What would happen if a geopolitcal crisis temporarily halted the production of advanced chips in Taiwan? This question has been on the table for years among executives and governments, but it’s gaining renewed urgency following analysis by Ben Bajarin (Creative Strategies), who in an interview with Stratechery described Intel as the “insurance policy” for all major US fabless companies — Apple, NVIDIA, AMD, Qualcomm, among others — facing risks from concentration around TSMC. The core idea is simple: ‘You don’t need it… until you do’.
The context: TSMC leads, and rightly so
Under normal circumstances, fabless companies do not seek alternatives: TSMC meets their needs and does it better, emphasizes Bajarin. The Taiwanese company is the benchmark for cutting-edge nodes — N2 today, A16 on the horizon — with a client list including the most demanding players globally. The demand for future generations is massive, and barring surprises, Apple, NVIDIA, and AMD will continue trusting their roadmaps to Hsinchu.
But that normality exists alongside a concentration risk that worries anyone considering business continuity: Taiwan produces the overwhelming majority of advanced semiconductors and TSMC concentrates a vital portion of that production. If something disrupts their supply capacity, the entire global electronics, AI, and cloud chain suffers.
Intel, “Plan B”… if it’s up to the task
This is where the insurance analogy comes in. Intel, due to geography and ambition, is the only player with genuine potential to provide advanced manufacturing capacity in the U.S. as a backup to TSMC. It’s not about replacing the Taiwanese company in performance or volume today, but rather having it operational in case worst-case scenarios materialize — critical product lines could keep running without interruption.
The nuance is key: for the insurance to work, Team Blue must meet three conditions in its foundry services:
- Sufficient capacity;
- Acceptable yields at advanced nodes;
- Cost and pricing competitiveness compared to TSMC’s benchmarks.
The short-term litmus test involves Intel 18A and 14A processes. If Intel can excel in performance, reliability, and industrialization for these nodes, it can build credibility with fabless clients seeking an alternative ramp-up for critical designs. Failing this, the backup remains underinsured.
TSMC’s counter: diversify without losing the lead
Nevertheless, TSMC isn’t sitting still. The company is expanding U.S. production, advancing cutting-edge processes in its American roadmap, and increasing investments in fabs and packaging outside Taiwan. The goal is adding resilience without sacrificing its technological edge. Still, moving manufacturing from East to West is a multi-year process requiring ongoing political and financial support: enormous capex, talent, and a robust local ecosystem.
Meanwhile, Washington is pressuring to balance production across regions — even proposing redistribution targets — and is offering subsidies for companies manufacturing on U.S. soil. In this landscape, Intel holds a structural advantage: it’s a domestic native company with fabs on U.S. soil, and it’s rebuilding its narrative as a “third-party foundry”.
Why “insurance” is the right word
Because in a normal situation, no one pays for capacity from a supplier they don’t need: TSMC satisfies and optimizes cost/performance better. But in the event of a catastrophic event — a temporary disruption, a geopolitical shock, a chain reaction incident — the presence of an operational supplier in the same country as your decision centers that can absorb at least part of your designs mitigates the existential risk: it avoids factory shutdowns, prevents the halting of hyperscalers, and prevents delays of one or two years in key product launches.
This is why Bajarin says: “You don’t need it… until you do”. And if you want it there when needed, it’s time to invest in that relationship now.
What’s left for Intel
Intel Foundry Services has gained institutional traction — from public contracts, to the interest of hyperscalers — and has aligned its narrative with concentration risks. But to become a genuine option, it needs to:
- Prove in volume that 18A and 14A meet TSMC-level performance at leading nodes with process stability.
- Standardize a catalog of IPs, flows, and tools that reduce friction in porting designs (PDKs, EDA, libraries, advanced packaging).
- Attract several “anchor logos” fabless for learning-by-doing: a few tape-outs are worth more than a thousand promises.
- Close the gap in cost per transistor and performance versus TSMC, even if the goal isn’t to dethrone them but to be a credible second source.
What if TSMC solves everything in the U.S.?
If TSMC consolidates high-performance process nodes in Arizona (and other sites), with high yields and controlled costs, the geographical risk diminishes, and the business case for Intel as a backup narrows. Still, redundancy and multi-sourcing remain good practices in critical supply chains: the mere existence of Intel Foundry with usable capacity maintains market discipline and enhances systemic resilience.
What experts think (and why CFOs care)
Thought leaders like Jim Keller have emphasized that Intel must demonstrate with silicon — not just with slides — that it has technology and operations back on track. CFOs and CSCOs are listening because supply assurance has become a financial objective after lessons learned from 2020–2022. Paying a premium for alternative capacity or for foundry options may be cheaper than risking prolonged supply disruptions impacting the bottom line.
Signals to watch in 2026
- Major fabless tape-outs on 18A/14A with commercial volume (not just risk production).
- Sustained yields and costs competitive enough — even if slightly above TSMC’s — to make second sourcing viable.
- Availability of packaging (EMIB, Foveros, CoWoS equivalents) for AI accelerators and complex SoCs.
- Advancement of TSMC in the U.S.: nodes, timelines, and volumes. The stronger its American footprint, the lower systemic risk, and the more strategic the need for Intel to refine its value proposition.
FAQs
Why don’t fabless companies already sign with Intel Foundry despite recognizing the risk?
Because TSMC meets and optimizes cost/performance better today. Switching foundries isn’t trivial: it involves porting flows, validating IPs, and taking on performance risks. A second source makes sense only if the risk of disruption outweighs the cost of diversification.
Which Intel nodes are the “fire tests” to attract Apple, NVIDIA, or AMD?
18A (and later 14A) are the exams: if these nodes show high yields, competitive packaging, and reasonable costs, Intel can offer credible contracts to high-volume, high-value fabless companies.
Isn’t TSMC already solving risk via fabs in the U.S.?
They are accelerating investments and advancing leading nodes in the U.S., but relocating capacity takes years and demands talent and an ecosystem. Until that volume stabilizes, maintaining an operational backup provider remains prudent.
What does “Intel is insurance” mean financially?
It implies paying a premium — reserved capacity, NRE, engineering — for options you don’t use in normal times but that prevent much larger losses if an adverse scenario occurs. It’s applying risk management to the semiconductor supply chain.
via: wccftech