Not All Supply Chain Risks Are Created Equal

In February 2022, when Russian tanks rolled into Ukraine, panic swept through the semiconductor industry. Ukraine and Russia together supplied somewhere between 50% and 70% of the world's semiconductor-grade neon gas, an obscure but essential ingredient in chipmaking. Prices spiked 600% almost overnight. Headlines screamed about another supply chain catastrophe. Policy experts nodded knowingly: this proved the fragility of globalization, the danger of geographic concentration, the urgent need for reshoring and industrial policy.

Then something unexpected happened. The crisis solved itself.

Within nine months, alternative suppliers in China, the United States, and Japan had ramped up production. Semiconductor manufacturers optimized their processes to use less neon. No major chip production was actually disrupted. The market, it turned out, worked exactly as economics textbooks said it should: high prices induced new supply, and equilibrium was restored.

This outcome should have prompted some serious reflection. Instead, it was mostly ignored. The policy machinery had already moved on, committed to a narrative in which any geographic concentration represents a critical vulnerability requiring government intervention. Billions of dollars in subsidies for "supply chain resilience" were already flowing. The CHIPS Act in the United States, the European Chips Act, similar programs in Japan and Korea, all premised on a simple idea: concentrated supply is fragile supply, and we need to spend whatever it takes to diversify.

But what if that premise is wrong? Or more precisely, what if it's only sometimes true? What if the Ukraine neon crisis and the Taiwan semiconductor concentration actually represent two fundamentally different types of problems, requiring radically different responses?

The Tale of Two Bottlenecks

Consider two scenarios, both involving extreme geographic concentration in semiconductor supply chains:

Scenario One: Ukraine Neon Gas

• Concentration: 50-70% of global supply from a single geopolitical region
• Disruption: Sudden, complete shutdown due to war
• Price response: 600% spike
• Resolution time: Nine months
• Government intervention required: None
• Lasting impact: Minimal

Scenario Two: Taiwan Advanced Chips

• Concentration: Over 90% of cutting-edge chip production from a single geopolitical region
• Disruption: Hypothetical but plausible conflict scenario
• Potential impact: Catastrophic, multi-year disruption to global technology supply
• Government response: TSMC building $40 billion facility in Arizona despite 50% cost premium
• Resolution time if Taiwan disrupted: 5-10+ years minimum

On the surface, these look similar. Both involve dangerous geographic concentration. Both face geopolitical risks. Both triggered alarm among policymakers. Yet they are fundamentally different problems requiring completely different solutions.

The difference isn't about geopolitics or geography. It's about the nature of what's being produced.

What Makes Something Actually Irreplaceable?

Neon gas purification, it turns out, is not particularly difficult. The technology has been around for decades. The equipment is standard industrial machinery. The knowledge required is available in chemical engineering textbooks. Multiple companies around the world had the technical capability to produce semiconductor-grade neon; they just weren't doing it at scale because Ukrainian suppliers, leveraging cheap Soviet-era industrial infrastructure, had lower costs.

When war shut down Ukrainian production and prices spiked 600%, the economic signal was clear and the response was swift. Chinese producers expanded capacity. American industrial gas companies dusted off mothballed facilities. Japanese firms optimized existing production. Within months, the global supply had rebalanced. The "irreplaceable" Ukrainian supply turned out to be quite replaceable after all.

Now consider what TSMC does in Taiwan. The company manufactures the most advanced semiconductors in the world using extreme ultraviolet lithography machines that cost $150 million each and are themselves produced by only one company (ASML in the Netherlands). The manufacturing process involves hundreds of steps with tolerances measured in nanometers, where a speck of dust can ruin millions of dollars in chips. The knowledge required isn't in textbooks because much of it is tacit, embedded in the routines and relationships of teams who have worked together for years, accumulated through decades of trial and error, optimization, and learning.

This knowledge cannot be downloaded, purchased, or reverse-engineered in any reasonable timeframe. When TSMC's Arizona facility comes online, it will have been five years in development, will cost 50% more to operate than equivalent facilities in Taiwan, and will initially produce chips one generation behind TSMC's most advanced Taiwan fabs, even though it's built by the same company using nominally the same equipment.

If Taiwan were disrupted tomorrow, the rest of the world could not simply ramp up alternative production the way they did with neon gas. The knowledge, the supplier ecosystems, the accumulated learning and optimization, are not easily transferable. Samsung and Intel have been trying to catch TSMC's most advanced capabilities for years, spending billions, and remain persistently behind despite having access to the same equipment suppliers and similar pools of engineering talent.

The Insurance That Looks Like Waste

This brings us to TSMC's Arizona investment, which from a narrow financial perspective looks almost irrational. Building a cutting-edge fab in Arizona costs approximately 50% more than building it in Taiwan. Labor costs are higher. Supplier ecosystems are less developed. The highly trained workforce has to be built from scratch. If you ran a standard return-on-investment calculation, Arizona would never make the cut.

Morris Chang, TSMC's founder, is not known for making irrational investments. So why do it?

The answer is that the Arizona fab isn't primarily a production facility. It's an insurance policy. More precisely, it's a financial option that pays off in a specific, catastrophic scenario: disruption to Taiwan. In the base case where Taiwan remains stable and accessible, Arizona will always be less efficient than Taiwan operations. The 50% cost premium is the price of the insurance policy, paid year after year. But in the tail-risk scenario where Taiwan becomes inaccessible, Arizona ensures that TSMC's most important customers (Apple, Nvidia, AMD) can still get advanced chips, and TSMC can remain a viable company.

Traditional corporate finance would reject this investment. The expected value, calculated by multiplying probabilities times outcomes and summing across scenarios, likely comes out negative. The base case where Taiwan is fine (high probability) involves paying the 50% cost premium forever (negative value). The disaster case where Taiwan is disrupted (lower probability) provides value (preservation of customer relationships and company viability), but when you multiply a modest probability times the benefit and subtract the ongoing cost premium, the calculation doesn't favor building Arizona.

Yet the investment makes perfect strategic sense when you account for the asymmetric nature of the risk. The downside of not having Arizona if Taiwan is disrupted is catastrophic: TSMC loses its most valuable customer relationships permanently, possibly faces existential threat as a company. The cost of having Arizona when Taiwan is fine is manageable: a margin reduction that TSMC can absorb given its profitability. This is the logic of insurance, not of expected-value maximization.

Now, crucially, this insurance logic only makes sense because of the specific nature of what TSMC produces. The underlying capabilities are genuinely difficult to replicate. There's no latent global capacity waiting to be activated by price signals. If Taiwan were disrupted, market mechanisms would not solve the problem within any reasonable timeframe because the barrier to entry isn't capital or willpower, it's accumulated knowledge and capabilities that require years to develop.

The Ukraine neon crisis proved that for some inputs, market mechanisms work fine. The TSMC situation demonstrates that for others, they don't.

How to Tell the Difference

So how do we distinguish between supply chain vulnerabilities that require insurance investments and those that markets can handle? The critical question is about substitutability, but at a deeper level than usually considered.

Ask these questions about any concentrated supply:

Can the technology be replicated with standard equipment and documented knowledge? If yes, price signals can induce alternative supply. Ukraine neon: yes, standard industrial gas processing. TSMC advanced chips: no, requires tacit knowledge accumulated over decades.

Does latent capacity exist globally that could be activated quickly? If yes, market mechanisms will resolve disruptions. Ukraine neon: yes, multiple countries had companies with relevant capabilities. TSMC advanced chips: no, only Samsung is remotely close and they're years behind at the cutting edge.

If prices increased 10x, would sufficient new supply emerge within 12-24 months? If yes, the bottleneck is temporary and market-solvable. Ukraine neon: proven empirically, 600% price spike yielded nine-month resolution. TSMC advanced chips: definitively no, even infinite capital cannot compress decades of learning.

Is the knowledge required to produce the input mostly explicit and transferable, or tacit and embedded in organizational routines? If explicit, new entrants can acquire it. If tacit, entry barriers are formidable. Ukraine neon: explicit, documented in chemical engineering literature. TSMC advanced chips: extensively tacit, embedded in team practices and institutional knowledge.

These aren't merely technical distinctions. They determine whether a supply chain concentration represents a structural vulnerability requiring insurance or a temporary geographic imbalance that markets will naturally correct.

The Resilience Theater Trap

Here's where current policy goes wrong. In the wake of COVID-19 supply chain disruptions and rising geopolitical tensions, a consensus has formed: geographic concentration is dangerous, and we need to reshore critical industries. This consensus doesn't distinguish between structural and geopolitical bottlenecks. It treats all concentration as equally problematic and all reshoring as equally valuable.

The result is what might be called "resilience theater": spending large sums of money to create domestic capacity for products that, if disrupted, markets would replace naturally within acceptable timeframes. It's security theater for supply chains, creating the appearance of action without meaningfully reducing actual vulnerability.

Consider the CHIPS Act in the United States, which allocated over $50 billion in subsidies for semiconductor manufacturing. Some of these investments target genuinely difficult-to-replicate capabilities, like TSMC's Arizona fab for cutting-edge logic chips or Intel's efforts to rebuild advanced manufacturing capability. These might well prove strategic, worth the substantial cost because they address structural bottlenecks that markets cannot easily solve.

But the CHIPS Act also funds many other projects: established semiconductor manufacturers expanding capacity they likely would have expanded anyway, investments in mature technology nodes that multiple companies globally can produce, subsidies for equipment and materials that have functioning global markets. For these categories, it's far from clear that government subsidies add strategic value beyond what market mechanisms would provide naturally.

If another Ukraine-style disruption hits a non-strategic input, the 600% price spike will induce exactly the same market response whether or not that input received reshoring subsidies. The subsidy will have effectively paid for capacity that would have emerged anyway, just slightly earlier and at greater cost. That's not strategy; it's just expensive.

The Cost of Getting It Wrong

The stakes here are substantial, and the costs of error are asymmetric.

Underinvesting in structural bottlenecks: If we fail to develop insurance options for genuinely irreplaceable capabilities, a disruption could be catastrophic. Imagine a Taiwan conflict that shut down TSMC with no alternatives available. The global technology industry would face immediate, severe constraints. Smartphone production, data center buildouts, artificial intelligence development, all would face multi-year delays. The economic cost would run into the trillions. This is the nightmare scenario that keeps national security planners awake at night, and it's not paranoia; it's a real possibility.

Overinvesting in geopolitical bottlenecks: If we spend billions subsidizing domestic production of inputs that markets would naturally provide alternatives for, we waste resources that could have been deployed more productively. The opportunity cost isn't just the direct subsidy; it's all the other things that money could have funded: basic research, infrastructure, education, debt reduction. Worse, these subsidies distort markets, prop up inefficient producers, and create political constituencies that fight to maintain protection long after any strategic rationale has passed.

Current policy seems far more worried about underinvestment than overinvestment. The operating assumption appears to be: when in doubt, subsidize domestic production. But this is wrong. Overinvestment in resilience theater directly enables underinvestment in genuine strategic needs by consuming limited political capital and fiscal resources.

If we spend $10 billion subsidizing semiconductor equipment that has functioning global markets, that's $10 billion not available for the much harder problem of developing domestic capacity in genuinely strategic materials or advanced packaging technologies. If we spread CHIPS Act funding across every plausible semiconductor-adjacent investment, we dilute the focus from structural bottlenecks that actually require government support to geopolitical bottlenecks where market mechanisms work fine.

What Proper Discrimination Looks Like

Some organizations have demonstrated the ability to correctly distinguish between these scenarios. ASML, the Dutch company that makes extreme ultraviolet lithography machines, reduced its dependence on Ukraine and Russia for neon gas to below 20% before the 2022 invasion, not through massive investments in captive neon production, but through diversification of suppliers and optimization of consumption. This was appropriate for a geopolitical bottleneck: acknowledging the risk while recognizing that market mechanisms would provide alternatives if needed.

TSMC, meanwhile, is paying a 50% cost premium to build redundant capacity in Arizona for something that absolutely cannot be quickly replicated: cutting-edge chip production. This is appropriate for a structural bottleneck: accepting the ongoing cost of insurance because the downside scenario cannot be addressed through market mechanisms.

The difference is in the analysis that preceded the decision. ASML apparently recognized that neon purification technology, while concentrated in Ukraine, was not fundamentally irreplaceable. Multiple companies globally had the capability; the question was just one of price. TSMC recognized that its most advanced manufacturing capability is genuinely difficult to replicate and requires years of development even when you're replicating your own processes in a new location.

This is the kind of discrimination that effective strategy requires. Not all risks are equal. Not all concentrated supply chains need the same response. Sometimes the insurance premium is worth paying. Sometimes it's wasted money that would be better spent elsewhere.

A Better Framework for Resilience

Instead of the current blanket approach of subsidizing reshoring wherever there's geographic concentration, policy should be more discriminating:

For structural bottlenecks (genuinely difficult-to-replicate capabilities with no latent global capacity): Insurance investments may be justified despite negative expected returns in the base case. The TSMC Arizona fab model is the template: accept ongoing cost premiums to maintain redundant capacity that preserves strategic optionality in tail-risk scenarios. Focus government support on these genuinely irreplaceable capabilities where market mechanisms demonstrably fail.

For geopolitical bottlenecks (concentrated supply of inputs with replicable technology and latent global capacity): Light-touch policies that reduce exposure without expensive domestic capacity buildouts. Subsidize diversification of suppliers, stockpiling for temporary disruptions, optimization to reduce consumption. But don't pay for full domestic production that will never be cost-competitive and isn't actually necessary because markets will solve disruptions naturally. The Ukraine neon crisis proved this works.

For everything else (unconcentrated supply or low-consequence inputs): Let markets work. Not every input needs a resilience strategy. Some things are concentrated because that's efficient and the risks are manageable. Some things can be easily substituted. Some disruptions have minimal consequences. Trying to eliminate all supply chain risk is a fool's errand that wastes resources and creates brittleness through overoptimization for the wrong threats.

The test should always be: If this supply were disrupted and prices increased 10x, would market mechanisms produce alternatives within an acceptable timeframe? If yes, we're looking at a geopolitical bottleneck where market forces will resolve the problem. If no, we're looking at a structural bottleneck that might justify insurance investments.

The Ukraine Lesson

The Ukraine neon crisis should be remembered not as proof that supply chains are fragile, but as proof that markets work when given the chance. A 600% price spike in a critical input, a 50-70% supply disruption from a war zone, and the global semiconductor industry kept running. No subsidies required. No emergency government intervention. Just price signals doing what price signals do: inducing new supply and optimization of consumption.

This is not an argument against all resilience planning or all government support for strategic industries. TSMC's Arizona fab, even at a 50% cost premium, makes perfect sense given the genuine irreplaceability of advanced semiconductor manufacturing. There are real structural bottlenecks that markets cannot quickly solve and where insurance investments are rational despite negative base-case returns.

But we need to distinguish between situations like TSMC (where insurance is rational) and situations like Ukraine neon (where markets work fine). Conflating them leads to massive waste: spending billions subsidizing domestic production of things that markets would naturally provide alternatives for, while potentially underinvesting in genuine strategic bottlenecks because we've diluted our resources across too many priorities.

The current policy consensus doesn't make this distinction. It treats all geographic concentration as dangerous and all reshoring as virtuous. That's not strategy. That's panic, dressed up in the language of resilience and national security.

Real strategy requires harder thinking: asking which concentrated supplies are genuinely irreplaceable, which disruptions markets can handle, where insurance premiums are worth paying, and where we're just engaging in expensive theater. The Ukraine neon crisis gave us a clear test case. The market solved a 50-70% supply disruption in nine months, with no subsidies and no industrial policy. That should inform how we think about supply chain resilience going forward.

Not all risks are equal. Not all concentration is dangerous. And sometimes, the best policy is to let markets work.