While the whole world watches chips and accelerators, the real money in AI infrastructure is being made by a modest furniture hardware manufacturer. 85-year-old Lin Tsung-chi, founder of Taiwan's King Slide Works, unexpectedly topped the island's rich list. Since the start of the year, his company's shares have surged nearly 280% — and this is not a speculative rally, but a consequence of a fundamental shift in data center architecture.
King Slide has historically been known for hinges and drawer slides, but the key business today is rail mechanisms for server racks. These components hold equipment weighing hundreds of kilograms and allow it to be pulled out for maintenance without disrupting cooling. Based on industry data, my estimates suggest the company controls about 80% of the market for high-performance server rails.
87% Margin: More Than Nvidia and TSMC
Demand for King Slide's products is growing alongside data center construction and the transition to more powerful AI systems. When a server rack costs millions of dollars, the price of rails takes a back seat — quality and reliability matter. This is reflected in the numbers: the company's gross margin reached 87% in the latest quarter, compared to around 50% a few years ago. For comparison, Nvidia's figure is ~75%, and TSMC's is 68%.
King Slide's Executive Vice President Jay C. Wang attributes this profitability to two decades of engineering development that enabled the creation of specialized mechanisms for server equipment. And these are not just words: the company has become a monopolist in its niche, and Forbes estimates the founder's fortune at $20.3 billion. Lin has even surpassed Terry Gou of Foxconn, whose wealth has also grown on the AI wave.
New Architectures — New Rails
Further demand growth is almost guaranteed. Hyperscalers are transitioning to their own AI accelerators, increasing equipment density, and implementing liquid cooling. Each new chip differs in dimensions, heat dissipation, and connectors — which means it requires a unique rack and unique rails. Examples are already evident: Google is expanding its agreement with Marvell, and Alibaba is raising $10.2 billion for AI infrastructure.
The scale is impressive: the announced capacity of gas power plants for direct power supply to U.S. data centers has grown from 97 GW to more than 189 GW by mid-2026. Nvidia is scaling its Vera Rubin platform to gigawatt-level capacities. King Slide is expanding production in Houston, but competition is intensifying: according to Daiwa Securities, the company's share of supplies for Nvidia systems could drop to 75% next year.
My view: the King Slide story is a classic example of how "secondary" players in the AI chain capture disproportionately high margins. But betting on a niche monopoly is risky: as soon as major operators certify alternative suppliers, pricing pressure is inevitable. The question is not whether the margin will decline, but how quickly and how deeply.