How did Google create modern AI - and can it still win the AI era?

The direct answer the hosts land on: Google invented nearly every component of modern AI, watched much of it escape through acquisitions that never happened and colleagues who walked out the door, and is now betting that the one full-stack position in the industry - plus founders who cannot be fired - is enough to win the era back. Ben Gilbert's summary of the company today: today they have just about the best collection of assets that you could possibly ask for. They've got a top-tier AI model with Gemini. They don't rely on some public cloud to host their model — they have their own, in Google Cloud, that now does $50 billion in revenue. The question the episode sets out to answer is whether that collection of assets can be turned into the same kind of dominance Google once had over search - or whether AI spreads its value across too many players for any one company, even this one, to capture it.

From twelve-hour translations to Google Brain

The story starts, as the hosts tell it, with a question asked in a Google micro-kitchen around 1999. A researcher wondered whether a system that predicts the next word in a sentence is actually demonstrating understanding - "So this conversation is happening — you know, this is 25 years ago — and Noam, the new hire, the, you know, young buck, he sort of stops in his tracks and he's like, 'Wow, if that's true, that's really profound.'" The new hire was Noam Shazeer, and the question would eventually become the large language model.

The first proof came through translation. Google's early neural translation model was so slow that a single sentence took twelve hours to render; Jeff Dean, the engineer behind much of Google's parallel infrastructure, reorganized the algorithm so words and sentences could be processed concurrently. "And Jeff's work with the team gets that average sentence translation time down from 12 hours to 100 milliseconds." Google Translate shipped, and the pattern was set: neural networks earned their place inside Google's products by making the existing business better.

The academic conversion came through Sebastian Thrun, who brought the field's most important skeptic into the building. "So Sebastian Thrun brings Geoff Hinton into the Google fold after this tech talk, I think first as a consultant over the next couple years, and then — this is amazing — later, Geoff Hinton technically becomes an intern at Google." Out of that relationship came Google's second flagship project of the era: "And that second project is called Google Brain."

The cat paper and AlexNet: when AI started paying

The hosts push back on the idea that the AI era began with ChatGPT. "Yes, the AI era started in 2012 and part of it was the cat paper. The other part of it was what Jensen at NVIDIA always calls the big bang moment for AI, which was AlexNet." The cat paper - Brain's demonstration that a giant unsupervised network could learn what a cat looks like without ever being told - became the foundation of YouTube's recommendation systems. AlexNet did something more visible: it won the 2012 ImageNet competition by a margin nobody had seen before. "So then the 2012 competition — along comes AlexNet. Its error rate was 15%. Still high, but a 10% leap from the previous best, being a 25% error rate, all the way down to 15 in one year. A leap like that had never happened before." "It's 40% better than the next best." The moment made Nvidia's gaming cards the substrate of research AI and, in Jensen Huang's telling, started the industry everyone else arrived in a decade later.

The $550 million hinge: the DeepMind auction

Between the cat paper and ChatGPT sits the deal the hosts treat as the most load-bearing event of modern AI. In late 2013, Mark Zuckerberg moved to buy DeepMind, a London lab with no products and a long way from the general intelligence it promised. "So, Mark calls and offers to buy the company. And there are various rumors of how much Mark offered, but according to Parmy Olson in her book, Supremacy, the reports are that it was up to $800 million. Company with no products and a long way from AGI." Elon Musk then tried to interpose himself: "He immediately calls up Demis and says, 'I will buy the company right now with Tesla stock.'" But Demis Hassabis wanted Google - and, the hosts argue, the feeling was mutual. "They do the diligence. They make the deal. Demis has true kinship with Larry, and it's done. $550 million US."

The counterfactual is the point. "I mean, I think if that had happened, DeepMind would not have gone down in the same way and probably OpenAI would not have gotten founded." In the episode's causal chain, Google buying DeepMind is what convinced the people who would otherwise have funded and joined a rival lab to build one inside Google instead - and it is why the lab that eventually shipped ChatGPT had to be built from scratch.

Ilya's warning and the secret chip

Scaling AlexNet's approach meant buying compute at a scale no research lab had requested. David Rosenthal narrates the plan: "So, they settle on a plan to order 40,000 GPUs from Nvidia." "For a cost of $130 million. 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." That order set Google on a collision course with its own economics, and the collision produced two of the most consequential infrastructure decisions in the industry's history.

The first was the TPU. When Google's new speech-recognition feature threatened to require doubling every data center the company operated, "There's a great quote where Jeff goes to Hölzle and goes: we need another Google." The arithmetic behind it was uncomfortable, as Ben Gilbert puts it: "it looks like we're just going to be shipping hundreds of millions, soon to be billions, of dollars over to Nvidia every year for the foreseeable future." The alternative was a custom chip for exactly one job - the tensor processing unit - built in a satellite office in Madison, Wisconsin, and kept quiet even inside the company. "So, the whole thing was a complete secret for at least a year before they announced it at Google I/O." "The other thing to know about the TPUs is they were done in time for the AlphaGo match. So, that match ran on a single machine with four TPUs in Google Cloud."

The second decision, made in the same period, was to let the outside world in. As the hosts prepare to tell the founding of OpenAI, they read from the press coverage of the dinner that started it: "to quote from an amazing Wired article at the time by Cade Metz, who would later write Genius Makers." Cade Metz's 2016 Wired report on the founding dinner recorded the obstacle in plain terms: the people most qualified to solve those problems already worked for Google, and no one at the dinner was quite sure those thinkers could be lured to a new startup, even with Musk and Altman behind it.

The transformer and the great value leak

Google's neural rewrite of Translate came first: LSTMs cut the error rate dramatically - "And indeed, in 2016, they incorporated into Google Translate these LSTMs. It reduces the error rate by 60%." - and the team threw away the old system wholesale. "There's actually this great New York Times Magazine story that ran in 2016 about it." Google's wholesale replacement of Translate's codebase with the neural network is the story the New York Times Magazine told in its 2016 feature on what it called the great A.I. awakening.

Then came the paper. "So, in perhaps one of the greatest decisions ever for value to humanity and maybe one of the worst corporate decisions ever for Google, Google allows this group of eight researchers to publish the paper under the title Attention Is All You Need." "As of today in 2025, this paper has been cited over 173,000 times in other academic papers, making it currently the seventh most cited paper of the 21st century." What Google did not do was productize it quickly - and the authors did not stay to try. "And also, of course, within a couple years, all eight authors of the transformer paper had left Google to either start or join AI startups, including OpenAI." The oral history of that period exists because the hosts' research relied on it: "Steven Levy wrote a piece in Wired about the history of this. And there are all sorts of quotes from the other members of the team just littered all over this piece, with things like Noam is a magician." Wired's 2024 oral history by Steven Levy collected the transformer team's own recollections of the paper's creation, including the verdict that "Noam is a magician."

The leak had a sharper edge inside the building: Google had a working chatbot years before ChatGPT. "And there is a chatbot in the like late teens, 2020 time frame that Noam has built within Google that arguably is pretty close to ChatGPT. Now, it doesn't have any of the post-training safety that ChatGPT does." The company knew what the interface could be. It chose, for safety and brand reasons, not to ship it - and the people who wanted to ship it left to build it elsewhere.

Microsoft pays, ChatGPT detonates: code red, Bard, and the merger

OpenAI survived its founding-era chaos because Microsoft rescued the scaling plan. At the 2018 Sun Valley conference, "they hash out a deal for Microsoft to invest $1 billion into OpenAI in a combination of both cash and Azure cloud credits." Microsoft invests another $2 billion in OpenAI as the compute bills grew, and the partnership converted a research collective with a dwindling bank account into the best-funded scaling experiment in the industry.

ChatGPT arrived quietly, as the hosts recount the timeline: ChatGPT launched on November 30th, 2022 - Thanksgiving week - as a research preview of an interface to the new GPT-3.5. "Within a week, less than a week actually, it gets 1 million users." Within two months, "it crosses 100 million registered users. The fastest product in history to hit that milestone." Inside Google, the reaction was existential. "Sundar issues a code red within the company" - and, as David Rosenthal frames it, the directive was the textbook incumbent response to disruption: "So this code red that Sundar issues to the company is actually a huge moment because what it means and what he says is we need to build and ship real native AI products ASAP. This is actually what you need to do in the textbook response to a disruptive innovation as the incumbent."

The scramble exposed how far Google had let its own lead rust. Its chatbot research had been capped for years - "For the version of LaMDA chat that is in AI Test Kitchen, they stop all conversations after five turns." - and the rushed Bard launch landed badly. Microsoft escalated while Google stumbled: in January 2023, "Microsoft announces they are investing another $10 billion in OpenAI and says that they now own 49% of the for-profit entity." Google's recovery came through reorganization. "Okay, so that first big decision that Sundar makes is unifying DeepMind and Brain. That was huge. Equally big, he says, I want you guys to go make a new model and we're just going to have one model that is going to be the model for all of Google internally, for all of our AI products externally. It's going to be called Gemini. No more different models, no more different teams. Just one model for everything." Even Google bringing Noam Shazeer home was part of the response: "And then Google ultimately in 2024 after ChatGPT launches, pays $2.7 billion, I think, to do a licensing deal with Character.AI, the net of which Noam comes back to Google." The hosts' reporting on how the reorganization landed comes hedged - "Obviously, this is all playing out in real time, but the point a lot of people at Google made is the Gemini situation is very different than the Google+ situation." - and the company's public posture since has been a deliberate two-sided bet: "It's this like very delicate dance of protecting the existing franchise while also building a hopefully non-cannibalizing as much as we can new franchise."

The parallel bet: from Larry's chauffeur to Waymo's five cities

While the language-model saga ran its course, Google ran a second decade-long bet on autonomy. The starting line was the DARPA Grand Challenge, and its first year was humbling: "So the first year in 2004 went fine. There were some amazing tech demonstrations on these really tight budgets, but ultimately zero of the 100 registered teams finished the race." Back at Google, "this Project Chauffeur that would become Waymo is the first project inside Google X." Sebastian Thrun's team set their benchmark with a challenge named for the co-founder who backed them: "He comes up with these 10 stretches of road in California that he thinks will be very difficult to drive. It's about a thousand miles and the team starts calling it the Larry 1000." The payoff took until fall 2020: "In October of 2020, they launched the first public commercial, no human behind the driver's seat thing in Phoenix. It's the first in the world. This is 11 years after succeeding in the Larry 1000." That arc is the hosts' proof that Google can hold a moonshot open for well over a decade.

Today the numbers are the company's own, and the hosts present them as such: "The study that Waymo just released last month showed that they have 91% fewer crashes with serious injuries or worse compared to the average human driver" - a company-released study, controlled for the fact that Waymos currently drive city surface streets. The operation has real scale: "They're operating in five cities, Phoenix, San Francisco, LA, Austin, and Atlanta. They have hundreds of thousands of paid rides every week." Whether autonomy becomes a business on the order of search remains open - but it is the clearest case in the episode of Google shipping a future it spent a decade funding quietly.

The $370 billion year: cloud turnaround and Gemini's hidden numbers

The business backing all of this is, by the hosts' snapshot, enormous. "over the last 12 months, Google has generated $370 billion in revenue. On the earnings side, they've generated 140 billion over the last 12 months, which is more profit than any other tech company." The market had noticed: "Google surged past their old peak of two trillion and just hit that three trillion mark earlier this month. They're the fourth most valuable company in the world behind Nvidia, Microsoft, and Apple." Its AI distribution numbers are real but deliberately opaque - the hosts complain Google hides the ball better than almost anyone they have covered - "What we do know is there are over 150 million paying subscribers to the Google One bundle." "So, they announced there are now 450 million monthly users of Gemini. Now, that includes everybody who's accessing Nano Banana."

The surprise inside the balance sheet is the cloud business, which the hosts credit to one of the least glamorous decisions in the episode: hiring Thomas Kurian. "In 2020, they crossed 13 billion in revenue, which was nearly tripling in three years. They hired like 10,000 people into the go-to-market organization. I'm not exaggerating that. And that's on a base of 150 people when he came in." The enterprise rebuild carried Google Cloud from roughly $4 billion in 2017 to a $50 billion-plus run rate growing 30% a year - and it matters for AI because the cloud is where Google can hand customers its own chips.

Four pillars, the Jensen tax, and founder control

The episode's strategic core is David Rosenthal's four-pillars claim: to play at the front of AI you need an application, a model, a chip, or a cloud - and Google is trying to hold all four. "Yeah, they have all of the pillars of AI, and I don't think any other company has more than one." The chip pillar is the quietest and possibly the most durable. Everyone renting Nvidia GPUs pays what the hosts call a premium with a nickname: "A lot of people refer to this as the Jensen tax or the Nvidia tax." Google's TPU program reprices that tax: "I have heard that Broadcom has something like a 50% margin when working with Google on the TPU versus Nvidia's 80%. But that's still a huge difference to play with. A 50% gross margin from your supplier or an 80% gross margin from your supplier is the difference between a 2x markup and a 5x markup." That spread is why investor Gavin Baker's argument, as Ben relays it, lands: "So Google being definitively the low-cost provider of tokens because they operate all their own infrastructure and because they have access to low-markup hardware, it actually makes a giant difference and might mean that they are the winner in producing tokens for the world." The scale behind the claim is growing fast: "between April of '25 and June '25, it went from a little under 500 trillion to a little under one quadrillion tokens. Technically 980 trillion" per month across Google services.

The bear case is about value capture, not value creation. "Google makes something like $400-ish dollars per user per year, just based on some napkin math, in the US." AI chat does not monetize like search pages do, and the market structure David expects is not winner-take-all: "They own 90% of the search market. I don't know what they own of the AI market, but it ain't 90%. Is it 25%? I don't know. But at steady state, it probably will be something like 25, maybe up to 50%." A company whose core franchise faces its first real substitution threat has to earn its ending - and the hosts' final answer is about governance, not product. The judge in the antitrust case declined to break anything material up: "Yeah, there was a US versus Google antitrust case. The judge first ruled that Google was a monopoly in internet search and then did not come up with any material remedies." The founders who have controlled the company through every decision in this story remain in charge, and their stated posture is the innovator's-dilemma antidote. As David puts his quintessence: "I mean, Larry and Sergey control the company. They have been quoted repeatedly saying that they would rather go bankrupt than lose at AI." Whether that is enough to win the era is the question the episode leaves deliberately, and correctly, open.

Interview Highlights

18 exchanges

Q
Why was AlexNet so much better than every other entry in ImageNet 2012, and what did Hinton's team figure out?
Ben Gilbert asked49:02
A
David Rosenthal answered
They realized the theory had been sitting around since the 1980s, but the hardware finally caught up. Deep neural networks needed massive parallelism, and Moore's Law had only taken CPUs far enough to build a few layers. The aha moment was rearchitecting neural networks for a completely different class of chip: consumer video game graphics cards, which are inherently highly parallel. It was counterintuitive — normally cutting-edge academic research ran on supercomputers, and instead they used gaming cards that retailed for a couple hundred bucks. Hinton, Krizhevsky and Sutskever walked into a local electronics store, bought two Nvidia GeForce GTX 580s — Nvidia's top-of-the-line gaming cards at the time — rewrote their algorithms in CUDA, Nvidia's programming language, and trained on those two off-the-shelf cards. That is how they did roughly 40 percent better than any other entry in the ImageNet competition. Jensen Huang calls this the big bang moment of the modern AI era: if two gaming cards can do this, imagine more of them, or a purpose-built chip. It set Nvidia on the path from struggling PC gaming accessory maker to the most valuable company in the world, and it showed the field that breakthroughs arrive when you match an old idea to new hardware — then spend years optimizing it.
Q
How did DeepMind end up bidding in Hinton's DNN Research auction while Facebook did not?
Ben Gilbert asked54:54
A
David Rosenthal answered
The timing is the giveaway: the auction ran concurrently with the NIPS conference — then called NIPS, now NeurIPS. Geoff Hinton ran it himself from his hotel room at the Harrah's Casino in Lake Tahoe, and Cade Metz's book The Genius Makers chronicles it. The bidding climbed with Microsoft dropping out and coming back in, while DeepMind dropped out early as promised: a startup with no cash. At the end it came down to Baidu and Google going at each other. At that point the three researchers looked at each other, decided where they actually wanted to land, and stopped the bidding at $44 million with the line: 'Google, this is more than enough money. We're going with you.' The plan had been to split it evenly, but Alex Krizhevsky and Ilya Sutskever went to Hinton and insisted he take 40 percent, keeping 30 percent each — which is how it ends up breaking down.
Q
How did the deal to sell DeepMind to Google actually come together — and why did Larry Page buy it?
Ben Gilbert asked01:19:17
A
David Rosenthal answered
Early in DeepMind's life, Elon Musk was sitting on a private jet with Luke Nosek — PayPal mafia member and DeepMind angel investor — reading Demis's email about DeepMind's AI learning, with no human training, a clever way to win at the Atari game Breakout: bounce the ball around the edge of the bricks and let it ricochet along the top. Larry Page was on that same plane — he and Musk were still close then — and when Larry asked what company this was, that is how Google's founder found out about DeepMind. The deal itself got done on kinship: Larry has always seen Google as an AI company, and Demis felt Larry got the mission. Google already had Brain shipping products, so Larry could say stay in London and keep working on intelligence, and DeepMind could plug into effectively infinite compute. Alan Eustace, carrying Larry's 'Let's really go big' blank check from the strategy offsite, flew to London in December 2013 — chartering a jet with a custom harness rig so that Hinton, who cannot sit, could lie on the floor during takeoff and landing — ran the diligence, and Google bought DeepMind for $550 million, with an independent oversight board for the mission. Had DeepMind stayed independent, it would be worth half a trillion dollars today.
Q
What were the second-order effects of Google buying DeepMind?
Ben Gilbert asked01:30:45
A
David Rosenthal answered
One person was really, really, really upset: Elon Musk — and maybe two if you count Mark Zuckerberg, who plays his cards closer to the vest. Elon had invested in DeepMind and wanted it for his own AI team at Tesla, and now Alan Eustace was flying around the world sucking up every AI researcher while Elon was already souring on Google. That pressure produced one of the most fateful dinners in Silicon Valley history, organized in summer 2015 at the Rosewood Hotel on Sand Hill Road by Elon and Sam Altman, then president of Y Combinator. The pitch: Google — and to a certain extent Facebook — have created a duopoly on AI talent, so let's do something about it. The one researcher they wanted that Google wouldn't let go was Ilya Sutskever; a counter-offer from Jeff Dean reportedly doubled whatever number was floated. Still, seven researchers left Google, and the group founded OpenAI as a nonprofit with a mission to build safe AGI benefiting humanity. A billion dollars was pledged by Musk, Altman, Hoffman, Livingston and Thiel — though it was later reported only about $130 million was ever actually collected, which was plenty for the research and compute of those first years.
Q
After OpenAI raided Google's AI talent, where did Google actually stand — and what happened when Krizhevsky arrived in 2013?
Ben Gilbert asked01:40:22
A
David Rosenthal answered
So back in 2013, when Alex Krizhevsky arrives at Google with Hinton and Ilya, he is shocked to discover that all machine learning runs on CPUs — infrastructure pushes back on GPUs, saying 'we're a CPU shop here.' So he does what any academic would: buys a GPU machine from an electronics store, sticks it in a closet down the hall, and trains like in academia — except Google pays the electricity bill. DistBelief, Google's first-generation deep learning system, needs rearchitecting for GPUs, and by spring 2014 Jeff Dean and John Giannandrea — who would later lead AI at Google and then Apple — hatch the plan to formally put GPUs into the fleet: 40,000 GPUs from Nvidia for $130 million. The proposal is elevated to Larry Page, who personally approves it over his finance team's objection, reasoning that the future of Google is deep learning. For perspective, Nvidia then is a $4-billion-revenue consumer graphics company worth about $10 billion — almost like Google handed Nvidia a secret. The GPUs pay for themselves many times over in Google Photos, Gmail typing suggestions and AdWords click prediction — and running neural networks at that scale exposes the next giant problem: giant matrix multiplications.
Q
Facing billions of dollars a year flowing to Nvidia, how did Google decide to build its own AI chip?
Ben Gilbert asked01:44:21
A
David Rosenthal answered
After Google ships speech recognition on Nexus phones, there's an amazing moment: it only runs on Nexus — no infrastructure exists to serve all of Android. Jeff Dean does the math — three minutes per day across a billion Android phones would require twice the data centers — and delivers the line: 'we need another Google,' or we build a new type of chip. The idea lands with Jonathan Ross noodling on FPGAs in his 20-percent time; that becomes a formal project, then a custom ASIC: the TPU. The trade-off: far more efficient than a GPU, but it only does one thing — the matrix multiplications behind neural nets. The core insight is reduced precision: 4586.8272 becomes 4586.8 becomes 4586, quantization handled in software, so the same power, memory and transistors yield far more calculations per second. The chip is designed, verified, built and deployed in 15 months — FPGAs as the stopgap test fleet, a hard-drive form factor so it slots into existing racks — built in Madison, Wisconsin, kept secret for over a year before Google I/O, and ready just in time for AlphaGo — one machine, four TPUs. Today Google runs an estimated two to three million TPUs against the roughly four million GPUs Nvidia ships in a year: an internal chip effort at almost Nvidia scale.
Q
Where did the transformer come from, and what had Google's language models been doing before it?
Ben Gilbert asked01:52:17
A
David Rosenthal answered
After Franz Och's Google Translate success and the post-AlexNet switch to neural language models, a Jeff-Dean-led group parachuting into Translate proved the concept and threw away the entire old codebase — the New York Times Magazine's famous 2016 story. That rewrite was state-of-the-art but RNN-based, and recurrent networks have a fundamental limitation: they forget too quickly, like reading a book and retaining only the last few pages. LSTMs — long short-term memory networks — fixed some of that with persistent memory, and in 2016 Google shipped an LSTM-based Translate with error rates down roughly 60 percent. But LSTMs are computationally intensive and don't parallelize — the opposite of where hardware was heading. So the Brain team searched for a better architecture. Jakob Uszkoreit's insight was to broaden 'attention': instead of attending to the previous few words, attend to the entire corpus, like professional human translators who read the whole text before translating. Expensive computationally, but extremely parallelizable. Noam Shazeer joined as co-author, pulled a Jeff Dean by rewriting the codebase from scratch, and the transformer crushed LSTM-based Translate — the bigger they made the model, the better the results. When Steven Levy profiled the team for Wired, his lead quote was Noam's three words: 'It works now.'
Q
What came of Google letting its eight researchers publish 'Attention Is All You Need'?
Ben Gilbert asked02:02:23
A
David Rosenthal answered
Obviously the title nods to the classic Beatles song about love. As of 2025 the paper has been cited over 173,000 times and sits seventh among the most-cited papers of the twenty-first century — every paper above it on the list has been out much longer. Within a couple of years, all eight authors had left Google to start or join AI startups, OpenAI among them. Noam Shazeer went off to found Character.AI and later came back to Google through a strange licensing-and-hiring arrangement reported on the order of a few billion dollars — a very, very expensive mistake. 2017 begins what you can call the five-year period of Google not sufficiently seizing the opportunity that it had created with the transformer.
Q
After Elon walked away, where did OpenAI's first billion dollars come from?
Ben Gilbert asked02:08:48
A
David Rosenthal answered
Sam Altman turns to one of the other board members of OpenAI: Reid Hoffman, who had sold LinkedIn to Microsoft a year or so earlier and now sat on Microsoft's board. Reid's intro: come talk to Satya. And do you know where Satya and Sam actually met? The Allen & Company Sun Valley Conference in Idaho, in July 2018. The deal: one billion dollars in cash, plus Azure cloud credits, and in exchange Microsoft gets an exclusive license to OpenAI's technology for Microsoft products. Structurally, the nonprofit OpenAI Inc. creates a captive for-profit, OpenAI LP, and Microsoft's investment goes into the LP; the board includes Hoffman, Altman, Sutskever, Greg Brockman, Adam D'Angelo and Tasha McCauley. Thus the modern OpenAI for-profit-nonprofit-question-mark is created — a structure people are still trying to figure out in 2025. It is, in a sense, the complete history of AI in one deal. And with Microsoft's billion dollars on the table, Microsoft is back as Google's mortal enemy.
Q
Google had a ChatGPT-like chatbot before ChatGPT — why could it never launch?
Ben Gilbert asked02:20:26
A
David Rosenthal answered
Noam Shazeer — the engineer who rearchitected the transformer and co-authored the paper — looked at what his team had built and said: we can use this for a lot more than Google Translate. The transformer paper's last paragraph says almost exactly that: the approach extends to images, audio, video. Google did none of it for a while. Noam's advocacy went further: throw away the search index and the ten blue links, and transform all of Google into one giant transformer model. He built a chatbot interface — before LaMDA; it was called Meena — in the late 2010s, and it was arguably close to ChatGPT in capability. What it didn't have was post-training for safety — its 'who should die'-style answers made it unshippable — and no RLHF, which is what ChatGPT had in 3.5. Even at OpenAI's quality bar it couldn't launch, for two business-model reasons. One: direct answers drop revenue versus ten blue links — a non-starter inside Google until 2021. Two: the legal risk of sitting between publishers and users; Google spent decades fighting the disintermediation perception, shipping info boxes only on non-monetizable queries. And no compelling reason to force it: nobody was really asking — the general public didn't know this was possible, and OpenAI didn't really know either. GPT was out there, but quietly.
Q
How did the ChatGPT launch actually happen?
Ben Gilbert asked02:25:34
A
David Rosenthal answered
Well, I think I do — I have it in my notes here. By late 2022 GPT-3.5 was useful, but how do you productize it? Sam Altman's take: we should make a chatbot — can someone just make a chat interface? Within roughly a week, someone wrapped GPT-3.5 API calls into a chat product. It was a magic product, completely unexpected, and the servers started tipping over; OpenAI found itself cutting deals with Microsoft in real time for more compute, Azure credits and cash advances. Expecting an API-and-B2B business, they threw a Stripe paywall over a weekend to dampen expensive consumer usage — and consumer revenue took off anyway. To say OpenAI had any idea what was coming would be completely false; Ben Thompson's line is that ChatGPT made them an accidental consumer tech company. There's an alternative motivation floating around too — a rumor that Dario Amodei and Anthropic were building a Claude chat interface, so launch before they do. It shipped November 30, 2022, as a low-key research preview; Sam tweeted 'Today we launched ChatGPT. Try talking with it here.' Within a week: one million users. By December 31: 30 million. By the end of January 2023 — two months in — it crossed 100 million registered users, the fastest product in history to that milestone.
Q
What did Sundar Pichai's December 2022 'code red' at Google actually mean?
Ben Gilbert asked02:34:06
A
David Rosenthal answered
Up until this point, Google and Sundar and Larry and everyone had been thinking about AI as a sustaining innovation, in Clayton Christensen's terms — it makes great products better, entrenches incumbents, and is predictable in its capital requirements. ChatGPT changed the category overnight: this is disruptive innovation, an existential threat, and the strengths of ten to twenty years of AI work suddenly look like liabilities, because Google has a lot of castles to protect — filters and safeguards stand between it and simply out-OpenAI-ing OpenAI. So the code red means: we need to build and ship real native AI products ASAP. It is the textbook incumbent response to disruption: don't bury your head in the sand, build comparable products, and be laser-operational about where the new product cannibalizes versus complements. What follows has been a multi-year ballet: protect search growth while creating the best AI experiences. AI Overviews appear for some queries but not all; AI Mode opens to some users but not all; Gemini ships as the full app, but Google.com does not redirect. The delicate dance is protecting the existing franchise while building a hopefully non-cannibalizing new one.
Q
How did Google respond in 2023 — and what were Sundar Pichai's two big decisions?
Ben Gilbert asked02:37:27
A
David Rosenthal answered
Code Red goes out December 2022. In January OpenAI hits a hundred million users and Microsoft puts in another ten billion — 49 percent of the for-profit entity — Google's mortal enemy owns the first true existential threat. In February Microsoft ships Bing with GPT, and Google scrambles: LaMDA plus a chat interface rebranded as Bard ships publicly. Bard is a bad product — missing the RLHF magic of tone and correctness tuning — and the launch video contains a factual error: one of history's worst keynotes, the stock down eight percent that day. In May LaMDA gives way to Brain's PaLM — better, still behind GPT-3.5 and March's GPT-4. Sundar makes two big decisions. First: merge Brain and DeepMind into Google DeepMind, in violation of the original deal terms — Demis becomes CEO, a decade of culture change in a stroke; read between the lines: the models came from the DeepMind side. Second: one model for everything — Gemini serves all of Google internally, all AI products externally. The logic is scaling laws — multiple model families repeat enormous costs — and unlike Google Plus, this is Google's technical wheelhouse. And naming the consumer service after the model — Bard becoming Gemini — is the Gmail move: this product is nothing but technology, a rallying point for Google culture.
Q
What is the origin story of Waymo, from the DARPA Grand Challenge to the Larry 1000?
Ben Gilbert asked02:46:57
A
David Rosenthal answered
Rewind to 2004 and the DARPA Grand Challenge — the seed pool of the self-driving revolution twenty years later: 132 miles of Mojave dirt roads, no humans allowed, a one-million-dollar prize — a break from DARPA's grants-only policy that took an act of Congress. In 2004, zero of a hundred-plus teams finished. In 2005 the pot doubled, twenty-two of twenty-three finalists beat the previous year's best, five finished. Two were Carnegie Mellon's; the other was Stanford's, led by Sebastian Thrun — commodity sensors on a nearly unmodified VW, innovating only in software — very googly. Their early sensor fusion: a precise narrow-field roof laser, a coarse wide-angle camera; a real-time overlay extrapolates safe-area colors forward to find a safe path. Stanford's car passed a Carnegie Mellon team autonomously. Larry hired Thrun — don't raise from Benchmark or Sequoia, we'll hire you — into Street View, and Thrun co-founded Google X, whose first project, Chauffeur, became Waymo. Larry: that self-driving car stuff, do it. Thrun resisted — multi-ton killer robots — until Larry asked what technical reason made it impossible — there wasn't one: 'I'm just afraid.' The Larry 1000 — ten difficult California stretches — was done by a dozen people within eighteen months of the 2009 start; the unreal part is the second 99 percent: ten years to productize.
Q
How did the US v. Google search-monopoly case land, and why did the court decline real remedies?
Ben Gilbert asked03:15:08
A
David Rosenthal answered
Between our Alphabet episode and this one, the United States government v. Google search-monopoly case played out. The judge ruled Google is a monopoly in internet search — and then declined to impose material remedies. There are some remedies, but immaterial ones: no Chrome spinoff, no stopping the tens of billions Google pays Apple for default placement. The reasoning, essentially: Google's a monopoly, but the cost of doing anything about that would have too many downstream consequences — and the judge cited the AI race: with tens of billions flowing into OpenAI, Anthropic and Perplexity, there's a new war on, leave it to the free market, and don't hamstring Google. David is skeptical: none of those challengers generates net income, fundraises don't last forever, cash burns fast — if the spigots dry up, Google faces no self-sustaining competition in search or AI. But the glib-but-meritorious counter: as flat-footed as Google was, avoiding a Microsoft-level distraction and judgment was worth it. There's even a domino theory where Ilya leaving Google leads to OpenAI leads to Google not being broken up — the antitrust forbearance actually saves Google. It's totally wild.
Q
Why did Google Cloud struggle for a decade, and how did Thomas Kurian turn it around?
Ben Gilbert asked03:22:55
A
David Rosenthal answered
We left cloud out of our Alphabet episode — it seemed peripheral then and turned out to be hugely strategic for AI. It started as App Engine in 2008: opinionated platform-as-a-service. AWS proved IaaS right, Microsoft pivoted Azure fast, and Google launched its IaaS only in 2012 — Compute Engine, then GCP. The knock: Google never figured out the enterprise interface — self-serve, advertiser-funded, never forced to delight enterprises. And cloud is a knife fight — lowest price, enterprise relationships, bundling, solutions; you say solution, I hear gross margin. Google was out of its habitat and wouldn't give away the crown jewels — no Bigtable, no Borg. Then came Thomas Kurian, ex-Oracle president, hired late 2018 — and everything changed: cloud took ten years to reach four billion; by 2020 it was thirteen billion, tripled in three years, with ten thousand go-to-market hires on a base of one-fifty. Google was a cloud company all along — best engineers, best infrastructure — but go-to-market and productization were too googly. Twenty twenty-two: twenty-six billion. Twenty twenty-three: a real third cloud, and profitable. Today: a fifty-billion-plus run rate, up thirty percent year over year, the fastest-growing major cloud — riding a religion about serving enterprises, multicloud strategy, and the AI tailwind: Google Cloud sells TPUs while everyone else begs Nvidia for GPU allocations.
Q
Why is Google Cloud the distribution mechanism for AI — and why is Google the only company with every AI pillar?
Ben Gilbert asked03:28:49
A
David Rosenthal answered
One is simply that cloud is the distribution mechanism for AI. To play at all in AI you need one of: a great application, a great model, a great chip, or the cloud — and Google is trying to have all four. No other company has more than one. Nvidia has chips — and, kind of, cloud, but not really. Meta has an application only, and at the moment it's out of the frontier-model race — though the hiring spree might change that. Amazon has the cloud, and maybe an application. Microsoft has the cloud and makes some models. Apple has nothing here; AMD has chips; OpenAI and Anthropic have models — neither has its own data centers at scale, and the chip efforts are not at scale either. Google, meanwhile, has data centers at scale, chips at scale, model usage at scale through google.com's AI Overviews, and applications at scale — every pillar of AI. And it has the very most net-income dollars to lose.
Q
What are the unit economics of AI chips — and why does Google's TPU margin matter so much?
Ben Gilbert asked03:36:45
A
David Rosenthal answered
Everyone pays Nvidia 75 to 80 percent gross margins, implying a four-to-five-x markup on the underlying hardware — the Jensen tax. Pricing power and scarcity: not building your own chips means paying a giant premium. Google pays Broadcom roughly a 50 percent margin on TPU work — the difference between paying a two-x markup versus a five-x one, which at this scale is a giant cost impact. Chips drive AI total cost of ownership: given the pace of next-generation models, depreciation runs five years or less; estimates put over half the cost of an AI data center at chips and their depreciation, people and R&D at a quarter to a third, power at two to six percent. Even Jensen, introducing Blackwell at GTC, said you really don't want Hopper — the H100s everyone chased at our last Nvidia episode. The economics are incredibly sensitive to your chip supplier's margin. One sanity check from Gavin Baker of Atreides Management: historically it hasn't mattered to be the low-cost producer — Google didn't win search by being cheapest, and Apple didn't win on cost. But this era might be different: AI companies run fifty percent gross margins at best, so Google being definitively the low-cost producer of tokens — own infrastructure, low-markup hardware — might decide who wins producing tokens for the world.

Back to Episode Overview View Podcast Channel

How did Google create modern AI - and can it still win the AI era?

Introduction