Controlled infrastructure scenario analysis — exploring how supply chain changes could propagate across the AI chip ecosystem.
All figures are illustrative estimates only. Sourced from public announcements and analyst reports — approximated for demo purposes. Not for investment, trading, or business decisions.
These scenarios explore hypothetical infrastructure changes and their potential downstream effects. They are analytical thought experiments — not forecasts, investment advice, or predictive models. All magnitudes are illustrative estimates.
A significant disruption to Taiwan semiconductor operations — whether geopolitical, natural disaster, or operational — affecting 30%+ of TSMC output for 6+ months.
TSMC fab output disrupted by ≥30% for 6+ months due to geopolitical or operational event. [PURELY HYPOTHETICAL SCENARIO]
Near-total halt of NVIDIA, AMD, Apple, Qualcomm chip production. No viable alternative at N3/N2 scale. [scenario est.]
AI cluster expansion halted for 18–36 months globally. Existing deployments continue but new buildout freezes. [scenario est.]
CHIPS Act emergency activation. Defense and national security chips prioritized. Consumer/commercial market severely constrained. [scenario est.]
Intel IFS becomes critical national security asset. Emergency government procurement. 18A readiness becomes existential. [scenario est.]
Samsung Foundry receives emergency orders it cannot immediately fulfill due to scale gap. Partial alternative. [scenario est.]
NVIDIA Rubin GPU demand for HBM4 significantly outpaces combined SK Hynix + Samsung + Micron HBM4 ramp capacity in 2026.
Rubin GPU demand for HBM4 exceeds total industry HBM4 supply by ≥40% in 2026. [scenario est.]
Rubin ramp constrained by HBM4 allocation. Hyperscaler delivery pushed 2–4 quarters. Premium pricing. [scenario est.]
SK Hynix HBM4 fully committed to NVIDIA. Samsung and Micron HBM4 also fully allocated. Combined still insufficient. [scenario est.]
AMD MI400 HBM4 allocation compressed. AMD either accepts smaller die or delays ramp. [scenario est.]
Google TPU v6 benefits from relative HBM supply advantage as custom silicon avoids competitive allocation. [scenario est.]
Intel 18A achieves competitive yields with TSMC N2, unlocking significant external customer wins and establishing Intel as a viable advanced foundry.
Intel 18A yields reach ≥80% risk production by Q2 2026 with confirmed external customer tape-outs. [scenario est.]
TSMC faces first meaningful competitive pressure at leading edge since 2016. Revenue share erosion at 2nm-class begins. [scenario est.]
Qualcomm/Broadcom consider Intel 18A as TSMC N2 alternative — diversification benefit. [scenario est.]
U.S. gains second world-class leading-edge foundry. Taiwan concentration risk drops meaningfully over 5-year horizon. [scenario est.]
Intel IFS credibly enters leading-edge foundry competition. External customer tape-outs accelerate. [scenario est.]
TSMC successfully executes its CoWoS-L expansion plan, doubling capacity from current levels by end of 2025. What changes downstream?
TSMC CoWoS-L capacity reaches 2× current levels by Q4 2025 through Chunan + Kaohsiung expansion.
HBM3e becomes the new binding constraint as CoWoS-L availability improves. Shifts bottleneck upstream. [scenario est.]
AMD gains additional CoWoS-L allocation, enabling accelerated MI350X ramp. [scenario est.]
B200 supply constraint eases significantly. Hyperscaler delivery timelines compress by ~2 quarters. [scenario est.]
Bottleneck score drops from ~82 to ~45. Ceases to be primary limiting factor for AI GPU supply. [scenario est.]
Geographic concentration risk unchanged — expansion still Taiwan-based. Sovereignty risk unaddressed. [scenario est.]
Amkor Arizona and CHIPS Act-funded facilities successfully achieve CoWoS-equivalent capacity, establishing domestic advanced packaging at meaningful scale.
U.S. domestic advanced packaging (CoWoS-equiv.) reaches 15% of current TSMC CoWoS-L capacity by 2027. [scenario est.]
Modest shift of packaging demand away from Taiwan for U.S. government-linked programs. Not commercially significant yet. [scenario est.]
Geographic concentration risk score for U.S. chips drops modestly. Long-term strategic diversification begins. [scenario est.]
Amkor Arizona becomes viable packaging alternative for defense + national security chip programs. [scenario est.]
U.S. defense programs gain domestic advanced packaging path. Strategic autonomy for classified AI programs. [scenario est.]