The MoatWide moat

TSMC is the company that makes the modern world's chips, and its moat is one of the widest in technology — not quite the pure monopoly of ASML upstream of it, but a dominance so complete at the leading edge that it functions like one. TSMC pioneered and perfected the 'pure-play foundry' model¹: it designs no chips of its own and sells no products under its own name, but instead manufactures, to order, the chips that other companies design — Apple's processors, Nvidia's AI accelerators, AMD's CPUs, and the silicon inside a vast swath of everything electronic. It is the contract manufacturer to the entire fabless industry, and at the most advanced nodes, where the hardest and most valuable chips are made, it is very nearly the only game in town: TSMC takes more than 70% of the whole foundry market's revenue, 72.5% in the June 2026 quarter², and upwards of 90% of leading-edge production. If you want a cutting-edge chip built, in practice you build it at TSMC.

The deepest layer is process-technology leadership. Making chips at the leading edge — today's 3-nanometer and 2-nanometer nodes — is among the hardest things humanity does, and TSMC does it better, earlier, and at higher yield than anyone else on earth. It has out-executed Samsung and left Intel, once the industry's manufacturing king, struggling to keep its own foundry ambitions alive. This lead is not static; TSMC extends it with every generation, pouring tens of billions a year into R&D and new fabs to stay a step ahead, and the gap to its rivals has, if anything, widened. Being the first to a working, high-yield new node means the most demanding customers — Apple, Nvidia — come to TSMC first, which funds the next node, which keeps them coming.

Beneath it lie scale, yield, and capital intensity, which together form a barrier few can even contemplate crossing. A single leading-edge fab costs on the order of twenty billion dollars, and TSMC's annual capital spending has climbed toward sixty billion and beyond. Only a company with TSMC's scale and cash flow can sustain that investment, and only decades of accumulated manufacturing experience produce the yields — the fraction of good chips per wafer — that make the economics work. A new entrant would need not just the money but the learning curve, and the learning curve cannot be bought.

Then comes the pure-play model itself, and the trust and ecosystem it creates. Because TSMC makes no chips of its own, it never competes with its customers — a crucial distinction from Samsung (which sells phones and memory) and Intel (which sells its own processors). A fabless designer can hand TSMC its most precious intellectual property knowing TSMC will not use it to build a rival product. Around that trust, TSMC has built a vast ecosystem of design tools, intellectual-property blocks, and advanced-packaging technologies (its CoWoS packaging is now essential to AI accelerators) that make designing for TSMC easier and switching away from it harder.

And around the whole thing wraps customer lock-in. Designing a leading-edge chip is a multi-year, deeply collaborative effort in which the designer and TSMC co-develop the chip against TSMC's specific process; the design is, in effect, built for TSMC's fab. Moving it elsewhere would mean re-engineering it for a different, inferior process, at enormous cost, risk, and delay — so customers stay, and the relationships deepen with every generation. The AI boom has made these bonds tighter still: Nvidia and the hyperscalers depend utterly on TSMC to build the accelerators the whole world is racing to buy.

The economics that result are extraordinary for a manufacturer: gross margins approaching 60% (and above it in the AI-fueled recent quarters), net margins near 50%, and a torrent of cash — even after the enormous capital spending. In 2025 revenue grew 36% to over $122 billion and net income leapt past $55 billion, and 2026 is guided higher still, above 40% growth, on insatiable AI demand.

The one risk the moat cannot touch is geography. The overwhelming majority of the world's most advanced chips are manufactured on the island of Taiwan, in the shadow of an unresolved and dangerous conflict with China. This concentration is the single largest risk in the whole business — a 'silicon shield' that protects Taiwan and simultaneously makes TSMC a chokepoint the great powers fret over — and it is why the stock trades at a visible discount to peers of similar quality. TSMC is diversifying, at vast expense, into Arizona, Japan, and Germany. But an investor in TSMC owns the best manufacturing franchise in the world's most important industry, priced with a permanent asterisk for the one risk no moat can wall out: the map. Inside what the moat can control, the number to watch is leading-edge share: while 90%+ of the world's most advanced chips come off TSMC's lines at yields no rival approaches, the discount is geography, not business. The year a Samsung or Intel node genuinely closes the gap — taking a flagship customer with it — is the year the asterisk stops being the only problem.