UBS Asia Hardware 2026 Outlook: Nuggets upstream components and foundries to avoid high-cost brands

The Zhitong Finance App learned that UBS released a research report stating that 2026 will be another year of strong growth in the AI supply chain. With the continuous expansion of capital expenses of hyperscale cloud service providers, rapid changes in AI hardware technology, and tight balance in the component supply chain, the Asian hardware industry is ushering in a new round of development opportunities.

Hyperscale cloud service provider capital expenditure remains strong

The report predicts that in 2026, the capital expenditure budgets of the world's top cloud service providers will increase significantly by 34% year over year, reaching about 424 billion US dollars. This impetus is mainly due to the continued expansion of cloud services and internet services, where the growth rate of cloud service revenue has rebounded to more than 28%.

Meanwhile, leading AI model makers such as OpenAI and Anthropic are actively investing to enhance model capabilities and maintain competitive advantages over giants such as Google. Despite market concerns about AI monetization and electricity supply, these factors are not enough to hinder the scale of spending in 2026.

Comprehensive win-win between AI GPU, ASIC and server

UBS believes 2026 will present a situation where AI GPUs and customized chips (ASICs) will prosper together. Nvidia's Blackwell platform (GB200/GB300) will enter large-scale delivery in 2026, and it is expected that approximately 60,000 racks will be delivered throughout the year. With the launch of the Rubin platform in the second half of 2026, the market will experience a smooth transition from Blackwell to Rubin. Rubin's modular cable-free computing tray design can speed up assembly and maintenance by 18 times.

Customized chips (ASICs), led by Google and Amazon, are being rolled out at an accelerated pace. Google's TPU v7 and Amazon's Trainium T3 are expected to be deployed on a large scale in 2026, and Meta is also planning to launch its ASIC solution in the second half of the same year.

In addition to the AI boom, traditional general-purpose servers also showed high single-digit to median recovery growth.

The “spring” of component vendors and the “headwind” of brands

UBS indicated that growth dividends are unevenly distributed upstream and downstream of the industrial chain. On the one hand, the overall scale of the market for rack power supplies, cooling, PCB/substrates, etc. is expected to continue to expand. On the other hand, the tight supply of various raw materials (especially memory) has created a “seller's market” for components, giving them stronger pricing power. As a result, component vendors will benefit doubly from the “sharp rise in volume and price.”

The supply of core components such as storage continues to be tight, forming a “seller's market”. Although this is beneficial to component suppliers, it also puts high cost pressure on PC and smartphone brands. Affected by the high price of memory, UBS lowered its 2026 PC shipment forecast, which is expected to decline by 4% for the year. Smartphone growth also slowed to 2% due to rising commodity prices.

Prefer AI hardware and component vendors

UBS believes that the 2026 stock price will still be driven by sales and profit growth rather than further valuation expansion. The bank still believes that the best opportunity exists among suppliers that can fully benefit from the increase in AI servers and increase the value or added value of stand-alone machines. The stocks it is optimistic about include server ODM vendors (Guangda, Hon Hai, Weichuang, Weiying) and component suppliers (Delta, Jianse, AVC), as well as Dalliguang, which has benefited from upgraded folding screens and variable aperture specifications in the lens field.

Considering the higher investment costs (especially storage), UBS is more cautious about brand companies (Asus, Lenovo, Gigabyte, MSI) and ODM and component companies (Renbao, Heshuo, Inventec, Jingshuo) that benefit less from AI servers.