Key takeaways
Where value chains quietly concentrate power
ASML, NVIDIA, and TSMC (Taiwan Semiconductor Manufacturing Company) dominate their respective markets. That much is consensus. What receives less attention is why these positions persist, and what separates them from companies that once looked equally dominant but eventually lost ground.
The distinction is structural irreplaceability. Market leaders can be displaced by better products or pricing pressure. Structurally irreplaceable positions persist because the systems they serve have no practical alternative and replication would require decades rather than capital.
This article examines what makes a technology position truly irreplaceable and offers a framework for distinguishing structural durability from competitive advantages that erode over time.
Durability as the scarce asset
Most investment analysis focuses on cyclical factors: interest rate sensitivity, demand elasticity, competitive positioning within a product cycle. These do matter, but they operate on a different timescale than structural factors—the forces that determine whether a position survives a crisis rather than merely recovers from one.
Structural risk differs from cyclical risk in kind, not degree. A cyclically dominant company loses share when demand shifts or a competitor launches a better product. A structurally dominant company maintains its position because the system itself depends on its continued operation.
The four characteristics of structural durability
Positions that persist across cycles, governments, and crises share four traits:
Non-replicability over non-competitiveness. The barrier is not that competitors lack resources. The barrier is that replication requires capabilities that cannot be assembled on any reasonable timeline.
Ecosystem dependency over customer loyalty. Users do not choose the product because of preference alone. They use it because their workflows, their partners, and their infrastructure assume it exists.
Time as a barrier, not capital. Sufficient funding cannot compress decades of accumulated knowledge, failed experiments, and supplier relationships into a shorter development cycle.
Strategic relevance beyond market demand. The position matters to governments, supply chains, and critical industries, not just to customers willing to pay a premium.
When all four characteristics align, the result is a position that looks less like market leadership and more like infrastructure.
ASML and industrial dependency
ASML holds a 100% monopoly on EUV lithography—the technology required to manufacture chips at the most advanced nodes. All sub-5nm logic chips from TSMC, Samsung, and Intel require EUV lithography from ASML.
This is not a competitive advantage in the conventional sense. ASML's position emerged from 30 years of R&D, a 5,000+ supplier ecosystem spanning multiple continents, and engineering complexity that makes replication impractical regardless of available capital.
Why competitors cannot catch up
China has invested tens of billions attempting to develop domestic EUV capability. Current estimates place Chinese efforts at least a decade behind, and the gap may be widening as ASML continues to advance.
The constraint is not funding. The constraint is that EUV development required solving thousands of interdependent technical problems over 30 years. Each solution is built on previous solutions, every supplier relationship developed through years of shared problem-solving.
For investors, the implication is direct: ASML's position represents industrial dependency, not market dominance. TSMC, Samsung, and Intel do not choose ASML because it offers the best value. They use ASML because no alternative exists, and none is likely to exist within any reasonable investment horizon.
NVIDIA and ecosystem lock-in
NVIDIA commands 80-92% of the AI accelerator market, with data center revenue reaching $115 billion in fiscal 2025. These numbers reflect hardware sales. The deeper moat lies elsewhere.
CUDA, NVIDIA's proprietary software platform, has accumulated 4 million+ developers over 20 years of continuous investment. Every major AI framework (PyTorch, TensorFlow, JAX) is optimized for CUDA. From academic research to production deployment, CUDA is the assumed foundation. Training pipelines are built on it. So is everything downstream.
The cost of switching
AMD offers competitive hardware with its MI300 series. Intel has invested billions in its Gaudi accelerators. Both face the same problem: customers switching from NVIDIA must retrain engineers, rewrite code, rebuild workflows, and accept performance uncertainty during the transition.
These switching costs compound at scale. A startup might experiment with alternatives. Most hyperscalers running inference at scale have found the transition cost prohibitive. Internal AWS data showed Trainium at just 0.5% of NVIDIA GPU usage as of early 2024, despite years of development. The result is that NVIDIA's pricing power persists even as hardware alternatives emerge, with gross margins holding above 73% through Q3 fiscal 2026.
The investment logic follows directly: NVIDIA's advantage is not that it makes better chips (though it often does) but that the cost of displacement extends far beyond the hardware purchase price.
TSMC and the fragility of concentration
TSMC's dominance presents a different pattern. The company holds 71% of global foundry market share as of Q3 2025, with quarterly revenue exceeding $33 billion. Nearly three-quarters of its sales come from 7nm nodes and below—the most advanced manufacturing processes available.
Apple, AMD, NVIDIA, and Qualcomm all depend on TSMC for their highest-performance chips. No alternative supplier can match TSMC's combination of process technology, manufacturing scale, and yield rates.
Durability and exposure
TSMC's position illustrates a tension inherent to structural concentration: the same factors that make an asset irreplaceable also make it systemically sensitive.
Taiwan's geographic location creates geopolitical exposure that cannot be diversified away. TSMC is building fabs in Arizona and Japan, but replicating Taiwan's manufacturing ecosystem (the suppliers, the talent pool, the operational knowledge) takes years even with unlimited capital.
For fund managers, TSMC's situation requires scenario thinking rather than standard valuation models. TSMC's technical lead is not in doubt. Pricing the tail risk of disruption to an asset thatglobal technology production cannot function without is the harder problem.
Regulation as confirmation
The European Commission designated seven gatekeepers under the Digital Markets Act: Alphabet, Amazon, Apple, ByteDance, Meta, Microsoft, and Booking. These companies now face obligations around data portability, interoperability, and self-preferencing.
Fines have followed. Apple paid €500 million for App Store restrictions, while Meta paid €200 million for its advertising consent model. These penalties represent a new phase of platform enforcement, but they do not represent dismantlement.
Why designation matters
Gatekeeper status confirms that a position has become systemically important. Regulators intervene precisely because the position cannot be contested through normal market mechanisms.
The pattern holds historically. Antitrust actions against Microsoft in the 1990s imposed behavioral remedies while leaving the company's core position intact. Investigations into Google's search practices have continued for over a decade without fundamentally altering its market share.
In technology markets, regulation has more often followed structural dominance than prevented it. Platform gatekeepers face behavioral constraints, not structural dismantlement, though historical exceptions (AT&T in 1984, Standard Oil in 1911) involved physical infrastructure that could be divided along geographic or operational lines—a model that maps poorly onto software ecosystems built around data and network effects.
By the time a company attracts regulatory attention for dominance, its structural position is typically secure. The obligations that follow constrain behavior at the margins without unwinding the underlying moat.
Systemic concentration beyond companies
China's technology research ecosystem illustrates a different form of structural concentration. They now lead in 66 of 74 critical technologies tracked.
Four areas carry high monopoly risk: cloud and edge computing, computer vision, generative AI, and grid integration technologies. This concentration operates at the national level, beyond the reach of company-specific regulation or competitive pressure.
The capital implication: some forms of structural risk cannot be mitigated through portfolio construction. They can only be understood and priced. For fund managers, the response is not avoidance but explicit scenario modeling and transparent LP communication about tail risk.
Four questions for portfolio review
The number of truly irreplaceable positions is smaller than commonly assumed. Capital crowds into perceived safety (e.g. large market caps, familiar names, consensus "quality") without always distinguishing between cyclical dominance and structural durability. This concentration "poses crowding and concentration risks that leaders should manage explicitly."
For allocators, the stress test matters more than the label:
What would displacement actually require? Not whether competitors exist, but whether they could realistically assemble the supplier relationships, engineering knowledge, and ecosystem integration within a fund's hold period. If the answer involves a generation of accumulated problem-solving, the position may be structural.
Where does the switching cost live? Hardware advantages erode. Ecosystem lock-in compounds. Identify whether customers face friction at the product level (solvable) or at the workflow, training, and infrastructure level (sticky).
How would you model disruption in IC? Standard DCF underweights tail risk for concentrated positions. Can you articulate a credible disruption scenario, assign it probability, and defend the resulting valuation adjustment? If the scenario feels implausible, that itself is data about structural durability.
Does your LP narrative distinguish "leading" from "irreplaceable"? Allocators have watched market leaders lose ground within a single fund cycle. "Non-replicable" describes a different thesis than "dominant." The language you use signals whether you've stress-tested the position or simply assumed persistence.
Bottom line
The companies examined here are market leaders that function as infrastructure. Their positions persist because the systems they serve have been built around them, and rebuilding those systems would take longer than any reasonable investment horizon.
For fund managers, the distinction between "market-leading" and "non-replicable" carries weight with allocators who have watched supposed moats erode in a single product cycle. Articulating structural durability with precision and the right framing is itself a competitive advantage.
Collateral Partners works with fund managers to translate complex investment theses into clear, compelling narratives. If your portfolio includes structural positions that deserve sharper articulation, get in touch.
