DAVOS, Switzerland, January 23, 2026, 18:23 CET
- Lenovo is seeking partnerships with multiple large language model providers to power devices, its CFO said at the World Economic Forum in Davos.
- The company plans to pass higher memory chip costs on to customers, Cheng said.
- Lenovo’s January tie-up with Nvidia targets faster rollout of liquid-cooled AI data centre systems.
Lenovo is hunting partnerships with multiple large language models around the world to power its devices as it tries to position itself as a global AI player, chief financial officer Winston Cheng said on the sidelines of the World Economic Forum in Davos. Cheng also said he sees an AI bubble in private and public market valuations and urged investors to watch operating costs, not just spending plans. 1
The push comes as device makers try to turn generative AI — software that can produce text and other content — into a standard feature on phones, PCs and wearables. It is also turning into a supply-chain and compliance problem, as rules and data limits start to vary by country.
Lenovo’s answer is to stay flexible. The company wants its products to work with different AI “brains” depending on what is allowed, what customers want, and who is easiest to integrate this quarter.
“We are the only company besides Apple with significant market share across both PCs and mobiles, and in the open Android and Windows ecosystems,” Cheng said. 2
Apple works only with OpenAI and Google’s Gemini, Cheng said, while Lenovo wants to sign up far more model developers. He named Humain in Saudi Arabia, Mistral AI in Europe and China’s Alibaba and DeepSeek as potential partners, adding: “We’re taking an orchestrator approach.” 3
Large language models, or LLMs, are software trained on vast amounts of text and code. They sit behind many chatbots and writing tools, and can summarise, answer questions and generate drafts, though they can also make mistakes.
Earlier this month Lenovo introduced Qira, a built-in, cross-device system meant to tie those models into a single experience across hardware. “Lenovo Qira is not another assistant, it’s a new way intelligence shows up across your devices,” Dan Dery, a vice president in Lenovo’s device group, said in a company release. 4
Lenovo has little interest in building its own flagship model, Cheng said, pointing to regulatory hurdles. He joined the company in 2024 and became CFO in April 2025, according to reports of his comments in Davos.
Memory chip prices are another pressure point. Cheng said costs were rising and Lenovo planned to pass the increases on to customers, while the company’s January partnership with Nvidia aims to help AI cloud providers bring data centres online faster using a liquid-cooled hybrid setup — “liquid-cooled” meaning heat is carried away by fluid rather than air. Cheng said the firms would focus on the “global deployment” of that capability and manufacture locally, and may consider launches in Asia or the Middle East. 5
But a multi-partner strategy carries its own headaches. Passing on higher component costs can squeeze demand, and stitching together several models across regions can slow product rollouts if one partner stumbles or regulators shift the goalposts.
