← Home
The Twenty Minute VC · May 26, 2026 · 01:01:38

Cerebras CEO on the Future of Data Centres, Token Costs and Memory | We are Not in an Infra Bubble & Dario Got a Bad Deal with Elon for Compute | Should US Companies Sell to China & Why Most Layoffs are AI Washed

Andrew Feldman, CEO of Cerebras Systems, discusses the semiconductor company's historic IPO and $70 billion market cap, driven by a $20 billion compute deal with OpenAI. The conversation spans infrastructure investment, the actual cost of AI tokens, competitive positioning of major tech companies, and systemic barriers to enterprise AI adoption—from memory pricing to legal friction.

This summary was generated from show notes and public descriptions, not from a full transcript review. Details may contain inaccuracies.

Novel

Memory Pricing Is a Critical Bottleneck10:39
The cost of memory chips (DRAM and HBM) is becoming a limiting factor in data center economics and token cost reduction.

Highlights

We Are Not in an Infrastructure Bubble05:58
The surge in AI infrastructure spending represents genuine demand-driven investment, not speculative bubble dynamics.
Anthropic's Deal with Elon Was Not Favorable08:58
Anthropic negotiated compute terms with Elon Musk that were available in the market but not strategically optimal.
Google Faces Structural Challenges in Token Production16:40
Despite vertical integration advantages, Google is less positioned than OpenAI to dominate token production due to organizational complexity and legacy business entanglement.
Lawyers Are the Biggest Barrier to Enterprise AI Adoption35:36
Legal and compliance departments, not technology limitations, are the primary blockers preventing enterprises from deploying AI systems at scale.
US Should Restrict Chip Sales to China for Strategic Autonomy39:20
Jensen Huang and Nvidia are wrong to argue for open chip sales to China; US strategic interests require limiting advanced semiconductor exports.

Editorial

Sam Altman's Superpower Is Capex Forecasting08:00
Altman's ability to accurately predict future compute and infrastructure spending needs gives OpenAI and partners a strategic advantage.
Most Layoffs Are AI-Washed Restructuring30:13
Companies are rebranding routine cost-cutting and restructuring as 'AI-driven efficiency' to provide cover for decisions driven by other factors.
The HR Function Transforms in an AI-Driven Workplace34:48
HR's role shifts from hiring, retention, and culture to workforce planning and retraining as AI automation eliminates and transforms job categories.

Misc

Cerebras IPO created 800 millionaires among employees
Sam Altman's superpower is forecasting capex spend for AI infrastructure
Memory pricing is becoming a critical bottleneck in data center economics
Was this useful?