Is AI a Bubble? | Gavin Baker on Data Centers, GPUs, and the AI Economy
Gavin Baker argues that AI is not a bubble, contrasting today's GPU infrastructure buildout with the 2000 telecom bubble's unused dark fiber. He discusses the positive ROI of AI spending by major tech companies, debates the future of model competition and SaaS, and analyzes the semiconductor competitive landscape dominated by NVIDIA versus Google's TPU.
Summary
In this fireside chat, Gavin Baker, Managing Partner and CIO of Atreides, addresses whether AI represents a bubble by drawing comparisons to the 2000 internet/telecom bubble. Unlike the telecom era when 97% of laid fiber was dark (unused), there are 'no dark GPUs'—all GPU infrastructure is actively being utilized. Baker notes that major tech companies spending on AI have seen significant return on invested capital improvements, with approximately $300 billion in annual free cash flow and $500 billion in cash reserves providing substantial financial cushion.
On frontier model companies, Baker emphasizes the importance of humility, noting that Google hadn't been founded when Netscape dominated the internet. He argues AI could be a sustaining rather than disruptive innovation for large tech companies that already possess data, capital, computing resources, and distribution. Regarding business model margins, he explains that AI labs will structurally have lower gross margins than traditional SaaS due to compute intensity and scaling laws, but this doesn't preclude them from being excellent businesses. He cites the successful cloud transition as precedent for companies accepting margin pressure.
Baker advocates for application SaaS companies to embrace declining margins rather than fight them, warning that resistance is counterproductive given AI's compute requirements. He notes that existing profitable businesses provide runway to develop AI products at breakeven. On consumer AI, Baker expresses concern that AI companies with new browsers may regret competing against Chrome's five billion users and Google's distribution advantage.
Regarding semiconductors, Baker identifies the core competition as between NVIDIA and Google's TPU, with Broadcom and AMD effectively going to market together. He views most custom ASICs as likely to fail without superior execution like Google achieved through three iterations. Baker predicts some high-profile ASIC programs will be canceled, particularly if Google begins selling TPUs externally. On business model evolution, he argues that outcome-based pricing will replace traditional models across industries—from customer service to affiliate commerce—closing inefficiencies that currently benefit advertising-based platforms like Google. Finally, Baker expresses optimism about robotics, predicting Tesla will compete against Chinese manufacturers with humanoid robots learning from video demonstrations.
About this episode
As part of our summer replay series, we're revisiting one of the standout conversations from Runtime, a16z's conference on AI infrastructure and the future of computing. Gavin Baker, Managing Partner and CIO of Atreides Management, joins David George to examine the biggest questions surrounding today's AI investment cycle. Is AI a bubble? What does the unprecedented buildout of data centers, GPUs, and compute infrastructure mean for the economy? And how should investors think about the companies building the next generation of AI? The conversation explores frontier models, Nvidia, Google, custom silicon, AI infrastructure, application software, robotics, and why Baker believes today's AI investment cycle looks fundamentally different from the internet bubble of the early 2000s. Along the way, they discuss the economics of GPUs, enterprise software, AI business models, and what comes next as AI moves from experimentation into the broader economy.
Key Insights
- Baker argues that unlike the 2000 bubble where 97% of fiber was dark (unused), there are currently no dark GPUs—all GPU infrastructure is being actively utilized, indicating real demand rather than speculative overcapacity.
- Baker claims that companies with the largest AI spending have collectively seen approximately 10-point increases in return on invested capital, with major spenders generating $300 billion annually in free cash flow and holding $500 billion in cash reserves.
- Baker contends that AI represents a sustaining rather than disruptive innovation for major tech companies because they already possess the essential ingredients: data, capital, computing resources, distribution, and talent.
- Baker argues that AI labs will structurally have lower gross margins than traditional SaaS companies due to compute intensity and scaling laws, but this lower margin profile need not prevent them from being highly profitable businesses.
- Baker asserts that the semiconductor market will be primarily determined by competition between NVIDIA and Google's TPU, with most custom ASIC programs likely failing unless they match Google's three-generation iterative improvement cycle.
- Baker warns that reasoning capabilities in frontier models have changed AI economics by enabling reinforcement learning flywheels similar to consumer internet companies, potentially favoring companies with large user bases for post-training optimization.
- Baker predicts that outcome-based pricing models will replace traditional pricing across industries, eliminating the systematic overpayment inefficiency that currently exists in advertising-based platforms like Google.
- Baker claims that humanoid robotics have won the morphology debate because they can learn from YouTube videos and allow humans to demonstrate tasks through suit-wearing, making them superior to other robot designs.
Topics
Transcript
Are we in an AI bubble? I do not believe we're in an AI bubble today. I was, depending on how you look at it, the privilege and the misfortune of being a tech investor during the year 2000 bubble, which was really a telecom bubble. And I think it's really helpful to compare and contrast today to the year 2000. The year 2000 internet bubble or telecom bubble was defined by something called dark fiber. At the peak, 97% of the fiber that had been laid was dark. Contrast that with today. There are no dark GPUs. The headlines change. The underlying questions don't. As AI investment continues to reshape the technology landscape, founders and investors are still grappling…
Full transcript available for MurmurCast members
Sign Up to AccessMore from The a16z Show
Before Blockchains, There Was State Machine Replication
Barbara Liskov, a Turing Award winner, discusses her pioneering work in distributed systems, including view stamp replication and Practical Byzantine Fault Tolerance (PBFT), which laid the foundation for modern blockchain protocols. She traces her career evolution from programming languages to distributed computing, and explains how these foundational protocols enable reliable systems even when components fail or act maliciously.
How Bitcoin Rewired a Classic Computer Science Problem
This episode explores how Bitcoin solved the Byzantine fault tolerance problem, a foundational challenge in distributed computing studied for 40 years, and traces how decades of academic research in consensus protocols are now directly informing the design of modern blockchain systems. The conversation highlights the convergence of classical distributed computing theory with practical blockchain implementations, particularly through the transition from proof-of-work to proof-of-stake protocols.
Mark Zuckerberg & Priscilla Chan: How AI Will Help Cure Disease
Mark Zuckerberg and Priscilla Chan discuss the Chan Zuckerberg Initiative's focus on accelerating scientific discovery through AI and biology tools rather than individual therapies. They explain their strategy of building shared infrastructure, data sets, and virtual cell models over 10-15 year horizons to enable faster progress against disease.
Adam Neumann: This Is How You Build Iconic Companies
Adam Neumann discusses his journey from a difficult childhood in Israel through the WeWork collapse to building Flow, a vertically integrated real estate company reimagining residential living through technology, design, and community. Mark Andreessen and Ben Horowitz explain why they backed Neumann post-WeWork, emphasizing how founders who overcome profound challenges often have the greatest potential, and detail Flow's expansion strategy across the US and Saudi Arabia.
Is Software Losing Its Head?
The podcast discusses how enterprise software is fundamentally shifting from being built around human interfaces to becoming 'headless'—optimized for AI agents to access data and logic directly. While incumbents like Salesforce are rebranding existing APIs as agent-ready, the real opportunity lies in new companies building between legacy vendors, leveraging AI to bridge organizational silos and capture the exceptions and context that make enterprise software truly sticky.