NVIDIA Part I: The GPU Company (1993-2006)

In one sentence: Part one of Acquired's NVIDIA trilogy (parts two and three cover 2007-2022 and beyond) — the story of a graphics-card company founded in a Denny's, born into a bloodbath of 90 undifferentiated competitors, that bet the entire company three times and nearly went bankrupt three times before emerging as the only protagonist its industry ever produced. The episode's counterintuitive verdict: through 2006, NVIDIA had definitely found product/market fit, but had not yet found their source of power — and the party quietly capturing the value all along was Microsoft, which risked nothing. Jensen Huang's own epigraph runs through the whole episode: "My will to survive exceeds almost everybody else's will to kill me." Ben's closing line: "It's a literally unbelievable story, except that it happens, so you have to believe it."

The Company on One Page

Founder Profile: Jensen Huang

Origins: Born in Taiwan in 1963; the family moved to Thailand in his early childhood. His mother spoke no English and taught the two boys ten dictionary words a day — the source, the hosts note, of Jensen's distinctive accent. At 9 he was shipped to America; his parents' savings only stretched to the cheapest school they could find — Oneida Baptist Institute in eastern Kentucky, which turned out to be a reform school, not a prep school. The brothers were the school's first foreigners and the town's first Chinese; nine-year-old Jensen and his roommate — a 17-year-old just out of prison, recovering from seven stab wounds — became fast friends: Jensen taught him math, he taught Jensen to lift weights, a habit kept since age nine. The family eventually reunited, landing in Tacoma, Washington, then suburban Portland, Oregon; Jensen took third at the table tennis junior nationals (photo in Sports Illustrated) and bused tables at Denny's through high school. His own gloss on those years: "Now, I don't get scared very often. I don't worry about going places I haven't gone before. I can tolerate a lot of discomfort." David's comment: that one sentence runs through his entire life.

Career prehistory: Into Oregon State at 16 to study EE, where he met his lab partner Lori — later his wife (the couple went on to fund the Jensen and Lori Huang Engineering Center at Stanford and give millions to OBI). Joined AMD in 1984, working on a then-blazing 1 MHz CPU — his own joke: "You could see it coming from a long way, and still coming, and still coming." AMD produced his first big mental flip: in school, full-stack (design plus manufacturing) seemed cool; inside it, he decided the cool thing was specializing in one slice of the value chain and using tools, platforms, and partners so anyone could build a chip — so he jumped to ASIC pioneer LSI Logic, where he made his name on the Sun account as the guy who could take a customer's chip vision and "turn it into reality and production."

How he thinks: Employees call his habit "CEO math" — every number gets rounded; his defense: "The details only matter if you understand the big picture first." He dislikes the word "vision," finding it exclusionary, and says "our perspective" instead. Today he says "intellectual honesty is the cornerstone of NVIDIA's culture" — which, David insists, refers precisely to the 1996 scene in Playbook entry 3. He is also deeply folksy: he nearly died crashing a Fast-and-Furious-style Toyota Supra ("Just one more way, he is like Elon Musk." — Ben), and he still loves Denny's, where he always orders The Super Bird. Ben's verdict: even after five years of a soaring stock and adulation, Jensen remains an underrated CEO.

Signature quotes:

"My will to survive exceeds almost everybody else's will to kill me."

"When technology moves this fast, if you're not reinventing yourself, you're just slowly dying." — and dying at the speed of Moore's Law, "the fastest of any rate that we know."

The Playbook

Each entry: story → insight → effect.

1. The curse of the "brain dead yes": fundable is not winnable

• Story: In 1993, a graphics-card startup was an automatic yes for a VC — 3D demand was proven by Doom and Jurassic Park, and the peripheral-card wave (sound cards, network cards, serial cards) had validated the "dedicated accelerator card for the PC" business model. Within months, 89 similar companies were funded — 90 undifferentiated competitors in one market. David: "The problem when something is a brain dead yes for a venture capitalist is that it's a brain dead yes for lots of venture capitalists."
• Insight: A brain-dead yes for one VC is a brain-dead yes for all of them; an idea's fundability and its winnability are different things. Ben layers on Buffett's ~2000 Fortune piece: America once had ~70 car companies, which collapsed to Ford, GM, and Chrysler — in the proliferation phase nobody differentiates, nobody builds power, and the survivors compete on thin margins defended only by scale.
• Effect: Of the ~90 VC-funded 3D graphics startups, exactly two survived — NVIDIA, after its tour through hell, and ATI, a Canadian company that bootstrapped its way in entirely outside the Silicon Valley ecosystem.

2. The first mover's standards problem: designing for a constraint that no longer exists

• Story: As the first company in, NVIDIA had no standards to follow and picked its own lane — NV1 bet on quadrilateral primitives and a proprietary API/SDK, and its architecture was contorted to conserve then-expensive memory. A year and a half later Microsoft's Direct3D made triangles the de facto standard and the 89 followers adopted it wholesale; meanwhile Moore's Law cratered memory prices, leaving NVIDIA's memory-frugal design costing $200 in memory components while competitors just threw $50 of cheap memory at the problem. Ben: "This is a case study of what happens when you get more clever than the rest of the industry."
• Insight: First movers are forced to choose their own lane; followers wait for the standard and draft behind it. And a chip must be designed for the cost structure of its ship date, not its start date — you have to "premeditate the exponential curve," designing for hardware two, three, four generations out.
• Effect: In 1996 Sega — "not comfortable with quadrilaterals either" — walked away; NV2 died half-built and the company was, in the hosts' words, screwed. The lesson later returned in positive form: NVIDIA's headquarters today is built out of triangles — "in homage to game developers, not quadrilaterals."

3. Intellectual honesty over attachment to the vision

• Story: At the bottom, Jensen sat down with his co-founders: "Guys, this is a pipe dream. We need to throw it all out if we're going to survive." They had sold 100+ engineers on a vision of defining the industry standard; "compete on the same standard as everyone else, purely on performance" is not an exciting vision for a Silicon Valley engineer, and Curtis and Chris initially resisted. The price of admitting failure: laying off 70% of the company. David: "Jensen, God, he's such a G. He's like, no, we're not going out like this." — with every card on the table against them, "literally anybody else, you pull the plug."
• Insight: When the platform owner has sentenced your original strategy to death, the only way forward is to admit it died. The admission costs real money and a shattered narrative — but paying it is the only way to reach the next round.
• Effect: The full pivot to the Direct3D standard, competing on nothing but performance — which produced the RIVA 128 and everything after it. "Intellectual honesty" became the stated cornerstone of the culture.

4. Simulation is NVIDIA's DNA (Ben's "one big one")

• Story: The normal chip cycle was two years — design, foundry prototype, rounds of physical testing, then tape-out. The company had nine months of cash. Jensen cut the physical-prototype loop entirely: he spent $1M — roughly a third of the money in the bank — on an unproven startup's chip-emulation software and hardware; NVIDIA was that startup's only customer (it later went out of business). The emulator rendered one frame every 30 seconds; someone sat at the screen checking output frame by frame — mind-numbing work. When validation finished, six months of runway remained. No prototype; straight to the foundry: "Make 100,000 units." The foundry: "You guys sure about that?" Yup.
• Insight: A process innovation forced by desperation hardened into a worldview: whatever can be done in simulation should not be done in the physical world. The DNA runs straight to the present — planes no longer go part-by-part into wind tunnels, drug discovery is transformed (look how fast the COVID vaccines came), "Simulation is an absolute miracle," and everything in the world is being compressed to 10x, 100x speed because it can be simulated rather than physically built. It runs straight into NVIDIA's Omniverse.
• Effect: The RIVA 128 taped out right the first time in 1997 and saved the company; the emulation-driven pipeline hardened from a one-time gamble into a standing process within the company.

5. The six-month metronome: in a commodity market, shipping speed is the only differentiation

• Story: With interfaces and programming models standardized by Microsoft, everyone was building the same thing — a commodity — and everyone knew what the next generation should be. NVIDIA's emulation pipeline compressed design-to-ship to six months per generation while the whole industry ran on 18-24 months.
• Insight: David nails the essence — NVIDIA and its competitors were designing against the same generation of technology assumptions, but competitors took 18-24 months to ship what NVIDIA shipped in six: same-generation designs, on the market a year and a half early. Which meant doubling performance at the same price point every six months, against Moore's Law's own 18-24-month doubling. "To say this is huge is the understatement of the century." At the time nobody — including Jensen — grasped the historical significance. Ben: "Necessity is the mother of invention."
• Effect: This became NVIDIA's true process power for the next decade (see the moat analysis); competitors' processes, orgs, and risk tolerances were locked into the old cycle and couldn't copy it for years.

6. Performance is the only thing buyers care about — and the whole system flows backward from consumer demand

• Story: The rushed RIVA 128 was "a freaking beast," but of Direct3D's roughly 24-25 blend modes, only about two-thirds worked; the rest crashed outright. NVIDIA ran a developer lobbying campaign: "Come on, what do you really need more than these eight for? These eight work great. You're going to love them. Just use those." Every developer complied.
• Insight: Consumers buy cards — and games — on graphics performance; developers target whatever card will be the price-performance king when their game ships. So developers will happily accommodate a chip's defects for performance. This was NVIDIA's first real market education (version one of the company wanted to "create technology and drive the market" while knowing nothing about it), and NVIDIA figured out "performance is everything" before anyone else — because it was forced to.
• Effect: RIVA 128 sold a million units in four months. The mid-2000s Far Cry phenomenon was the same logic at its extreme: gorgeous engine, mediocre game (David lasted under ten minutes), players stacking a thousand gasoline barrels and shooting them to watch underpowered rigs die — one title pulling both software and hardware sales.

7. Fight Intel's "integrate and extinguish" with intelligence that can't be commoditized

• Story: Intel's standing playbook: open the PCI ecosystem, let peripherals bloom, watch which ones get consumer traction, then integrate them into the motherboard or CPU — game over for the peripheral startups. Sound Blaster sound cards and network cards were the precedents (Ben: "These things are dead-end businesses."). Graphics cards had no obvious exemption. NVIDIA's answer came in two moves: in 1999 it marketed the GeForce 256 as the first "GPU" — which "wasn't really true yet" (it wasn't programmable), half marketing bravado, half what David calls "a big middle finger to Intel and this whole CPU-dominant world"; then GeForce 3's programmable shaders and the CG language made it true.
• Insight: A fixed-function accelerator will eventually get integrated; the only defense is to grow programmable intelligence in the chip and graduate from accelerator to processing unit — David's summary of the declaration: "This is no sound card. This is not going to get commoditized." Underneath sits the architectural truth: CPUs are serial (David's analogy: the I Love Lucy chocolate assembly line, wrapping one piece at a time) while pixels don't depend on each other — graphics is natively parallel, a completely different competitive vector from the CPU makers. Ben plants the flag: at the time, everyone assumed graphics was parallelism's only big use case — "put a pin in that."
• Effect: Intel's ~1999 discrete cards became some of the worst-reviewed in history (Ben: "Talk about not your core competency.") and it retreated to integrated graphics — big in markets that don't care about graphics, never competitive where anyone does. CG bound developers' shader code to NVIDIA hardware — the first experiment in CUDA-style ecosystem lock-in. Name the category first, make it true later.

8. Ally with the giants instead of fighting them — and the Microsoft mirror

• Story: Version one of NVIDIA died trying to out-platform Microsoft with its own APIs and standards; the GeForce 3 era did the opposite — the Xbox deal (~$500M a year plus a $200M prepayment) bought Microsoft's backing for the CG programmable-shader language and a shared front against Intel. David: "an A+ move" — the original vision wasn't wrong; the positioning was. But Microsoft extracted its pound of flesh: low-margin revenue that was half the company's ~$1B top line.
• Insight: Ben's summary cuts deepest: "NVIDIA fought for their life and won." David: "Multiple times." Ben: "And Microsoft just leveraged the crap out of their amazing position and probably achieved about the same outcome." "Microsoft wants to play Switzerland… we want to commoditize all of our suppliers." Microsoft sat back while NVIDIA and ATI took all the risk of developing the computer-graphics market, then harvested. Both of NVIDIA's first two existential wars — beating 90 competitors, and proving the GPU wouldn't be commoditized by Intel — made Microsoft enormous money.
• Effect: Ben's setup for Part II: NVIDIA was forced to grow insane competencies in those wars, which ultimately produced CUDA and powered the machine-learning revolution; Microsoft never had to grow that DNA — just as it never grew mobile DNA, and got beaten by Apple. The riskless harvester's price is that it builds no muscle.

9. Capital-allocation discipline: turn down low-margin "great opportunities"

• Story: After the original Xbox, NVIDIA has supplied exactly two more consoles in its entire history: the PlayStation 3 and the Nintendo Switch — three, ever. Jensen's official framing (Ben, paraphrasing his talk): "There are a lot of things we could spend our resources doing. If I don't think that we can do anything really unique and special and really change the world, then we have better things to spend our resources on." Ben's translation: "No, there are crap margins in that, I'm not doing that."
• Insight: With finite resources, every allocation has to pass two screens at once — near-term cash flow and long-term strategy; large but low-margin revenue that builds no unique capability is a temptation to be refused.
• Effect: The Xbox deal itself had taught the lesson at full tuition — it depressed company margins for years (29% gross margin in 2004), and when the Xbox 360 generation went elsewhere, revenue flattened on cue. David's image: out of the Intel frying pan, into the Microsoft fire — exaggerated, he admits, "but there's a lot of truth in it."

10. In a business at the mercy of Moore's Law, the only way to live is ruthless reinvention

• Story: Ben's periodization: NVIDIA's history runs in 6-10-year epochs — a meteoric rise powered by one contrarian move, then a taper, then a forced reinvention. Three so far in this episode: the original rise before 90 competitors flooded in; software emulation and the six-month cycle; programmable shaders — then another plateau.
• Insight: Ben: "Jensen really figured it out early that they were in a business that was totally at the mercy of Moore's Law… There is no way to stay ahead other than ruthless self-examination and completely ending and rebounding the business." David's compression: "Yup, ship faster and reinvent."
• Effect: Ben: "That, to me, is why they survived." It is also why NVIDIA made three separate bet-the-company pivots — the alternative was joining the other 89.

11. Democratize the tools for developers: the original vision comes true

• Story: The 1992 Denny's business plan really had three layers: the accelerator hardware, the API/SDK/framework layer, and a developer ecosystem — and the third was exactly the soft spot LSI CEO Wilf Corrigan skewered in Jensen's resignation meeting: "Who's going to use these and what for? Who makes PC games? Is there a developer ecosystem for this?" (In 1993, outside id Software's Carmack, there barely were 3D PC game developers.)
• Insight: Jensen's company-level thesis (he refuses to call it a vision): 3D graphics would become a new storytelling medium — not pre-recorded, different on every playthrough, and the only entertainment medium that is networked, hence the only truly social, interactive one. The counterfactual is Carmack: without this industry's full software-and-hardware toolchain, telling stories in this medium required being that vanishingly rare creature who is both a genius developer and a great storyteller — what NVIDIA's marketing calls "Da Vinci and Einstein together in one person."
• Effect: The mission carried NVIDIA's first 20 years; today gaming is a $180B/year business, the largest entertainment medium, bigger than Hollywood and music combined. Jensen, Chris, and Curtis's original vision came true — a new artistic platform for storytellers.

12. "Wrong, intellectually, and yet right": the founder premium

• Story: Jensen walked into Sequoia having read three chapters of a how-to-start-a-business book, with half a business plan, and completely botched the pitch — "barfs all over Don." As a dejected Jensen headed for the door, Don Valentine called after him: "Well, that wasn't very good, but Wilf says to give you money. Against my best judgment, based on what you just told me, I'm going to give you money. But if you lose my money, I'll kill you."
• Insight: David's retrospective — Sequoia and Sutter Hill's market thesis was wrong (90-way melee, no differentiation): "They were wrong, intellectually, and yet they were right. Why were they right? They were right because, frankly, of Jensen." The greatest venture returns come in two flavors: the NVIDIAs (team, plan, and thesis all look right; the path is crooked but it works) and what Sequoia's website once called "The Misfits" (the apparently uninvestable) — "They do the Steve Jobs and they do the Jensens." Ben's founder lesson on the mechanics: "Getting a reference from the CEO of a portfolio company is a really good way to come in with a venture capitalist already leaning toward investing" — after Wilf's "Don, I got a kid who's going to come see you, stand by," botching the round was "literally impossible."
• Effect: From a $6M post-money to the eighth most valuable company on Earth; Mark Stevens — the Sequoia GP who joined NVIDIA's board, and David's professor at Stanford GSB — is still on the board and still personally holds the stock. One of the best venture returns in history, full stop.

Moat Analysis (the 7 Powers framework)

7 Powers is Hamilton Helmer's strategy framework (7 Powers: The Foundations of Business Strategy): seven structural advantages that let a company sustain differential returns. Acquired runs every company through the checklist — and this episode's verdict is, unusually, a negative.

The key judgment (Ben): "NVIDIA had definitely found product/market fit, but had not yet found their source of power." And: "They had the inkling of how they could get power, but it was not yet implemented" — the inkling being the CG → CUDA line. Hamilton Helmer and his Strategy Capital colleague Chenyi give the moment its name: after climbing the product/market-fit mountain, founders face a completely separate second journey — developing power ("a whole second invention," in Ben's words). In 2006, NVIDIA was stranded between the two peaks.

Bull & Bear (standing in 2006, just after the AMD-ATI deal):

Bear: 2004 gross margin of 29% (against the 66% this same GPU hardware business earns today) — every drop of economic potential competed away, zero pricing power; ~$500M a year of low-margin Xbox revenue equal to half the company, with the Xbox 360 generation already lost and revenue flat; OPEX surging into HPC/scientific computing, a segment nobody could size — "Jensen, do you really know what you're doing?"; Intel announcing Larrabee, a full-fledged GPU assault; ATI catching up on programmable shaders while the industry whispered that NVIDIA had taken its eye off the gaming ball. The killer question: "Are you betting the farm on scientific computing?" David: "How big is that market?" Bull (Ben): The answer is yes — bet the farm — and that is itself the bull case: "It turns out, scientific computing would be so much more than scientific computing." Everything else in computing accelerates enormously when parallelized. "That is 100% the bull case and 100% of what happened."

Grading: The framework (Ben): given the 1993-2006 market opportunity in computer graphics, how well did NVIDIA exploit it? Value creation: amazing. Value capture: better "than anyone else as far as I could figure out" among 90 companies doing the same thing. The real question is whether someone else on the value chain captured far more — "Would you rather have been Microsoft than NVIDIA?" David's final grade: A — not A+. The A: they were essentially the only company that survived (ATI survived too, in a very different way), and they almost inarguably created and shepherded the entire industry. The missing plus: because of Microsoft — until the DOJ case, the real winner was the company harvesting from the sidelines. Ben: "Hard to argue with it."

Deep Cuts

• The full naming story: Chip design files were saved with the extension .NV — "next version." The founders went through the dictionary for words containing NV and found the Latin invidia — envy: "we'll be the envy of the industry." Drop the leading I: NVIDIA. The green logo followed for "green with envy." Ben files it as a classic Rich Barton "empty vessel name" — with vid buried in it for video.
• Wilf Corrigan's desk: When Jensen resigned from LSI, its CEO (and former Fairchild Semiconductor CEO) asked the killer questions — who's going to use these, and is there a developer ecosystem? — then said: "All right, you'll be back, I'm going to hold your desk." And then picked up the phone anyway: "Don, I got a kid who's going to come see you, stand by." For scale on why that call mattered: LSI's IPO returned $153M to Don Valentine in a day — David estimates roughly 10x the entire Sequoia fund (Fund II or III, $10-15M), possibly the biggest IPO venture return in history at the time.
• The emulation bet in full: $1M, about a third of the cash; NVIDIA was the emulation startup's sole customer ("look, we literally have no options") and the startup later folded — but the chip it saved became NVIDIA's turning point. David: "This is lunacy, what they're doing." No prototype, straight to 100,000 units. Bonus etymology: "tape-out" comes from the days when photolithography masks were literally assembled with tape.
• The Morris Chang phone call: TSMC's salespeople had been ignoring NVIDIA ("TSMC only works with the best and NVIDIA is not the best" — early on it used a second-tier foundry; David guesses possibly UMC, a speculation), so Jensen wrote Morris Chang a physical letter and mailed it to Taiwan. Chang read it and phoned the NVIDIA office directly — mid-chaos, the whole team hand-testing RIVA 128 units as they came off the line. A few seconds of silence, then Chang heard Jensen yell: "Everybody shut up. Morris Chang is on the phone." The multi-year deal came the following year; TSMC has been NVIDIA's foundry ever since, and the two men remain close.
• AMD nearly bought NVIDIA: In 2006, AMD's first choice for a graphics acquisition was NVIDIA — the stock had gone sideways for years and the deal penciled out. Jensen insisted he be CEO of the combined company; the talks detonated, AMD bought ATI, and the rest is history. Forbes ran the cover story "Shoot to Kill." David: one of the great what-ifs.
• The Stanford quantum chemist's phone call (Jensen's favorite story — flagged in the episode as probably apocryphal/composite): a researcher whose molecular models took weeks on Stanford's supercomputer; his gamer son told him to buy a stack of GeForce cards at Fry's and port the model to CG — it ran in hours, matching the supercomputer's answer exactly. "I just want to thank you, Jensen, for making my life's work achievable in my lifetime." David: "For sure, it's something that Jensen made up. Maybe did, maybe didn't." Ben: "It's a composite, but every word of it is true in spirit." Either way it is the teaser for Part II — scientific computing had arrived as a market, with researchers literally hacking together retail GeForce cards.
• The Keyhole investment's double logic: In 2006 a startup nobody would fund pitched Jensen, who got it instantly: "Oh, my God, I see this is the future. This is a simulation. You are creating a model of the earth in software and people can just navigate around the earth." Half worldview match (an earth model in software = simulation, straight toward today's Omniverse), half ecosystem self-interest — keep it alive because it demoed NVIDIA's technology better than anything else. Google later bought it and made it Google Earth.
• Loose gems: GeForce came from the internal naming contest's winning entry, "Geometry Force"; at his first Stanford talk after donating the engineering building, Jensen opened with "I've donated, we have this nice building now, so I have no more money" (David: "I'm penniless. Right, Jensen." — net worth ~$20B); and Don Valentine's "View from the Top" GSB talk — the one David rewatches every year, where Don holds up Alfred Lin's resume — happened to fall on the day the Jensen and Lori Huang Engineering Center opened. Don's comment: "Jensen did a building."

Era & Industry Trivia (tangents worth keeping)

• The stage in 1992-93: SGI workstations ruled 3D — military or Jurassic Park budgets only (SGI founder Jim Clark later did Netscape); consoles were still in the 2D Super Nintendo era; Windows 95 was a year and a half away; and id Software in Texas (John Carmack, John Romero) had just put 3D on consumer PCs through sheer engineering heroics with Wolfenstein 3D and Doom (1993) — the proof point behind the Denny's pitch.
• Direct3D and the graphics API wars: Watching Doom take off, Microsoft decided 3D belonged inside Windows itself (the open-source rival being OpenGL). The attitude, as David voices it: "Developers want to do 3D graphics directly into Windows without any of this crufty middleware from some no-name company like NVIDIA." The 89 followers' logic was brutally rational: "I would be dumb not to."
• Prehistory of the accelerated-computing wave: An ASIC trades general-purpose flexibility for one function hard-coded into physical design; sound cards, network cards, and graphics cards were all the same move — offloading work from the CPU to dedicated silicon. The ASIC's modern descendant is the FPGA, which David plants as some people's bear case on today's NVIDIA (saved for a later episode). Before TSMC's founding in 1987, the chip industry was entirely full-stack.
• The gaming market forks: By 1997 (internet risen, dot-com bubble not yet burst, PlayStation 1 out, PS2 not yet), gaming split into the standardized console market and the smaller but far higher-spend hardcore PC market. Trivia within the trivia: Sony had used the term "GPU" on the PlayStation before 1999 — but nobody marketed it as a category until NVIDIA did.
• Keyhole → Google Earth → Google Maps: Google assembled Google Maps by mashing up three acquisitions — Where2, Keyhole, and ZipDash — each in the $20-30M range (Acquired did an early Google Maps episode).
• TSMC's 30th anniversary gala (all three hours on YouTube): Morris Chang himself plays talk-show host to the CEOs of NVIDIA, ARM, ASML, Qualcomm, and Broadcom — the pillars of the TSMC ecosystem. Conspicuously absent: AMD (no Lisa Su). Jensen tells the letter-and-phone-call story on stage.
• Venture, then vs now: There were no multi-stage venture rounds — "you raised VC, and then hopefully got profitable and went public." NVIDIA raised ~$20M total before its IPO, starting from the $2M round. Against LSI's $153M single-day return, Ben sighs: "Boy, has venture changed as an asset class."
• The DOJ case: The Microsoft of the 1990s-2000s was power incarnate — until the antitrust case genuinely defanged it. David: "probably for good for the ecosystem."
• Carve-out Easter eggs: David recommends the just-released Elden Ring (from the Dark Souls studio, with George R.R. Martin contributing the backstory) — a ready-made proof that games have become the biggest, most ambitious storytelling medium, in perfect rhyme with the episode; Ben returns to Starting Strength (Mark Rippetoe) after a decade — "apparently inspired by Jensen," who has been lifting since age nine.

Cross-domain Notes

This episode has a real crossover with the PH (geopolitics) domain — in the direction of a business-domain, bottom-up supplement to PH narratives, not evidence for PH claims. First, NVIDIA/GPUs/chips are core entities in the PH domain's AI-power-structure narrative (ai-power-structure) and in concept pages like technate and ai-apocalypse: where PH argues top-down about whose hands compute concentrates in and to what end, this episode supplies the corporate-history origin of that compute substrate — parallel computing, the CG → CUDA lock-in lineage, and the TSMC bond all germinate in these thirteen years. Second, the episode carries its own structural observation relevant to PH: the contrast between Microsoft's riskless platform harvesting and NVIDIA's fight-for-your-life muscle-building is necessary prehistory for understanding how AI-era compute power later landed in NVIDIA's hands. Third, the cross-episode Easter egg: Trader Joe's founder Joe Coulombe sat on Denny's board while a high-schooler named Jensen Huang bused tables there — the connector planted in Trader Joe's：反常识的杂货帝国 pays off in this episode (Denny's: Jensen's job, his founding booth, and The Super Bird).

Pages Worth Creating

• Entities: Jensen Huang（黄仁勋） (founder page: the full OBI–Denny's–three-bets arc), TSMC：纯代工模式发明者 (Morris Chang and the foundry model; Acquired has a dedicated episode)
• Episodes: NVIDIA 之二：机器学习公司（2006-2022） (2007-2022: CUDA and the machine-learning revolution), NVIDIA 之三：AI 时代的黎明（2022-2023） (the trilogy's conclusion)
• Concepts: 7 Powers 护城河框架 (Hamilton Helmer's framework, Acquired's standard analytical toolkit), Counter-Positioning（反向定位） (the recurring Power across the series; the six-month emulation pipeline vs the industry's two-year cycle is this episode's variant)

Source · acquired