I've been paying close attention to the trends in AI server concept stocks recently and found that this sector is really worth in-depth research. As global investments in AI infrastructure continue to expand, the server market is expected to maintain double-digit growth through 2026, but the question is— which companies are truly worth buying?

I’ve organized the industry chain and found that the server industry roughly divides into three parts: complete machine assembly, infrastructure, and key components. Among them, I am most attracted to companies that have both technological advantages and deep ties with major clients like NVIDIA or Google.

First, let's talk about complete machine assembly. Foxconn, as the world's largest contract manufacturer, holds over 40% global market share in AI servers. In the first three quarters of 2025, its consolidated revenue reached NT$5.5 trillion, up 16% annually. Management estimates that AI business revenue will reach a trillion yuan scale by 2026, which is quite astonishing. Quanta is also strong in cloud servers, mainly serving Google and AWS. In Q3 2025, revenue surpassed NT$500 billion, with a net profit of NT$15 billion for the quarter. Wistron is more focused, with 100% of its business in data centers. Last year, revenue hit NT$950.6 billion, up 163% annually, with EPS reaching NT$275.06, setting a profit record for Taiwanese stocks.

In terms of infrastructure, Vertiv’s liquid cooling technology is key. Once AI racks exceed 100kW power consumption, their cooling distribution units have basically become standard. In Q3 2025, net sales reached $2.68B, up 29% year-over-year, with order backlog reaching $9.5 billion, providing high revenue visibility for the future.

Honestly, the current AI server concept stocks have already surged significantly, and many leading stocks still have high P/E ratios. I looked at Wall Street analyst forecasts—Celestica’s average target price is $374.50 (potential upside of 22%), Vertiv’s target price is $206.07 (potential upside of 27%), and Foxconn’s target price is NT$306 (potential upside of 30%). The numbers still have room to grow, but risks are also considerable.

What’s most noteworthy is that investors’ focus is shifting from revenue growth to profitability and ROI. If signs of an AI bubble burst appear, or if the market shifts from growth validation to profit validation, these stocks could experience sharp declines. Additionally, attention should be paid to cloud service providers’ investment scales, progress in non-x86 architecture chips, edge AI development, and the impact of China’s AI chip localization on the supply chain.

My personal view is that AI server concept stocks still have opportunities through 2026, but selecting the right targets is crucial. Companies with liquid cooling technology, high-density computing solutions, and deep collaborations with chip giants are likely to remain favored by capital. However, entering now requires more caution—don’t be fooled by short-term gains; focus on actual profitability and order visibility.