Why Selective De-risking Is Harder Than Decoupling

For AI and tech builders, founders, and investors, the real problem is not whether geopolitics affects operations. It is how to redesign a multinational footprint without turning suppliers, compute, data, and market access into a permanent drag on speed and scale.

That is why the decoupling-versus-de-risking distinction matters. Decoupling aims to cut dependence cleanly. De-risking tries to keep operating across borders while reducing exposure to coercion, chokepoints, and regulatory surprises. The catch is that de-risking rarely looks elegant in practice: it changes procurement, footprint design, supplier governance, and how much inefficiency leadership is willing to absorb.

Start with the policy intent, not the slogan

The EU’s position is explicit: “The EU does not intend to decouple from China. Our policy is one of de-risking, not decoupling.” ¹ For multinational leaders, that line is less about diplomacy than operating logic. It says the goal is not a clean break. It is selective vulnerability reduction.

SIPRI’s framing makes the same point more operationally: de-risking means continuing to “work and trade with China,” while mitigating threats to resilience, prosperity, and security. ² That is the tension most global firms now face too. You may want scale, market access, and interdependence. You do not get to assume those benefits come without strategic exposure.

A broader International Business Review analysis describes the shift from the “efficiency-driven globalization of 1980–2016” toward strategies emphasizing resilience and local responsiveness. ³ The business consequence is straightforward: the old optimization target was unit cost. The new one is operating continuity under constraint.

"AI compute is becoming a modern equivalent to military defense spending."

— 1 Minute Signal coverage of Julia McCoy ⁴

That is the same strategic shift in a different domain. Once resilience becomes a national or corporate priority, the operating model stops being a pure efficiency machine.

Why “China+1” is not a synonym for exit

A common mistake is to treat diversification as a softer version of decoupling. It is not that simple.

East Asian Institute coverage of China+1 is useful because it states the distinction plainly: “China+1 and de-risking do not imply a full decoupling from China. Firms and governments are not replacing China; they are simply responding to a world in which disruption is no longer exceptional.” ⁵ That is the real shift. China may remain core, but it is no longer enough to hold a global operating model together by itself.

Acclime China’s analysis highlights the hidden cost of superficial diversification: “For many manufacturers, offshoring to Southeast Asia has simply added a processing step in the middle of a supply chain that still begins in China, one that now carries its own tariff exposure and higher operational costs.” ⁶ In other words, a footprint change can create the appearance of resilience without actually removing concentration risk.

Novex Global’s warning lands from the operator’s side of the ledger: “The biggest risk isn’t staying in China too long — it’s diversifying without proper supplier vetting to hit a deadline.” ⁷ That is the execution trap. Reactive diversification can lower political exposure while raising quality failures, coordination overhead, and supplier risk.

The practical lesson is that de-risking is not a slogan. It is a design and sequencing problem.

The real tradeoff is resilience versus efficiency

The best supply-chain literature in the source set keeps the economics honest. KPMG’s analysis is blunt: “The Washington Consensus era of ever-increasing trade liberalization is over. Persistent trade shocks, both policy-driven and otherwise, continue to reshape global supply chains.” ⁸ It adds the line leaders should be asking internally: “The question is not whether to pay for resilience, but how much, and where is that investment best deployed.” ⁸

That is the right framing. Resilience is not free, and not every risk deserves the same budget.

KPMG also captures a dynamic that should make executives cautious about simple relocation narratives: “The deficit has shifted, not shrunk and what once came directly from Shenzhen now arrives via Hanoi, Monterrey and Jakarta. Decoupling has now become recoupling through intermediaries.” ⁸ For multinational planners, that means moving production does not remove dependency by default. It often just relocates it into a different network with fresh tariff, compliance, and supplier dependencies.

"Supply chains are networks, not chains, and risk is concentrated."

— GOV.UK ⁹

That is why broad, undifferentiated decoupling can be so expensive. It may reduce one category of exposure while introducing new costs in lead time, inventory, certification, and quality assurance. The price of resilience depends on where the bottleneck actually sits.

Multinationals usually choose between exit, freeze, or reconfiguration

The most useful academic lens in the source set comes from Journal of International Business Studies. It describes three broad responses to geopolitical tension: divestment, freezing operations, or staying while reconfiguring resources. ¹⁰

That framework maps cleanly to corporate decision-making:

• Divestment is the cleanest decoupling. It removes the host-country exposure entirely, but gives up future optionality.
• Freezing is a de-risking move. You stop expansion, but preserve the subsidiary and its resources if conditions improve.
• Staying and downgrading is the most complex route. It means isolating resources from the global network or limiting certain advantages to host-country-specific products. ¹⁰

The same article’s core warning is worth keeping in view: “Continued association with a stigmatized location leads to cross-border stigma translation risk that threatens the MNE’s network resources and non-location-bound firm-specific advantages (FSAs).” ¹⁰ That is the hidden cost of geopolitical entanglement. The risk is not only physical disruption; it is reputational spillover across the wider firm.

This is where the comparison becomes more concrete. In sectors with strong localization requirements or high regulatory friction, freezing or selective reconfiguration can preserve more value than an exit. In sectors where the asset base is highly mobile and the reputation risk is contagious, decoupling becomes easier to justify.

De-risking is targeted, not blanket redundancy

This is where the supply-chain resilience research matters most. GOV.UK’s Foresight report says effective resilience should focus on “targeted resilience at the point of failure, rather than spreading resources broadly.” ⁹ That is the key execution insight. Resilience is a resource allocation problem, not a moral preference.

The report’s broader framework uses three capabilities: absorptive capacity, adaptive capacity, and restorative capacity. ⁹ That combination is more useful than a generic “diversify everything” instruction. In practice, it means:

• buffers where failure would cascade,
• multiple suppliers where substitution is realistic,
• recovery playbooks where disruption is likely,
• and no unnecessary redundancy where the system already has slack.

That is also where sector and asset type matter. Some operations can absorb more OpEx-heavy monitoring and dual sourcing. Others require asset-heavy regionalization or vertical integration. The point is not to choose one universal posture. It is to match the response to the concentration of risk and the firmness of the bottleneck.

The contrast shows up clearly in the source set. KPMG notes that a fire at a U.S. magnesium supplier halted one automaker’s production, while Taiwan still produces 92% of sub-10nm semiconductors. ⁸ Those are not the same problem. One calls for a narrowly targeted supplier response. The other implies a deeper structural dependence that may justify regionalization, inventory redesign, or, in the most extreme cases, exit from a market or technology stack.

Governments are de-risking too, and that changes corporate options

One reason this debate matters now is that governments are no longer passive background conditions. They are increasingly shaping the operating environment directly.

The EU’s 2026 economic security framework expands screening for sensitive areas such as dual-use items, critical raw materials, and critical infrastructure, while also adding outbound investment monitoring for sectors like AI, semiconductors, and quantum technologies. ¹¹ The European Parliamentary Research Service describes the shift as “a paradigm shift from a reactive to a more coherent, joined-up and proactive deployment of EU tools.” ¹²

That matters for multinational operators because de-risking is no longer just an internal supply-chain decision. It is being constrained, and sometimes enabled, by screening rules, procurement policy, and investment oversight.

A concrete example from pharmaceuticals shows both the appeal and the limits of policy tools. CFR’s coverage of U.S. pharmaceutical procurement argues that federal purchasing power is still optimized mainly for price, even though it could be used to incentivize domestic manufacturing. ¹³ But the source is careful not to oversell the fix: using procurement to set industry-wide standards remains “an aspirational theory rather than a tested policy with documented outcomes.” ¹³

That caution belongs in any executive playbook. Procurement can support de-risking when stable demand helps a supplier base scale. It can also fail if the economics are too thin or the buyer does not change its incentives. For multinationals, the relevant lesson is narrower: public procurement can reinforce a resilient operating model, but it is not a substitute for one.

1 Minute Signal coverage of Julia McCoy’s discussion of sovereign compute adds the same lesson from the infrastructure side: countries are trying to reduce dependence on foreign cloud providers because sovereignty, industrial growth, and supply chain resilience are now strategic priorities. ⁴ For builders, that is a reminder that market access increasingly depends on where regulators and states decide dependence is acceptable.

A simple decision matrix for multinationals

If you need a practical rule, use this sequence:

1. Exit when exposure is existential.
   If the host market creates unacceptable legal, reputational, or security risk, divestment is the cleanest response. ¹⁰

2. Freeze when you need optionality.
   If the market still matters but current conditions are too uncertain, halt expansion while preserving assets and decision rights. ¹⁰

3. Reconfigure when the bottleneck is specific.
   If the problem is concentrated in one supplier, route, or node, use targeted resilience rather than broad retreat. ⁹

4. Diversify when substitution is real.
   China+1, friend-shoring, and regionalization only work if the new node actually reduces concentration risk instead of just adding a new processing step. ^{5, 6}

5. Don’t confuse motion with de-risking.
   Moving a factory, changing a country label, or adding a second sourcing line does not automatically reduce risk. It may simply redistribute it. ^{7, 8}

What builders should take away

For AI and tech leaders operating across borders, the headline is not “decouple or don’t.” It is that the hard work is in deciding what to preserve, what to isolate, and what to abandon.

The evidence points to a few disciplined conclusions. De-risking is usually better than blanket decoupling when the business still needs cross-border scale. But it works best when it is targeted, not broad-brush. The right response depends on how concentrated the bottleneck is, how location-bound the asset really is, and how exposed the firm is to geopolitical or regulatory spillovers. ^{9, 10, 14}

That is a more demanding operating model than the old globalization playbook. It is also the one the sources support.