delivered the first Nvidia DGX deep learning system outside Nvidia to OpenAI.
“I delivered the very 1st one outside of our own company to OpenAI.”
Undated
Huang says AlexNet, designed for computer vision, was the first widely seen deep learning network and required Nvidia GPUs to produce.
“the 1st deep learning network that we, uh, that the industry, a lot of people saw, was AlexNet, and AlexNet was, uh, designed for computer vision.”
Undated
Nvidia acquired Mellanox to change how computers are connected and work together for AI workloads.
“we bought a company called Mellanox so that we can change the way computers are connected and the way can computers can work together.”
Undated
Huang says AI-based weather simulation at Nvidia runs thousands of times faster than numerical methods, citing 10,000 times elsewhere in the discussion.
“we're already simulating weather now. Ten zero, 50,000 times faster than you using numerics”
Undated
Huang says Nvidia expects to improve its engineers' productivity by a factor of ten using AI.
“we're gonna improve the productivity of our engineers by a factor of ten.”
Undated
Huang estimates the R&D budget for one of Nvidia's large chip generations at about $5 billion, more than building a rocket.
“The R D budget for it is probably something like $5 billion”
Undated
Huang says AI has learned the language of proteins, allowing synthesis of proteins with desired functions for uses such as drugs and breaking down plastics.
“Just as we've learned the language of humans, we've now learned the language of proteins”
Undated
Huang says Nvidia faced existential crises maybe five, six or seven times in its first 15 years.
“The 1st 15 years, the adversity were incredible. Um, maybe five, six, seven times, it was existential.”
Undated
Kevin Scott said NVIDIA has been Microsoft's compute and network partner since NVIDIA acquired Mellanox.
“NVIDIA's been, you know, our compute and network partner since they bought Mellanox”
Undated
Huang says NVIDIA's first two chips, NV1 and NV2, used forward texture mapping with curved surfaces rather than triangles, an approach that proved wrong.
“MV1 and MV2 were based on forward texture mapping, no triangles but curves, and it tessellated the curves.”
Undated
Huang says GeForce Now was NVIDIA's first data center product, followed by remote graphics for enterprise data centers, then CUDA-based supercomputing products.
“GeForce Now was NVIDIA's first data center product. And our second data center product was remote graphics”
Undated
Huang says every NVIDIA accelerator is architecturally compatible with the others, which he says no other accelerator maker has done.
“We are the only accelerator on the planet where every single accelerator is architecturally compatible with the others.”
Undated
CUDA stands for Compute Unified Device Architecture, and Huang says it grew out of an earlier NVIDIA architecture called UDA.
“CUDA is Compute Unified Device Architecture.”
Undated
Huang credits John Carmack's Quake and Tim Sweeney's Unreal as the first two killer applications for consumer 3D graphics.
“Carmack and Sweeney both between Unreal and Quake created the first two killer applications for consumer 3D.”
Undated
One of the hosts says NVIDIA's stock went through two or three drops of more than 80% in the public markets.
“you went through two, two if not three, 80% plus drawdowns in the public markets.”
Undated
Huang says NVIDIA's market value fell to about two to three billion dollars for a period because of its investment in CUDA.
“we traded down at about a couple of two, three billion dollars in market value for a while because of the decision we made in going into CUDA”
Undated
Gandhi said GPUs are widely available and many engineers are trained on them, which is a consideration when choosing between TPUs and GPUs.
“Because there is this that was, you know, externally available, and a lot of people are trained up.”
Undated
Gandhi said the NVIDIA GPU shortage comes up in customer conversations with Google Cloud.
“It definitely is mentioned, uh, as we think about it in the conversations, uh, it's always the the discussion that comes up.”
Undated
Jensen Huang, an electrical engineer, launched Nvidia with two co-founders at a Denny's restaurant.
“An electrical engineer by training, he launched the company with two co founders at a Denny's restaurant.”
Undated
Jensen Huang says accelerated computing improves performance by roughly 1,000 to 100,000 times every ten years.
“in the case of accelerated computing, we're moving somewhere between a thousand to 100,000 times, uh, every ten years.”
Undated
Jensen Huang says Nvidia's AI GPUs weigh about 70 pounds and contain about 35,000 parts.
“GPUs that we build for AI, uh, weighs 70 pounds. Uh, they have 35, zero parts.”
Undated
Jensen Huang says eight of the roughly 35,000 parts in Nvidia's AI GPUs come from TSMC.
“Uh, out of those 35, zero parts, eight of them come from TSMC.”
Undated
Jensen Huang says Nvidia's AI GPUs consume 10,000 amps and are so heavy that robots are used to manufacture them.
“They, uh, consume 10,000 amps, uh, obviously, they're so heavy.”
Undated
Jensen Huang says Nvidia's computers are so advanced that a supercomputer is needed to test another supercomputer.
“The computers are so advanced that it takes a supercomputer to test another supercomputer”
Undated
Jensen Huang says Nvidia has nearly gone out of business a handful of times in its history.
“actual experiences of nearly going out of business a handful of times.”
Undated
Jensen Huang says Nvidia cut roughly three or four layers of management by flattening its organization.
“we probably reduced, you know, three, three or four layers of management, uh, out of our company just by doing so.”
Undated
Jensen Huang expects quantum computers to work as accelerators connected to Nvidia's accelerated computing systems.
“when it gets here, uh, it'll likely be an accelerator that's connected to the Nvidia accelerated computing systems.”
Undated
Nadella said Azure will use Nvidia and AMD chips as well as Microsoft's own silicon and system architecture.
“we'll partner with Nvidia, we'll partner with A M D, we'll have our own silicon, uh, we will have our own system architecture.”
Undated
Kanjun Qiu says Imbue's investors include Nvidia, Eric Schmidt and a nonprofit institute, rather than venture investors.
“our investors are Nvidia, Eric Schmidt, uh, nonprofit, Haotester Institute.”
Undated
Garman says NVIDIA runs its own AI training clusters on AWS because AWS has the most stable infrastructure.
“NVIDIA actually runs their AI training clusters in AWS because we actually have the the most stable infrastructure of anyone else.”
Undated
Jensen Huang says he answered no when asked whether he would start NVIDIA again knowing what he knows, and that the answer was taken out of context.
“I said, absolutely not. Of course, it was taken out of context”
Undated
Rana El Kaliouby cites NVIDIA CEO Jensen Huang's view that a person who harnesses AI, not AI itself, will take your job.
“It's not going to be an AI that takes your job, it's going to be a person who knows how to leverage and harness AI that will.”
Undated
Nvidia's first product, the NV1, was a commercial flop.
“Their first product, right, is a flop. And it's called the NV one.”
Undated
After its early product failures, Nvidia cut its staff from over one hundred employees to forty.
“He goes down from over one hundred employees to down to forty.”
Undated
Nvidia bought a $1 million Ecos emulator to test chip designs faster, which cut its cash runway from nine months to six.
“an Ecos machine cost $1 million and would cut NVIDIA's runway from nine months down to six months.”
Undated
At the time of the episode, Nvidia had about 29,000 employees.
“NVIDIA is Jensen with 29,000 lives. Right now, NVIDIA has 29,000 employees.”
Undated
Jensen Huang has about sixty direct reports and refuses to hold one-on-one meetings with them.
“he has sixty direct reports, and he refuses to do one on one.”
Undated
Under Nvidia's three-teams, two-seasons structure, the company could release a new chip every six months while competitors took about eighteen.
“Because this allows the company to release a new chip every six months, their competitors are releasing a chip every eighteen months.”
Undated
Huang reads over a hundred "top five things" emails from employees each day to see what is happening inside Nvidia.
“So every day, he reads, you know, over a hundred top five emails to get a snapshot at what's happening inside of the company.”
Undated
Huang reviews a list of top contributor candidates and approves special one-off stock grants that vest over four years.
“Okay, Jensen will review the list of top contributor candidates and give out special one-off grants of stock that vest over a four-year period.”
Undated
Nvidia's "ship the whole cow" strategy repackaged chips that failed top-tier quality tests into cheaper, slower products instead of discarding them.
“This strategy was dubbed ship the whole cow, a reference to how butchers find ways to use almost every part of the carcass, not just the prime cuts.”
Undated
A researcher named Harris, whom Nvidia hired, coined the term GPGPU, meaning general-purpose computing on GPUs.
“Harris is the guy that coins the phrase GPGPU, which means general purpose computing on GPUs”
Undated
Nvidia created CUDA, short for compute unified device architecture, a programming model that made GPUs easier to use for general computing.
“this new programming model for chips that they call the compute unified device architecture. It's called CUDA.”
Undated
Making its GPUs CUDA-compatible took four years and cut Nvidia's gross margin from forty-five percent to thirty-five percent.
“The move was extremely expensive. It took four years and the company invested so much in converting its GPUs for CUDA compatibility that its gross margin fell from forty five to thirty five.”
Undated
Nvidia chief scientist David Kirk gave over a hundred talks in a year to promote teaching parallel computing and CUDA.
“David Kirk gave over a hundred talks over the course of a year on this.”
Undated
Aaron Levie attributes to NVIDIA's Jensen Huang the idea that the IT department becomes the HR department of AI.
“Jensen at NVIDIA kind of put it the the best, which is, you know, effectively, the IT department becomes the HR department of AI.”
Undated
Nikesh Arora cited plans to spend $500 billion on large AI data centers as evidence of the scale of AI investment.
“You can see Nvidia, you can see infrastructure, people wanna spend $500 billion building large data clusters, data centers.”
Undated
Elad Gil says Nvidia chief Jensen Huang has told the show he has about 40 direct reports and does not hold one-on-one meetings.
“we've talked to Jensen a few times and he's pointed out, for example, he has like 40 direct reports, he doesn't do one on ones”
Undated
Chris Benson credited Nvidia's GPUs with supporting the continuing evolution of neural network technology.
“NVIDIA's come about able to make GPUs that could support the continuing evolution of the technology.”
Undated
DJ Patil said Andrew Ng was one of the first people to advocate using GPUs for deep learning.
“he was one of the first people to advocate and using GPUs for deep learning.”
Undated
Andrew Ng said China's recent ban on certain Nvidia chip imports signals movement toward independence from TSMC in Taiwan.
“it's been interesting to see China recently banned certain imports of Nvidia chips, which is a strong signal that China is moving toward independence from TSMC in Taiwan”
Undated
Rene Haas is an electrical engineer who worked at Nvidia and Mythic before joining Arm.
“An electrical engineer by training, he also worked at Nvidia and Mythic before joining Arm.”
Undated
Rene Haas says Nvidia chose Arm CPUs to pair with the accelerators in its Grace Hopper and Grace Blackwell chips.
“when NVIDIA made the choice with their Grace Hopper chip and then Grace Blackwell, they needed a CPU to pair up with the accelerator, and and they chose ARM”
Undated
Rene Haas says TSMC is by far the main provider of AI compute chips, including everything Nvidia builds.
“Uh, TSMC is by far and away the provider of all these AI compute chips. Everything NVIDIA builds is from there.”
Undated
Agrawal said AI companies would need to earn about $1 trillion in revenue to justify the current capital spending on AI infrastructure.
“you've gotta earn about $1 1 trillion of AI revenue for all of this capex to be worth it”
Undated
Anima Anandkumar was previously principal scientist at Amazon Web Services and senior director of AI research at NVIDIA.
“Previously, she was principal scientist at Amazon Web Services and senior director of AI research at NVIDIA.”
Undated
Anima Anandkumar said her group built the first high-resolution AI-based weather model in collaboration with Nvidia, Caltech and Berkeley Lab.
“We built the first high-resolution AI-based weather model, but that was in deep collaboration with my team at Nvidia, Caltech, uh, Berkeley Lab.”
Undated
Anima Anandkumar said the AI weather model is tens of thousands of times faster than physics-based forecasting and can run on a desktop GPU.
“it's tens of thousands of times faster. So what would take a supercomputer can be put in now a desktop GPU.”
Undated
Anima Anandkumar said Nvidia open sourced its AI weather model under a permissive Apache license.
“But we immediately open sourced it, right? So it's faster, it's accurate, you know, and because of that open sourcing in a permissive way with Apache license.”
Undated
Boeckem says Nvidia is a long-standing partner of Hexagon in physical AI.
“as Nvidia is a good partner for for us. We have been on the trajectory trajectory, uh, with them, um, for for a longer time now.”
Undated
Brian Elliott introduces Sid Pardeshi as a prolific inventor at NVIDIA who has long been interested in building large software systems.
“since Sid was a a prolific inventor at NVIDIA, he's been he's been thinking about building large scale software systems since he was a little little boy”
Undated
Nvidia was already selling millions of GeForce GPUs a year when it decided to put CUDA on them, according to Jensen Huang.
“We were already selling millions and millions of GeForce GPUs a year.”
Undated
Jensen Huang says Nvidia's market value fell to about $1.5 billion after CUDA's cost hit the company.
“our market cap went down to like $1.5 billion, and so we were down”
Undated
Jensen Huang says CUDA raised the cost of GeForce GPUs by about 50% when Nvidia had roughly 35% gross margins.
“increased our cost by 50 and that con consumed, and we were a 35 gross margin company”
Undated
Jensen Huang says NVLink 72 let Nvidia place a model of 4 to 10 trillion parameters in a single computing domain, driven by the arrival of mixture-of-experts models.
“We could now take an entire 4 trillion, 10 trillion parameter model and put it in one computing domain, as if it's running on one GPU.”
Undated
Jensen Huang describes Nvidia's safety rule that an agent should have at most two of three capabilities: access to sensitive information, code execution and external communication.
“Agentic systems can s can access sensitive information, it can execute code, and it can communicate externally.”
Undated
Jensen Huang says Moore's law would have improved computing about 100 times in ten years, versus a million-fold gain Nvidia achieved.
“And so in the last ten years, Moore's law would have progressed computing about a hundred times in the last ten years.”
Undated
Jensen Huang says he persuaded several memory-maker CEOs about three years earlier to invest in HBM, then used mainly in supercomputers, as mainstream data-center memory.
“About three years ago, I was able to convince several of the CEOs that even though at the time HBM memory was used quite scarcely, you know, and and barely by supercomputers.”
Undated
Jensen Huang says about 200 suppliers contribute to Nvidia's Vera Rubin rack, which has roughly 1.3 million components.
“we have a couple of hundred suppliers that contribute the technology that goes into our 1.3 million component rack”
Undated
Jensen Huang says power grids are built for rare worst-case days and typically run around 60% of peak.
“Most of the time, we're nowhere near the worst case condition, and we're probably running around, call it 60 of peak.”
Undated
Jensen Huang says he started a management method called the speed of light about 30 years ago, which tests every design against physical limits.
“a method, uh, that I started, uh, 30 years ago, and it's called the speed of light”
Undated
Lex Fridman cites Nvidia's Vera Rubin pod as containing 7 chip types, 5 rack types, 40 racks, 1.2 quadrillion transistors and nearly 20,000 Nvidia dies.
“we're talking about 7 chips, 7 chip types, 5 purpose-built rack types, 40 racks, 1.2 quadrillion transistors, nearly 20,000 Nvidia dies”
Undated
Jensen Huang says Nvidia and its partners can produce about 200 Vera Rubin pods per week.
“probably kind of crank out about 200 of these pods a week, just to put in perspective”
Undated
Lex Fridman describes Nvidia's Nemotron 3 Super as a 120-billion-parameter open-weight mixture-of-experts model.
“Nemotron 3 Super, which you can also use inside Perplexity to look stuff up now, which is, uh, 120 billion parameter open weight, uh, MoE model.”
Undated
Jensen Huang says Nvidia has done hundreds of billions of dollars of business with TSMC over three decades without a contract.
“Three decades, I don't know how many tens, hundreds of billions of dollars of business we've done through them, and we don't have a contract.”
Undated
Jensen Huang credits about 43,000 Nvidia employees and several million developers for CUDA's success.
“it wasn't three people that that made CUDA successful, it was 43,000 people that made CUDA successful”
Undated
Jensen Huang says Nvidia GPUs are already operating in space, doing imaging work on satellites.
“we're already there. Uh, NVIDIA GPUs are the 1st GPUs in space”
Undated
Jensen Huang says it is possible for Nvidia to reach $3 trillion in annual revenue in the near future.
“is it possible for NVIDIA to be a, you know, 3 trillion dollar revenues company in the near future? The answer is, of course, yes.”
Undated
Jensen Huang recalls a CEO telling him that a fabless semiconductor company could not exceed $1 billion in revenue, a prediction Nvidia later passed.
“it's theoretically impossible for a fabless semiconductor company to exceed $1 billion”
Undated
Jensen Huang names Doom as the most influential game, saying it turned the PC from an office tool into a gaming device.
“I would say Doom from a from a our the intersection of the cultural implication as well as the industry, turning a PC into a gaming device.”
Undated
Samaya AI recently announced new investment from Nvidia's venture arm NVentures and from Databricks Ventures.
“you also announced recently a new investment from N Ventures, that's Nvidia's, uh, venture arm, as well as Data Bricks Ventures.”
Undated
HPE co-engineered with Nvidia on a NIM to make its finance AI give consistent, deterministic answers.
“we actually co-engineered with Nvidia, uh, on a nim to help crack determinism, because you just can't get finance data wrong.”
Undated
Peter Koerte says Siemens and NVIDIA aim to cut a car air-drag fluid dynamics simulation from eight hours to minutes.
“instead of waiting eight hours for a complex computational fluid dynamic simulation, let's say of the air drag on a car, instead of eight hours, we can reduce that to minutes.”
Undated
Siemens announced its AI acceleration examples for design and manufacturing with NVIDIA at CES.
“these are kind of examples that we have announced also at CES earlier this year with NVIDIA.”
Undated
Deere's autonomous tractor runs embedded NVIDIA Orin GPUs, which Hindman says must be hardened for shock, vibration and temperature.
“we run embedded GPUs, NVIDIA, uh, uh, orin, uh, platform GPUs on that on that product.”
Undated
Sarah Guo says Onyx co-founder Gil previously built synthetic data and worked at Nvidia.
“Your co founder Gil came out of building synthetic data and working at Nvidia.”
Undated
Hinds says Nvidia CEO Jensen Huang does not hold one-on-one meetings with his direct reports, a practice she says works at Nvidia because of its strong mission clarity.
“The fact that Jensen doesn't have one on one meetings with his direct reports, that does not work in most organizations.”
Undated
Daniel Whitenack said Nvidia and Intel rose that day as buyers rushed to purchase hardware.
“What did gain was, uh, uh, the, you know, Nvidia, Intel chip chip providers rising as people again, you know, panic, buy some of these, uh, these hardware solutions.”
Undated
Arvind Krishna describes Nvidia's Jensen Huang as funding ecosystem players to keep them within Nvidia's vertically integrated stack.
“He's trying to fund everybody in the ecosystem to ensure that they don't stray from the vertically integrated stack that he's producing.”
Undated
CrowdStrike used Nvidia's Nemotron model as one of the main bases for its blue defensive model.
“NemoTron is one of the main models that we leveraged on on the blue side.”
Undated
Kurtz says CrowdStrike and Nvidia have worked on the SafeMind models together for about six months.
“We've been working on it for the last six months, and we kinda went through how this would work.”
Undated
Some teams run disaggregated inference, using Nvidia chips for the prefill step and Cerebras chips for the decode step.
“some interesting, you know, people doing these, like, hybrid, you know, disaggregated inference, where they run, you know, NVIDIA for the prefill and and Cerebras for the decode”
Episodes
- 15 Dec 202510 Years of Acquired (with Michael Lewis)2h 47m
- 6 Oct 2025Google Part III: The AI Company4h 4m
- 27 Jan 2025TSMC Founder Morris Chang2h 55m
- 30 Sep 2024Acquired LIVE from Chase Center (with Daniel Ek, Emily Chang, Jensen Huang and Mark Zuckerberg)2h 25m
- 18 Dec 2023Holiday Special 20232h 17m
- 16 Oct 2023NVIDIA CEO Jensen Huang1h 31m
- 6 Sep 2023Nvidia Part III: The Dawn of the AI Era (2022-2023)2h 55m
- 19 Dec 2022Holiday Special 20222h 21m
- 20 Jun 2022The Playbook: Lessons from 200+ Company Stories1h 5m
- 20 Apr 2022Nvidia Part II: The Machine Learning Company (2006-2022)2h 12m
- 6 Apr 2022Platforms and Power (with Hamilton Helmer and Chenyi Shi)1h 24m
- 28 Mar 2022Nvidia Part I: The GPU Company (1993-2006)2h 0m
- 3 Nov 2021Complexity Investing & Semiconductors (with NZS Capital)1h 44m
- 23 Mar 2026#494 – Jensen Huang: NVIDIA – The $4 Trillion Company & the AI Revolution2h 32m
- 13 Jan 2025#376 Jensen Huang: Founder of Nvidia1h 40m
- 16 Oct 2026Donald Trump: Ass Man1h 16m
- 22 Oct 2026Why AI Compute Comes in Rectangles — Erik Bernhardsson, Modal CEO1h 25m
- 21 Apr 2026Industrial AI for the Physical World: Siemens’s Peter Koerte27m
- 21 Jul 2026AI Agents in Banking: UBS Former Chief Information Officer54m
- 10 Apr 2026AI Agents in Finance with HPE's Chief Financial Officer (CFO) | CXOTalk 91455m
- 15 Jan 2026Robots and Physical AI: Strategy and Technology 2026 | CXOTalk #90551m
- 12 Aug 2025Are You Considering Chinese AI In Your Strategy?32m
- 14 Nov 2023Nvidia’s CEO On What It Takes To Run An A.I.-Led Company Now23m
- 13 Jan 2025Why CEO Matt Garman is willing to bet AWS on AI1h 10m
- 9 Apr 2025Palo Alto Networks CEO: Global Cyber Threats, Digital Safety, and Tech Leadership51m
- 26 Jun 2024Dario Amodei CEO of Anthropic: Claude, New models, AI safety and Economic impact1h 11m
- 13 Mar 2024Microsoft Chairman & CEO: AI, chip shortage, empathy, and poetry37m
- 12 Apr 2023Jensen Huang Founder and CEO of Nvidia47m
- 6 May 2026John Deere's AI vision for future farms39m
- 1 Apr 2026How AI is reshaping Wall Street37m
- 24 Dec 2025AI’s journey from the lab to the marketplace27m
- 17 Dec 2025‘Like Disneyland when it just opened’: Arm CEO Rene Haas on the AI revolution24m
- 26 Nov 2025Andrew Ng on winning the AI race, with DJ Patil26m
- 9 Oct 2024Debunking AI myths, with Jeremy Kahn39m
- 22 Oct 2026The New Apex Predator: CrowdStrike CEO George Kurtz on Adversaries Armed With AI Agents45m
- 24 Sep 2026Re-Founding Incumbents for the AI Era with Sequence Holdings Co-Founder and CEO Michael Lee42m
- 28 May 2026Building an AI Guardian for Enterprise with Onyx Security CEO Maxim Bar Kogan41m
- 4 Oct 2025The Impact of AI, from Business Models to Cybersecurity, with Palo Alto Networks CEO Nikesh Arora58m
- 29 Aug 2024Cloud Strategy in the AI Era with Matt Garman, CEO of AWS42m
- 23 Oct 2023What Google Cloud Can Teach Enterprises Developing & Rolling Out AI Tools, With Kawal Gandhi32m
- 24 May 2023Going Full Send on AI, and the (Positive) Impact of AI on Jobs, with Kevin Scott, CTO of Microsoft54m
- 21 Oct 2026Arvind Krishna & Nikesh Arora: Quantum Threat, Open vs Frontier AI, AI Cyber Warfare & SaaS Collapse40m
- 23 Dec 2025AI Enterprise - Databricks & Glean | BG2 Guest Interview45m
- 19 Mar 2026295. AI, strategy, and the future of work: What business leaders need to know35m
- 8 Aug 2024Reinventing AI Agents with Imbue CEO Kanjun Qiu48m
- 10 Jun 2026Babysitting the Machine: Glean's Rebecca Hinds on the Hidden Human Labor of AI at Work1h 45m
- 5 Feb 2026Infinite Code Context: AI Coding at Enterprise Scale w/ Blitzy CEO Brian Elliott & CTO Sid Pardeshi1h 53m
- 29 Jan 2025Aaron Levie, CEO of Box, on Box AI, Enterprise Enthusiasm, and the Evolution of SaaS50m
- 6 Aug 2026Models, Harnesses, and Multi-Agent Systems49m
- 23 Jul 2026Surviving the New Economics of a Post-Agentic World35m
- 10 Nov 2025Are we in an AI bubble?49m