The In-Sourcing 72 Hours — Tesla Terafab, Intel's Santa Clara EUV Expansion, and Korea's $919B Sovereign Fund Rerouted the Silicon Supply Map
Capex shifted from merchant-vendor order books to sovereign and buyer-owned fabs — a week that put Intel on both sides of the trade at once.
Why these 72 hours are different
The dominant theme of US semiconductor news last week was not who buys the next GPU. It was who takes ownership of the silicon supply chain itself. Three headlines all pointed the same way.
First, South Korea announced a $919 billion sovereign commitment to build 18.4GW of AI data center capacity by 2035 (DCD, 7/1). This is not hyperscaler capex — it is state capex. Samsung Electronics and SK Hynix moved almost immediately to back a domestic chip-hub initiative (Google News, 6/29).
Second, Tesla hired Gary Jiang — a 17-year Intel veteran — to lead Terafab, its in-house chip manufacturing initiative (Google News, 6/30). Tesla did not stop at Dojo. It has begun staffing an actual fab.
Third, Intel broke ground on a Santa Clara expansion for next-generation EUV capacity (Google News, 7/1). Intel is in-sourcing itself.
Intel's two-sided exposure
In the same 72 hours, Intel appears as both defender and talent donor. The Santa Clara EUV build-out is the capex step that keeps its process-node story alive. But the crew that would run that capacity just started leaking to Tesla. That combination — capital present, talent leaving — is the classical failure mode of the Intel foundry narrative.
One data point cuts the other way. Combined market cap for Micron, Intel and AMD (. The market is repricing the CPU, memory and fab layer — not only the GPU vendor.
If this analysis was helpful · ☕ Support Us · ✈️ Telegram