The history of the global artificial intelligence market continues to spring surprises, and one of the most striking examples is the rise of Taiwan's King Slide Works. The company's founder, 85-year-old Lin Tsung-chi, unexpectedly topped the island's rich list. The reason is both mundane and brilliant: his firm, which started out making furniture hardware, has captured the lion's share of the market for critical components in the servers that power modern neural networks.
Since the start of the year, King Slide shares have surged nearly 280%. The secret to its success lies in rail mechanisms for server racks. These are not just metal guides but high-precision engineered products capable of holding equipment weighing hundreds of kilograms and allowing it to slide out for maintenance without disrupting delicate cooling systems. Based on industry data, I estimate the company controls around 80% of the global market for rails used in high-performance servers.
Margins higher than Nvidia's
King Slide's financial metrics look sensational. In the latest quarter, gross margin reached 87%, whereas just a few years ago it barely exceeded 50%. For comparison, Nvidia's figure is around 75%, and TSMC's is 68%. Such profitability stems from simple market logic: when a single AI server rack costs millions of dollars, clients are willing to pay for rail reliability rather than skimp on it. As analysts rightly note, quality here trumps price.
As a result, Lin's wealth, according to Forbes, is estimated at roughly $20.3 billion, allowing him to overtake Foxconn founder Terry Gou. Notably, the latter's fortune also grew thanks to AI infrastructure, but it was not enough.
New architectures — new opportunities
I see several fundamental factors that will sustain demand for King Slide's products in the coming years. First, hyperscalers are actively shifting to their own AI accelerators. Google has already secured an option to buy shares in Marvell Technology as part of an expanded chip development agreement, while Alibaba is raising $10.2 billion to advance its AI stack. Each such chip is unique in size and heat output, requiring custom rack designs and, consequently, new rails.
Second, data center density is growing. The declared capacity of gas-fired power plants being built in the U.S. for direct power supply to data centers has increased from 97 GW to over 189 GW in just six months. Nvidia, in turn, is scaling its Vera Rubin platform to gigawatt-level capacities. All of this demands a reworking of server structures, including mechanical components.
However, competition is intensifying. According to Daiwa Securities estimates, King Slide's share of shipments for Nvidia system kits could drop to 75% next year as the chipmaker diversifies its suppliers. This will allow data center operators to bargain over prices. Nevertheless, the company is expanding production in Houston and expects to remain directly tied to AI investment.
My takeaway: King Slide's story is a classic example of how the "second tier" of suppliers wins from a technological revolution. While everyone watches the chips, those providing the physical infrastructure for them are amassing fortunes. And this trend, by all appearances, is far from over.