Nov 21, 2024 · 1h 3m · bg2-pod
Ep20. AI Scaling Laws, DOGE, FSD 13, Trump Markets | 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, Brad Gerstner and Bill Gurley analyze major shifts across technology, macroeconomics, and government policy, focusing on AI scaling laws, autonomous vehicle regulation, the DOGE efficiency framework, and the 2025 market outlook.
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 98.9% of the talking time here. How this is scored →
Brad and Bill as informed peer 7.7 Guest teaching 2.3 Guest disagreement 2.5 Brad and Bill pushing back 2.8
speaking balance: gold is Brad and Bill, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 36:35 Gurley challenges Gerstner on shared autonomous vehicles
Gurley dismissively rejects the premise that ordinary Tesla owners will casually rent out their personal cars for robotaxi fleets, citing inconvenience and household dynamics.
Hardest push from Brad and Bill ▶ 37:15 Gerstner counters with fleet arbitrage economicsGerstner refuses Gurley's skepticism by explaining how fleet arbitrage operators will exploit low leasing rates to build commercial fleets just like early Uber drivers did.
Biggest teaching moment ▶ 3:20 Gurley outlines technical constraints on LLM scalingGurley educates listeners on the specific mathematical and data ceiling constraints limiting LLM pre-training advances beyond simple compute expansion.
Brad and Bill hold their own ▶ 52:02 Gerstner demonstrates federal budget deficit mathGerstner breaks down exact multi-year revenue projections and budget cuts required to reach a balanced federal budget without draconian disruptions.
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 |
|---|---|---|---|---|---|---|
| BG² Pod Opening Title Sequence | 8 | 3 | 3 | 4 | Bill Gurley and Brad Gerstner analyze the potential slowing of LLM pre-training scaling laws, contrasting statements from AI lab leaders with Nvidia earnings commentary. Gurley articulates technical constraints regarding parameter saturation, context windows, and synthetic data quality. Both hosts operate as deep domain experts comparing hardware and architectural nuances. | |
| Hardware Infrastructure, AI Memory Unlocks, and Enterprise Cloud Growth | 8 | 2 | 2 | 2 | The conversation shifts into inference workloads, memory innovations, and hyperscaler software growth. Gerstner details metrics across Microsoft Azure and Snowflake to argue software acceleration remains intact despite pre-training commoditization. Gurley adds technical nuance around hardware clusters, context windows, and alternative chips. | |
| Federal FSD Regulation, Tesla v13 Progress, and OEM Market Structure | 8 | 3 | 4 | 4 | Gerstner reviews miles per critical disengagement data for Tesla FSD v13 versus Waymo, predicting an impending consumer robotaxi pivot. Gurley provides skeptical friction regarding fleet management, teleoperations, consumer willingness to share private vehicles, and OEM market consolidation. The debate features active sparring over consumer behavior and OEM survival. | |
| Autonomous Vehicle Liability and Insurance Reform | 7 | 2 | 2 | 2 | Gurley highlights the necessity of tort and liability limits for autonomous vehicles before transitioning to the newly announced Department of Government Efficiency (DOGE). Gerstner reviews the historical context of impoundment authority and executive branch powers. Both co-hosts align on the disruptive potential of aggressive deregulation and fiscal scrutiny. | |
| Federal Audits, Deficit Math, and Balanced Budget Feasibility | 7 | 2 | 2 | 2 | Gerstner details mathematical modeling on balancing the federal budget over a 4-year period through modest cost deceleration and revenue growth. Gurley emphasizes the absurdity of major federal agencies routinely failing statutory audits. Both hosts agree on structural efficiency benchmarks modeled after corporate standards. | |
| 2025 Market Macro Setup: Deregulation, Tariffs, and Tech Dispersion | 8 | 2 | 2 | 3 | Gerstner provides a macro outlook for 2025 equity markets, evaluating the divergence between consumer-facing retail exposed to tariffs and margin-expanding enterprise tech. Gurley questions consumer strength amidst rising auto loan delinquencies. The co-hosts conclude that wide dispersion will reward disciplined stock picking over broad index investing. |
Statements from this episode (20)
Gurley: Scaling variables in fitting algorithms eventually stops adding value
“Historically, Mathematical algorithms that do some type of fit, and this is a very sophisticated form of that, but when you take the variables up to a certain level, they, it stops adding value. You just get too close to the fit.”
Gurley: Nvidia's competitive differentiation is greatest at largest cluster sizes
“One, the NVIDIA differentiation, as we've talked about, is greatest at the largest cluster size.”
Gerstner: 2025 AI model gains will increasingly come from post-training and inference
“So I think at least As we look at 2025 for NVIDIA, for the supply chain, and for the model companies themselves, we think we're going to continue to see improvements in the models. It won't all come from pre-training. We think there's an increasing amount comi…”
Gerstner: Microsoft said almost all $10B AI revenue is inference
“Microsoft has said of its ten billion dollars in AI revenue, it's almost all inference.”
Gerstner: Nvidia stated half its revenue comes from inference
“We know Nvidia has said half of their revenue is inference”
Gerstner: Next giant AI breakthrough will be agents and actions
“We will have better and better models, but I think the next giant breakthrough will be with agents and their actions.”
Gerstner: OpenAI could 2x or 3x ChatGPT users without core model improvements
“OpenAI could double, triple the number of users of ChatGPT with frankly, not making any improvements in the core model because consumers today barely scratched the surface on core model capabilities.”
Gerstner: Tesla FSD 12.5 delivered 100x improvement in MPCI
“By the middle of this year when they released 12.5, it showed a hundred X improvement in miles per critical disengagement.”
Gerstner: Tesla FSD 13 will deliver another 10x MPCI improvement
“And then with the launch of FSD 13 in the next couple of weeks, we expect another 10 X improvement. In terms of that MPCI”
Gerstner: Tesla Robotaxi will likely launch in select cities by Q2 2025
“I think next year really is the year of achievement of a safety standard that allows RoboTaxi to go into action, probably in Q two. They'll pick a couple of cities.”
Gerstner: Tesla Robotaxis will require Hardware 4 compute
“It's going to require hardware for in order to run all these cars.”
Gerstner: Autonomous shift could shrink global automakers from 20 to 3 or 4
“I think there is a path where this is that event, that extinction event, tectonic event that leads, takes us from 20 OEMs down to three or four. And if we sat here in five or six or seven years and we were on a path to three or four global automakers, it would…”
Gurley: Waymo and Tesla Licensing Tech to Incumbent Automakers Seems Likely
“Both Waymo and Tesla have ended at potentially licensing. And so, that would be an interesting maybe that happens down the road between the, you know, the historical incumbents and these two new companies and whether that type of model plays out seems quite li…”
Gurley: Autonomous Vehicles Face Severe Threat Without Federal Liability Reform
“I do think that we have this huge problem in America where litigation is so rampant and I can just imagine some lawyer in front of a jury talking about how the computer killed somebody and trying to extract tens or hundreds of millions of dollars. And that cou…”
Gurley: Americans will back DOGE once wasteful spending is highlighted
“And I think when you highlight to the American population, really idiotic spend,
And they're going to get behind this thing.”
Gerstner: Bipartisan budget committee members want to support DOGE
“Over the weekend, I talked to several members of Congress, both in the Senate and in the House on, on the Budget Committee, and I will tell you there is strong agreement, and it wasn't just from Republicans, right? There were people who were trying to figure o…”
Gerstner: Musk and Ramaswamy Called a 4-Year Balanced Budget Plan Unambitious
“And I shared that with Vivek and Elon, and they both had separate and independent reactions, which is not nearly ambitious enough, right?”
Gurley: Milei May Have Had Fastest Impact of Any Country's President
“If what he said is true about what's been accomplished, it may be the fastest impact a president of a country has ever had on, on, on the trajectory of that company country.”
Gerstner: Altimeter shorted Target over tariff fears and weak discretionary consumer spending
“On the one hand, we were short a little bit of target because we're worried about tariffs and we're worried about the retail consumer holding up, especially around You know, these highly levered consumers when it comes to housewares and things that Target has …”