In this conversation from a16z’s Runtime conference, Gavin Baker, Managing Partner and CIO of Atreides Management, joins David George, General Partner at a16z, to unpack the macro view of AI: the t">
"Is there an AI bubble?” Gavin Baker and David George
Key Takeaways
The current AI boom differs from the 2000 dot-com bubble due to real demand and GPU utilization.
Major tech companies perceive AI as an existential race, committing substantial capital to infrastructure.
NVIDIA and Google are identified as primary competitors in the AI infrastructure and chip development sector.
AI business models are projected to have lower gross margins compared to traditional SaaS due to compute costs.
New AI-driven business models are emerging, particularly in customer service and consumer interactions.
The evolution of consumer AI and AI browsers is challenging established internet models.
Deep Dive
Gavin Baker contrasts the current AI boom with the 2000 telecom bubble, noting high GPU demand versus past 'dark fiber' surplus.
Unlike 2000, every available GPU unit is currently being utilized in the market.
NVIDIA's current valuation is reported as substantially lower than Cisco's peak valuation in 2000.
Companies investing heavily in AI infrastructure collectively hold $500 billion in cash and generate $300 billion annually in free cash flow.
NVIDIA's primary competitor in the AI sector is identified as Google, leveraging its TPU chips and AI leadership with products like Gemini.
Google's AI business, including Gemini and search overviews, is considered substantial and likely larger than OpenAI's.
Google's internal chip development, including TPUs and Trainiums, is crucial for its competitive AI ambitions.
NVIDIA's strategic responses include supporting labs like Anthropic and seeking improved AI efforts from Meta or a détente with China.
The forecast for AI infrastructure suggests lower gross margins compared to historically high SaaS company margins, due to compute-intensive AI scaling laws.
AI frontier labs are expected to have lower gross margins than traditional SaaS or internet era businesses, given fundamental differences in scaling laws and compute costs.
The guest's perspective on application SaaS has evolved, anticipating significant winners that adapt to AI, potentially accepting lower gross margins.
AI is challenging the traditional internet model where companies direct users to external sites.
AI browsers are anticipated to evolve, potentially integrating more functions directly, though current iterations require refinement for tasks like shopping.
Caution is expressed regarding new AI browser companies given the dominance of existing platforms like Chrome, which has billions of users.
Established players like Google are expected to integrate AI cautiously, having learned from past product failures like Google Buzz.
The competitive landscape for AI hardware includes NVIDIA, Google's TPUs, and the partnership between Broadcom and AMD.
NVIDIA is evolving from a semiconductor company to a systems and data center-level provider.
Broadcom is positioning itself to offer alternative AI infrastructure, including fabrics and TPUs, to companies like Meta.
Amazon's Annapurna team and AMD are noted as important supporting players in the AI infrastructure market.
Platform shifts combined with business model changes are creating disruption opportunities, such as task resolution pricing in customer support.
Customer service is identified as an accessible starting point for new AI business models due to abundant textual data and LLM proficiency.
AI agents could be compensated based on outcomes like customer satisfaction or first-call resolution.
Consumer business models may evolve to include affiliate fees, exemplified by personalized AI assistants for vacation planning or gift recommendations.