While AI stocks took a beating in late June, one company barely flinched. Taiwan Semiconductor Manufacturing Company, the world’s most important chipmaker, saw its shares pull back roughly 7% from record highs near $476 to around $434 on July 2. Nvidia and AMD both faced steeper declines on June 26 as traders hedged positions and locked in profits across the AI sector. TSMC, the company that actually fabricates the chips powering the AI revolution, held comparatively firm.

Why TSMC keeps shrugging off the selloff

Every cutting-edge AI accelerator from the major chip designers runs through TSMC’s foundries, which gives the company a uniquely insulated position when sentiment sours on individual AI names.

CEO C.C. Wei said in April 2026 that AI demand remains “extremely robust.” The company backed it up with numbers. TSMC revised its full-year 2026 revenue growth expectations to exceed 30%. Capital expenditures were set between $52 billion and $56 billion. TSMC now projects AI-related revenue to grow at a 55% compound annual growth rate through 2029, up from previous guidance of 45%, driven by rising volumes, stronger pricing power, and the emergence of agentic AI technologies. Recent monthly sales data showed a 30% increase.

The tokenized TSMC play on Ethereum

TSMX, a tokenized version of TSMC shares, trades on Ethereum, offering on-chain exposure to the chipmaker’s performance without touching a traditional brokerage. TSMX prices have recently ranged between $440 and $480, roughly tracking the underlying equity. Earlier in 2026, the tokenized shares hit all-time highs exceeding $880. The divergence between TSMX’s all-time high above $880 and its current range around $440-$480 also raises questions about liquidity dynamics in tokenized markets versus traditional ones. Thin order books on-chain can amplify moves in both directions, making tokenized stocks simultaneously more accessible and more volatile than their underlying assets.

What this means for crypto investors

When TSMC raises its capex guidance to $52-$56 billion, it’s effectively greenlighting the next generation of chips that will run both centralized and decentralized AI workloads. The 55% CAGR projection for AI revenue through 2029 provides a useful benchmark for anyone trying to size the AI compute market.

For traders watching the broader AI rotation, TSMC’s resilience relative to Nvidia and AMD suggests the market is differentiating between companies that design AI chips and the one that makes them all. The foundry model, where TSMC collects revenue regardless of which designer wins market share, functions like a picks-and-shovels play during a gold rush.