Three cases document a structural triangle. UC-103 (The Silicon Moat) shows that TSMC controls 70% of global foundry and virtually all advanced AI chip packaging, concentrated on one island 130km from mainland China. UC-104 (The Foundry Gambit) shows that the Western reshoring attempt — Intel 18A, $7.86 billion in CHIPS Act funding, 9.9% government equity stake — has produced a technical achievement but not a commercial customer. UC-105 (Silicon Sovereignty) shows that US export controls have accelerated China’s domestic semiconductor ecosystem: SMIC producing 5nm-class chips via DUV multipatterning, Huawei targeting 600,000 Ascend 910C units in 2026, Nvidia’s China share halved from 90% to 50%. The prognostic question is whether these three forces converge into a single outcome: the permanent bifurcation of the global semiconductor supply chain into two incompatible ecosystems. A Western stack built on TSMC, running CUDA, designed with Synopsys/Cadence EDA tools, and packaged with CoWoS. A Chinese stack built on SMIC, running CANN, designed with domestic EDA, and assembled with domestic packaging. If the bifurcation becomes permanent, every AI model, every data centre, every chip-dependent product will exist in one ecosystem or the other. The window for managed convergence is open. These five WATCH triggers measure how fast it is closing.
This prognostic case is the capstone of a four-case semiconductor cluster and connects to four additional cases across the library. Each case contributes a structural thesis that the bifurcation window measures.
The eight cases span five sectors (semiconductor, AI infrastructure, automotive, geopolitical, macro) and four case types (diagnostic, amplifying, at-risk, prognostic). The convergence point is the physical layer: every case depends on chips, and the chip supply chain is splitting. The bifurcation is the meta-pattern that connects the bottom layer to everything above it.
These metrics establish the state of the semiconductor bifurcation as of March 22, 2026. Each will be checked at review (September 22, 2027).
The bifurcation is not a prediction. It is a measurement of forces already in motion. UC-103 through UC-105 document the three structural forces — concentration, reshoring failure, and adversary self-sufficiency — that are converging toward ecosystem separation. This prognostic case measures how fast the convergence is occurring and identifies the inflection points that would make the bifurcation permanent.
Hardware ecosystems can reconverge — you can switch fabs, you can change chip suppliers. Software ecosystems cannot easily reconverge once developers have committed. Huawei open-sourced CANN in 2025. Chinese developers are optimising AI workloads for Ascend architectures. Once training pipelines, inference stacks, and enterprise software are built on the domestic stack, the switching cost to return to CUDA becomes prohibitive. The hardware bifurcation is concerning. The software bifurcation would be permanent.
The bifurcation accelerates through three failure modes: (1) TSMC disruption forces the Western stack to confront its concentration risk with no alternative, while China’s parallel stack is unaffected; (2) Intel fails to secure 14A customers and the Western reshoring thesis collapses, increasing TSMC dependency; (3) SMIC yields improve and domestic HBM scales, removing China’s remaining dependency on foreign components. Any one accelerates bifurcation. All three together make it irreversible within the 18-month window.
Bifurcation is not inevitable. Trade de-escalation could reopen chip commerce. Intel 14A success could create a genuine TSMC alternative, reducing the geopolitical pressure that drives the split. SMIC yield problems could stall China’s sovereign stack. HBM domestic production could fail to scale. The TRADE_RECONVERGENCE trigger is included specifically to measure the positive scenario. But the structural forces currently favour bifurcation, and each month of separate ecosystem development increases the switching cost that makes reconvergence harder.
The semiconductor cluster is the physical layer underneath 62% of the case library. UC-065 (AI infrastructure capex) flows through TSMC. UC-043 (hyperscaler debt) finances chip purchases. UC-084 (AI code generation) runs on GPUs manufactured through the supply chain documented here. UC-092 (Europe’s automotive decline) depends on chip supply. The bifurcation is not an isolated geopolitical story. It is the substrate risk that connects to every technology, finance, and infrastructure case in the library. If the chip supply chain splits, the cascades are not confined to semiconductors. They propagate through every dimension of every connected case.
-- The Bifurcation: Prognostic Semiconductor Ecosystem Split
-- Semiconductor Cluster Case 4 of 4 (UC-103, UC-104, UC-105, UC-106)
FORAGE semiconductor_bifurcation
WHERE tsmc_foundry_share > 0.65
AND intel_external_revenue < 100_000_000
AND smic_5nm_production = true
AND huawei_ascend_target > 500_000
AND nvidia_china_share_decline > 0.40
AND domestic_software_stack_emerging = true
AND export_controls_tightening = true
AND western_reshoring_uncommitted = true
ACROSS D4, D6, D3, D1, D2, D5
DEPTH 3
SURFACE bifurcation
WATCH tsmc_disruption WHEN material_taiwan_ops_disruption AND delivery_delay_ge_4_weeks = true
WATCH intel_14a_anchor WHEN tier1_customer_binding_volume_commitment = true
WATCH smic_5nm_yield WHEN n_plus_3_yield_ge_50_pct AND commercial_product_shipped = true
WATCH asml_service_change WHEN duv_servicing_restricted_or_ceased = true
WATCH trade_reconvergence WHEN bilateral_agreement OR entity_list_relaxation OR nvidia_sales_restored = true
DRIFT bifurcation
METHODOLOGY 85 -- $100B+ in CHIPS/EU/Japan subsidies, comprehensive export controls, Intel 18A technical achievement, multilateral coordination
PERFORMANCE 35 -- TSMC share growing, Intel no customer, SMIC advancing, equipment still flowing, software stacks diverging, bifurcation forces accelerating
FETCH bifurcation
THRESHOLD 1000
ON EXECUTE CHIRP prognostic "Two semiconductor ecosystems forming. Western: TSMC 70% + CUDA + Synopsys/Cadence + CoWoS. Chinese: SMIC 5nm DUV + CANN + domestic EDA + domestic packaging. 8 cross-case references across 5 sectors. 5 WATCH triggers: TSMC disruption, Intel 14A anchor, SMIC yield milestone, ASML service change, trade reconvergence. The software lock-in is the irreversibility mechanism. The window is open. 18-month review measures how fast it is closing."
SURFACE analysis AS json
SURFACE review ON "2027-09-22"
Runtime: @stratiqx/cal-runtime · Spec: cal.semanticintent.dev · DOI: 10.5281/zenodo.18905193
If something you are dealing with looks like this case, write.