Is AI a Bubble? | Gavin Baker on Data Centers, GPUs, and the AI Economy

Is AI a Bubble? | Gavin Baker on Data Centers, GPUs, and the AI Economy – The a16z Show

Mentioned

$GOOGL $NVDA OpenAI $AMD $AMZN Anthropic $AVGO Twitter

Market views

Bullish

NVDA— NVIDIA is trading at reasonable 40x P/E (vs Cisco's 150-180x in 2000), has no 'dark GPUs' (all capacity utilized), and customers have seen 10x increase in ROIC from GPU spending. Jensen is described as one of the two best CEOs and playing a strong hand well with rational strategic moves.
GOOGL— Google has seen 150x increase in tokens processed in 17 months, owns the leading TPU chip (NVIDIA's biggest competitor), has DeepMind and Gemini which may be the leading AI company today taking 15-20 points of traffic share, and has Chrome with 5 billion users providing massive distribution advantage.
META— Meta is seen as existential player in AI with strong commitment, described as willing to invest heavily to win. Has data, distribution, compute, and capital advantages as an incumbent that could make AI a sustaining innovation for them.
INFRA— AI infrastructure layer is viewed as safer investment at beginning of technology wave. Major companies investing have $300B annual free cash flow and $500B cash on balance sheet, providing substantial buffer for continued buildout.
FIGMA— Figma successfully communicated gross margin decline from AI distribution to investors who responded positively, demonstrating that public markets will reward companies that embrace lower margins for AI growth.
BROADCOM— Broadcom is effectively partnering with AMD to compete against NVIDIA by building Ethernet-based fabrics and enabling custom ASICs, positioning as key enabler for companies wanting alternatives to NVIDIA's full-stack solution.

Bearish

ASICS— Most high-profile custom ASIC programs are predicted to fail in the next three years, especially if Google starts selling TPUs externally. It took Google three generations to get TPU right, suggesting difficulty for others.
SAAS— Application SaaS companies that try to preserve 90% gross margins will lose to AI competitors. Gross margins are structurally declining due to compute intensity of AI, and companies refusing to accept 60% margins with higher revenue will be disrupted.
ADVERTISING— AI-driven affiliate fee models will squeeze out inefficiencies where advertisers systematically overpaid Google. The shift to outcome-based pricing will be a business model degradation for traditional advertising.

Not investment advice.