No Priors: Artificial Intelligence | Technology | Startups
Re-engineering the Semiconductor Supply Chain with Intel CEO Lip-Bu Tan
18 Jun · 45 min
Key points
- Intel CEO Lip-Bu Tan says he took the job at 66 because saving Intel matters to the semiconductor ecosystem and the United States.
- Tan says Jensen Huang invested $5bn in Intel, a stake now worth $25bn, and the US government is also a major shareholder.
- Tan says AI will have a bigger impact than the internet, but power, helium, memory shortages and years-long fab buildouts are the bottlenecks.
Intel's CEO says chip supply, not demand, limits AI
Intel's turnaround is a bet that AI is limited by chip supply, not demand, and that the US needs domestic manufacturing to meet it. Lip-Bu Tan is rebuilding the balance sheet, culture and foundry discipline first. He is also reaching into new materials and packaging because physical limits mean cost and power will not scale like performance.
Takeaways
- Tan took Intel to save a strategic American asset. Tan, 66, was told to retire but took the job because Intel matters to the semiconductor ecosystem and the United States. Trump first asked him to resign over a conflict of interest, then let him continue after hearing his background.
- The first fix was balance sheet and culture. Tan started by strengthening the balance sheet, with the US government as a major shareholder, plus help from SoftBank. Jensen Huang put in $5bn, now worth $25bn. Internally he pushed accountability, faster decisions, and put all engineering under himself.
- Foundry is a bet on trust and discipline. Tan doubled down on foundry because domestic manufacturing is critical and a supply chain cannot rely on one or two players. Winning customers requires yield, defect density and cycle time discipline. The 14A process is 1.4 nanometers, with one and 0.7 nanometer planned.
- AI demand is real and supply is the constraint. Agentic AI is lifting CPU demand, with the CPU to GPU ratio in training moving from one to eight toward one to four or one to one. Power, helium and memory shortages limit growth, and fabs take years to build.
- New materials are needed as physics bites. Performance can double but cost and power cannot, so Tan is hiring material scientists. Intel is investing in gallium nitride, silicon carbide, indium phosphide, glass packaging and diamond foundry. Power conversion from 40 volts to 1 volt is a named bottleneck.
- Musk's TeraFab shares Tan's view of the gap. Tan and Elon Musk both think semiconductor infrastructure has not kept pace with AI. Musk is building his own fab using Intel's technology and process help, since his cars and robots need a great deal of silicon.
- Tan targets a 10x return by 2030 to 2032. Tan sets a 10x goal over five to ten years and expects Intel's value to surface by 2030 to 2032. He points to Cadence, which returned about 76 times when he stepped down as CEO. Intel is near a $600bn market cap.
Tan's investment thesis
- Find a problem that really needs solving.
- Find a partner to work with on it.
- Judge how large and durable the application is.
- Double or triple down on the large ones.