Texas Instruments is making the fabless industry's counter-argument in concrete. Seven new fabs in Texas and Utah, $60 billion committed, targeting 95% internal manufacturing by 2030. The bet is that for analog and mixed-signal silicon, owning your fab is cheaper than the foundry coordination tax at scale. The data TI is pointing at: Q1 2026 datacenter revenue up 90% year over year, industrial up 30%, both driven by embedded analog in power management, sensing, and signal chain. At those growth rates, TI's fab capex is a capacity play, not a diversification hedge.
The mechanism is the analog supply chain's structural difference from digital. Analog process nodes do not race Moore's Law. A 45nm BCDMOS process runs for 20 years. TI's internal foundry model gives them a fixed depreciation curve on nodes that competitors are still fabbing at TSMC or GlobalFoundries, paying market price for wafers. When analog demand is growing 30-90% annually across industrial and datacenter, the marginal cost of an internal wafer versus a merchant wafer becomes the deciding variable. TI's math is that the breakeven on the $60B lands before 2030. The 95% internalization target also removes the lead-time exposure that bit analog supply chains in 2021-2023, when foundry allocation for mature nodes was the bottleneck, not demand.
The structural risk is demand forecasting at seven-fab scale for a market with historically cyclical demand. TI is betting on sustained AI datacenter and industrial build-out through the end of the decade. If the AI capex cycle corrects before the fabs ramp, TI carries idle capacity on owned capital instead of cutting foundry purchase orders. That is the named loser scenario: TI's balance sheet in a 2028 analog downcycle with 95% captive utilization and falling demand. Fabless analog competitors have more flexibility in that scenario. The market will know by 2027 which bet was right, when the first three of the seven fabs are online and utilization data is public.