Mar 2, 2025 · 1h 3m · bg2-pod
Grok 3, AI Memory & Voice, China, DOGE, Public Market Pull Back | BG2 w/ Bill Gurley & Brad Gerstner · Bg2 Pod
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of BG², tech investors Bill Gurley and Brad Gerstner analyze the rapidly evolving frontier AI landscape, Google's search cannibalization, US-China trade policy, and the macroeconomic impact of federal spending cuts driven by DOGE.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Brad and Bill hold 98% of the talking time here. How this is scored →
Brad and Bill as informed peer 8.3 Guest teaching 1.1 Guest disagreement 0.3 Brad and Bill pushing back 1.1
speaking balance: gold is Brad and Bill, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 36:00 Critique of US AI Export Regulations
Gerstner strongly criticizes US export controls, arguing the Commerce Department's diffusion framework will backfire by guaranteeing global adoption of Huawei's AI chips.
Hardest push from Brad and Bill ▶ 4:10 Challenging Grok 3 Pre-training NarrativeGurley rejects the premise that Grok 3's fast benchmark rise proves pre-training still has significant headroom, asserting it hit the same ceiling as other frontier models.
Biggest teaching moment ▶ 39:44 Satya Nadella on Hyperscaler Compute ScalingSatya Nadella provides a detailed breakdown of how AI agent workloads exponentially increase compute needs via nested program invocation.
Brad and Bill hold their own ▶ 45:40 Gurley on AI Microeconomics and DepreciationGurley delivers an incisive analysis of AI unit economics, highlighting the dangerous 20x price collapse between frontier and non-frontier models.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Brad and Bill as informed peer | Guest teaching | Guest disagreement | Brad and Bill pushing back | Why |
|---|---|---|---|---|---|---|
| Welcome & Podcast Logistics | 8 | 1 | 1 | 2 | Gurley and Gerstner open with an in-depth technical discussion of Grok 3's cluster size in Memphis and pre-training scaling limits. | |
| Grok 3 Consumer Adoption, App Downloads, and Voice Mode | 8 | 1 | 0 | 1 | The hosts analyze consumer app store momentum and compare Grok 3's dedicated voice/UI productization against Meta AI's in-app search integrations. | |
| GPT-4.5, Search vs. AI Wars, and Consumer Scale | 9 | 1 | 0 | 1 | Gerstner and Gurley cite OpenAI metrics (400M weekly active users, $11-12B revenue run rate) and draw parallels to 1990s search engine market consolidation. | |
| Google Search Cannibalization and the Death of SEO | 9 | 1 | 1 | 1 | Gurley and Gerstner discuss the collapse of SEO, 20-40% drops in Google organic clicks, and Google's innovator's dilemma regarding blue links versus AI overviews. | |
| Meta's Position, Product Execution, and the Mobile Parallel | 8 | 1 | 0 | 1 | The hosts discuss Meta's mobile transition history and whether Google can mount a similar structural pivot in search agent workflows. | |
| DeepSeek vs. Llama, Anthropic, and X Platform Strengths | 8 | 1 | 0 | 1 | Gerstner notes DeepSeek R1 disrupting Llama in enterprise inference deployments, while evaluating X's real-time news surface area. | |
| X/Grok Momentum and Perplexity Acquisition Potential | 7 | 1 | 0 | 1 | The hosts evaluate Perplexity's M&A prospects, high private valuations ($8-9B), and potential channel conflicts for acquirers like Microsoft. | |
| Four Vectors for AI Lock-In: Memory, Voice, Features, and Network Effects | 8 | 1 | 0 | 1 | Gurley details four vectors for AI defensibility (memory, voice, features, network effects), while Gerstner shares a personal test of advanced voice mode prompting. | |
| Transition to Product Companies, OpenAI Losses, and Apple's Strategy | 8 | 1 | 1 | 1 | The hosts examine leaked OpenAI financials showing $20B annual projected cash burn, evaluating Apple's device lock-in versus AI execution risks. | |
| US-China AI Competition and Diffusion Framework Restrictions | 9 | 1 | 1 | 1 | Gurley and Gerstner critique US policy regarding China, arguing that the diffusion rule backfires by expanding Huawei's global AI footprint. | |
| AI Microeconomics, CapEx vs. OpEx, and Data Center Scale | 9 | 3 | 1 | 2 | Features Satya Nadella explaining AI infrastructure compute ratios and agent-driven test-time demand, followed by Gurley and Gerstner dissecting hyperscaler CapEx and lease hedges. | |
| Business Resiliency, Unit Economics, and Capital Discipline | 9 | 1 | 0 | 1 | Gurley breaks down the microeconomics and rapid depreciation of GPU models, contrasting excess capital with Tesla's unit-cost discipline. | |
| DOGE, Federal Spending Cuts, and Historical Precedent | 8 | 1 | 0 | 1 | Gerstner analyzes macroeconomic effects of DOGE federal budget downsizing, comparing current austerity to 1990s Clinton surplus dynamics. | |
| Economic Impact of Austerity, Tariffs, and Government Software Cuts | 8 | 1 | 0 | 1 | The hosts discuss how government spending cuts impact vendor revenues, citing 50% drops in government airline travel and software seat reductions. | |
| Global Defense Cut Proposals, Market Caution, and Fiscal Rebalancing | 8 | 1 | 0 | 1 | Gerstner and Gurley discuss global defense budget proposals, Buffett's record cash reserve, and U.S. Treasury rotation dynamics. |
Statements from this episode (42)
Gurley: xAI's Memphis facility was the largest contiguous cluster in the world
“And it was the largest
You know, contiguous cluster in the world, and there was a lot of chatter about that ahead of time.”
Gerstner: Only five or six players can compete in frontier AI
“When we say they all, I mean, we're really only talking about five or six players who have a chance to be in this game at this point.”
Gurley: Building bigger clusters and more LLM parameters won't yield much
“I've been concerned that the way an LLM works and the way it's optimized that building bigger clusters and more parameters won't buy you much.
Right.
Whether I've said it or not, Ilya said it, Andreessen said it, other people said the same thing, and so to m…”
Gerstner: Grok 3 hit number one on the iPhone App Store
“Grok three rocketed to the top of all app downloads you know, on the iPhone charts.”
Gerstner: Meta AI productization is just an unimpressive search box
“As incredible as I think Zuckerberg and Meta are and the advancements they made, I have not particularly been impressed by the productization of Meta AI, right? It's basically just a search box stuck at the top of Instagram or stuck in my WhatsApp thread.”
Gurley: Only DeepSeek and Grok broke into App Store top 10
“And only deep seek and Grok Of all the others have shown the ability to break into the top 10 on the app store download.”
Gerstner: Winning consumer AI platform will capture 70% to 80% market share
“So I don't think it's going to be win or take as much I don't think we're going to see a 99% monopoly here, but I do expect that, you know, we're going to see 70 or 80% share go to the winner.”
Gerstner: Public companies report Google organic clicks down 20-40% YTD
“We have public companies now reporting, that are reporting their Google organic clicks are down 20 to 40% year to date.”
Gerstner: SEO is dead on Google Search
“SEO is dead. And just, you know, those are the free links. That was the core product that used to attract everybody to Google. And the idea that SEO is basically now gone is pretty, pretty.”
Gerstner: ChatGPT cannibalized 80% of my Google search activity
“The amount of activity that I used to do on Google has been 80% cannibalized by ChatGPT because there's search embedded within ChatGPT.”
Gerstner: Google's product velocity on consumer AI has been unimpressive
“Their velocity on product has not been impressive. The velocity on consumer has not been impressive, so they have them. They've had them for a long time, Bill. They had ChatGPT before ChatGPT.”
Gurley: ChatGPT cannibalizes informational search, not high-monetizing commerce searches
“Informational searches are what, Check GPT cannibalize first, not the commerce searches, which is where most of the money is on the paid.”
Gerstner: Google's search moat is distribution and mindshare, not technology
“Google's moat was not a technological moat with search. Their moat was a distribution moat. Their moat was a mindshare moat.”
Gerstner: Letting ChatGPT launch first was Google's mortal sin
“The only thing that could attack Google was never anything head on. It had to be an orthogonal attack from something that was 10 X better, a hundred X better because it gave us answers instead of blue links, right? That's why it was such a mortal sin for them …”
Gerstner: DeepSeek has displaced Meta's Llama in enterprise AI experimentation
“I heard from several inference players that you and I are friends with that all of a sudden DeepSeek rather than Lama is the enterprise open source model of choice that everybody's experimenting with and playing with.”
Gerstner: Meta Will Overcome Slow Start in AI and Deliver Products
“I think they will come through, and remember, when it comes to almost all product stuff, stories, copy, you know, catching up with Snapchat, or whether it's reels catching up with TikTok, they've always showed up to the party late, but they are grinders, and t…”
Gerstner: Anthropic has conceded the consumer AI market
“No, they've, they pretty much seeded the game on consumer.”
Gerstner: X is the top news platform in every country worldwide
“X to me is so interesting because they have a platform that is the number one news platform in every country on the planet.”
Gerstner: OpenAI is accelerating and widening its consumer AI lead
“The odds on favorite there has to be open AI. I think they continue to widen their gap, by the way, Bill, I think they're accelerating at scale, but that's where the race is.”
Gurley: Perplexity looks like an acquisition candidate for tech incumbents
“It kind of looks like an acquisition candidate to me. I don't know if anyone can agree on price, but for one of these other players that hasn't been as successful from a product standpoint.”
Gerstner: AI memory dramatically increases consumer switching costs
“If you get memory, the switching costs explode.”
Gurley: Leaked OpenAI forecasts show $20B annual losses in 2025 and 2026
“There was someone published the internal forecast of OpenAI, which included, I think, in 25 and 26, losing twenty billion each year.”
Gerstner: Apple has opted out of building frontier AI models
“Apple has self-selected out of the race, right? They're not building a big model.”
Gerstner: ByteDance's chatbot is the number one AI assistant in China
“Their AI, you know, their chat GPT equivalent is number one in China, right?”
Gerstner: China will develop frontier AI despite US restrictions
“China is going to have frontier AI and almost all of the things we do to try to slow them down and stop them are backfiring on the United States.”
Gerstner: US diffusion rules guarantee global adoption of Huawei AI chips
“And it almost guarantees a Huawei level belt and road initiative around the world. And the world's going to run on Huawei AI chips, which gives them then the demand That they need to build a frontier AI chip.”
Gurley: Excessive Washington regulation is the number one risk for Nvidia
“And if I owned Nvidia, my number one concern. Would be excessive regulation coming out of Washington. My number one concern.”
Gerstner: AI model inference will deliver strong gross margins at maturity
“I would just posit for you that you'll be able to come up with a variable expense structure using the right mix of models that will be a great margin. May not be as high as retrieval was for Google, but still a great margin.”
Gerstner: Microsoft is spending 80% of free cash flow on CapEx
“Microsoft, I think, is spending 80% of their free cash flow, right, on CapEx.”
Nadella: AI agents will exponentially increase compute infrastructure demand
“When you think of an AI agent, it turns out the AI agents Is going to exponentially increase compute usage because you now are not even bound by just one human invoking a program. It's one human invoking programs that invoke lots more programs. And so that's g…”
Gerstner: Frontier AI race is closed to everyone except Musk and Altman
“And so if you said, is this still an open sport? I'd say no way, right? I don't know anybody else other than the Elon and Sam at this point.”
Gurley: Price difference between current and older AI models is 20x
“The price difference between today's model. Yes. And yesterday's model is 20 X.”
Gerstner: AI models rapidly depreciate once off the frontier
“So it is a fast depreciating asset. The second you're off the frontier.”
Gerstner: US collected $56 billion in tariffs last year
“Last year we had fifty six billion dollars of tariffs imposed on other countries. That's the amount of revenue we collected from tariffs.”
Gerstner: DOGE will downsize federal government by 40% to 50%
“The message is very clear that I think there's going to be a downsizing of the federal government to the tune of, let's call it, 40 or 50%.”
Gerstner: Clinton downsized government by 10-20% and reached $230B surplus
“I don't know the exact percentage of federal employees they let go. It was like between 10 and 20%, but we had a balanced budget, ah, you know, in three fiscal years, we had a two hundred and thirty billion dollar surplus.”
Gerstner: US will cut $500B to $1T from federal spending over next few years
“It does mean that we're probably going to take five hundred billion to a trillion dollars out of federal spending over the course of the next couple of years.”
Gerstner: I am holding half my normal equity exposure amid macro uncertainty
“My risk profile is lower than our standard risk profile. What do I mean by that? Very simply, you know, I own half as much as I would normally own at a point in time.”
Gerstner: Public markets could see a 10% to 15% drawdown from slowing growth
“Which means we could see just a random run of the mill, 10 to 15% drawdown, right, in the markets while the market gets its head around the fact that the economy is going to grow a little slower.”
Gerstner: Government airline ticket sales at one major carrier are down 50% YTD
“I won't say the exact company, but I talked to an airline the other day. And at this airline, their number of government tickets sold year to date is down 50%.”