Apr 29, 2024 · 1h 5m · bg2-pod
Ep8. AI Models, Data Scaling, Enterprise & Personal AI | BG2 with 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, hosts Brad Gerstner and Bill Gurley, along with guest Sunny Madra, examine Meta's landmark Llama 3 launch, the economics of open-source versus closed AI models, enterprise cloud adoption trends, agentic workflows, and the impact of massive CapEx spending on tech valuation and macroeconomics.
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 78.9% of the talking time here. How this is scored →
Brad and Bill as informed peer 7.9 Guest teaching 4.3 Guest disagreement 2.1 Brad and Bill pushing back 3.1
speaking balance: gold is Brad and Bill, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 32:51 Sonny directly rejects the hosts' traditional ML premise
Sonny directly rejects Bill and Brad's view that traditional ML yields better enterprise ROI, arguing generative AI democratizes high-end capabilities for non-technical firms.
Hardest push from Brad and Bill ▶ 43:56 Bill challenges Brad on Altman's scaling statementsBill refuses Brad's assumption of endless model scaling by citing Sam Altman's direct comments and pressing Brad on why he refuses to take Altman at his word.
Biggest teaching moment ▶ 5:26 Sonny details 10x-50x token price collapseSonny educates the hosts with concrete infrastructure pricing data, demonstrating how Llama 3 70B reduced input and output token costs drastically compared to GPT-4.
Brad and Bill hold their own ▶ 56:12 Bill dismantles VC IPO criticism with LP mechanicsBill demonstrates elite venture market mastery by reframing the IPO delay debate around LP paper marks, fund vintage dilution, and secondary market liquidity valves.
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 |
|---|---|---|---|---|---|---|
| Welcoming Guest Sunny Madra | 7 | 5 | 1 | 1 | Brad and Bill introduce Llama 3 and set the framing around open-source disruption. Sonny provides operational data from Groq on developer migration and token pricing economics. | |
| Open Source vs. Closed Model Monetization | 8 | 2 | 1 | 2 | Bill details his thesis on the commoditization of closed models and the collapse of the consumer $20/month subscription model due to Meta AI being free. Sonny and Brad concur with this market dynamic. | |
| AI Infrastructure, CapEx Escalation, and Strategic Moats | 8 | 5 | 1 | 2 | Bill probes whether Meta's massive CapEx guidance is a deliberate competitive signal to clear the field. Sonny explains how over-investing in training reduces downstream inference costs, while Brad outlines four pillars of AI competitiveness. | |
| Meta's Open Source Strategy & Model Economics | 8 | 4 | 2 | 4 | Bill differentiates Meta's open-source strategy from standard infrastructure by comparing it to Google's Android scorching the strategic earth. Brad adds nuance regarding revenue-sharing with hyperscalers and potential proprietary shifts. | |
| Enterprise AI Adoption and On-Premises Repatriation | 7 | 6 | 1 | 2 | The hosts discuss enterprise cloud growth and Barclays survey data on workload repatriation. Sonny adds insider color on why enterprises do not trust hyperscalers not to train on private data. | |
| Agentic AI, Multi-Step Reasoning, and Enterprise Value Creation | 8 | 6 | 5 | 5 | Sonny explicitly disagrees with the hosts' suggestion that classic ML captures value better than GenAI, arguing GenAI democratizes complex tasks with single prompts. Bill pushes back using Sonny's commoditization arguments. | |
| Personal AI Assistants, On-Device Models, and Latency Optimization | 8 | 5 | 4 | 6 | Bill and Brad debate whether LLM scaling has hit a wall or if Sam Altman is bluffing about model sizes. Sonny explains how on-device small models and low latency unlock consumer use cases like the Humane pin. | |
| Tesla FSD & Robotaxi Update, Plus Tech CapEx Earnings Reactions | 8 | 3 | 2 | 3 | Brad analyzes Tesla's Robotaxi and Uber/Airbnb hybrid network strategy. Bill injects skepticism regarding regulatory driverless licenses compared to Waymo, while Brad defends Meta's capital reinvestment. | |
| Venture Capital, Late-Stage Liquidity, and the IPO Market Debate | 9 | 3 | 2 | 3 | Brad and Bill critique Dan Primack's editorial on VCs delaying IPOs. Bill delivers an in-depth breakdown of LP paper marks, secondaries acting as pressure release valves, and structural venture dynamics. |
Statements from this episode (24)
Madra: Llama 3 became Groq's most popular model in 48 hours
“And for us, we saw within the first 48 hours, it become the most power, most popular model that we run on Grok. And so really replacing what? Replacing Mixtrel eight by seven for us, which was, you know, generally considered the best open source model at that …”
Madra: Developers replace OpenAI with Llama 3 with no performance loss
“And people are doing a direct replacement with open AI across the board. They come to us. So they come to, you know, all the different providers and they replace out open AI. And they don't really see any performance impact or any reasoning impact”
Madra: Llama 3 70B is over 10x cheaper than GPT-4
“Price performance, right? You get probably on, from a GPT-IV, you're more than 10 times cheaper. Right. And you're. Yeah. Yeah. 10 times cheaper. And well, let me just tell you, GPT-IV is 10 dollars per million tokens input and 30 dollars per million token out…”
Gerstner: Over 50% of OpenAI's revenue comes from consumers
“Over 50% of the revenue is, I think, from consumer.”
Gurley: Free Meta AI eliminates consumer willingness to pay $20 for ChatGPT
“As long as Meta's free, I don't think anyone pays the 20 dollars. And I will say that is as things sit today. If some crazy feature comes along, you know, we've talked about personal memory, maybe, maybe that comes back, but for now, It feels dead.”
Gurley: Altman and Amodei speak in platitudes while Zuckerberg discusses technical specifics
“Dario and Sam talk in these high level platitudes about how this stuff's going to cure cancer and we're all going to, Not have to work anymore. And Zuck was down in the weeds in the meat.”
Gurley: Consumer AI Tools Lack Differentiation Needed to Reach Escape Velocity
“Right now, it's hard to believe that any of them, including the Google one, like, are going to have escape velocity because the differentiation, I'm not seeing, maybe you guys are, I'm not seeing an element of differentiation on the consumer facing tool that's…”
Gerstner: Very few companies will be able to build frontier AI models
“I think that the cost structures that are going to be associated with frontier level models, there are going to be very few companies on the planet that are going to be able to build those models. Because I think the latest discussions, whether it's you know, …”
Gerstner: Cloud providers are paying revenue shares to open-source model creators
“I can definitely confirm that the data clouds are paying a revenue share. To the open models.”
Gerstner: Closed AI model startups besides OpenAI will struggle to raise funding
“I think it's going to be very difficult for new entrants to be venture backed because to, you know, open AI will continue to get funding because they have this incredible team. They have a hundred million people using the product and paying them for the produc…”
Gerstner: Economic value is in enterprise relationships, not AI models
“The value is not in the model. Right. Just like the value is not in storage. Right. You could say storage is a part of the AWS cloud, but that there's not a lot of value in that thing unto itself. The value is in the enterprise relationship. The value is in ri…”
Gerstner: 64% of Fortune 500 companies use Azure OpenAI
“64% of Fortune 500 customers are now Azure OpenAI customers, which I thought was pretty extraordinary.”
Madra: 82% of enterprises use or plan to use open-source AI
“82% of the respondents said they are either already on open source or will move to open source.”
Gerstner: 83% of CIOs plan to repatriate workloads on-prem
“So this among enterprise CIOs moving back to hybrid and on-prem, the number was the 83% of respondents said that they were going to repatriate at least some of their workloads. Right. Back to on-prem and that was up from 49% or 43% in 2020.”
Gurley: Compelling enterprise AI use cases rely on traditional ML, not LLMs
“When I meet a company and see them using AI in a way that feels like ultra compelling from us improvement of their own strategic business position, it's almost always a more traditional AI model that's running a very particular optimization problem. It's not a…”
Gerstner: GPT-5 will introduce early memory and agentic actions
“Like I think in GPT-V, it's not going to be the final state, but I think you're going to see the beginnings of memory and the beginnings of actions. Right. And this is, you know, months away and you and I have a bet on this.”
Gerstner: Wearable AI devices are one year away from capable on-device models
“We're a year away, probably max, from that thing being able to have a billion parameter or five hundred million parameter model that basically has all the capability you need it to have.”
Gurley: Tesla has not applied for regulatory driverless permits
“Tesla's, one, they haven't even applied for those things, but we don't have them driverless on the road yet.”
Gerstner: The GPU shortage will not turn into a glut in 2024
“Every supply shortage does ultimately result in a glut. But people have been calling for this glut now for you know, 12 months anyway. And they're calling for it again this year. We're not going to see it again this year.”
Gerstner: Meta raised net income to $55B while cutting headcount to 69K
“In two years, that company has gone from twenty two billion in net income to fifty five billion in net income. They've reduced their headcount from 85,000 people to 69,000 people.”
Gurley: Massive secondary sales eliminate the primary employee pressure driving IPOs
“When you support massive secondaries, you're taking the number one pressure out of the system that used to lead founders to want to go public because their employees are like, I need liquidity. So you do a release valve, and you take that away”
Gurley: No individual venture capitalist can force a company to IPO
“No single VC's gonna stand up and make a company go public. That's not gonna happen.”
Gerstner: Azure AI hits $4B run rate with 7% growth contribution
“Azure AI contributed seven percent of growth this quarter. So it's now translates into about a four billion run rate business You know, that didn't even get broken out until five quarters ago.”
Gerstner: AI benefits concentrate in hyperscalers over smaller tech companies
“My hunch is, That the largest companies in technology, back to your network effects and your scale advantages, I'm not sure that smaller tech technology companies are seeing the benefits that the largest technology companies are. You know, so we'll see that as…”