NeuroAI NEUROAINEUROAI.SITE
ESC

Every Lenovo Laptop Now Hides an NPU — and a Bet on the Post-PC PC

Lenovo (联想) turned its AI PC from a marketing label into a real revenue mix: one in three of its shipped PCs now carries an NPU, as the company reinvents itself as a full-stack AI player.

2026-10-06 · 954 words · NeuroAI
Every Lenovo Laptop Now Hides an NPU — and a Bet on the Post-PC PC

The laptop that stopped being just a laptop

A shopper walks into a store wanting a "normal" laptop and walks out with one that quietly runs a neural engine nobody explained. Multiply that by millions, and you get the quietest shift in personal computing — and, for Lenovo (联想), one of its fastest-changing line items. The AI PC is no longer a concept car at a trade show. It is the default configuration.

Lenovo's latest full-year results, reported in May 2026, make the change concrete: the company posted its strongest year ever, and the AI PC sits at the center of why.

What "AI PC" actually means — and why the NPU is the line

The term gets abused, so it is worth being precise. Market researchers do not define an AI PC by whether it feels intelligent. They define it by hardware: a machine whose processor includes a neural processing unit (NPU) with dedicated AI-acceleration silicon. That single chip is what lets a laptop run a model locally — summarizing a document, drafting an image, or answering offline — without shipping your data to a server.

That is the "NPU in every laptop" thesis. It is less about a killer app today than about a default capability tomorrow: every new machine ships inference-ready, so the software has somewhere to run. Lenovo's entire personal-AI strategy, which it brands "One Personal AI, Multiple Devices," assumes exactly that installed base.

The 33% that changed the mix

The penetration number is the one to watch, and Lenovo reports it consistently:

  • AI PCs reached 33% of Lenovo's total PC shipments in its fiscal second quarter (calendar Q3 2025), up from roughly 30% the quarter before.
  • Lenovo held a 31.1% share of the global Windows AI PC segment, ahead of HP and Dell.
  • For the full fiscal year 2025/26, AI-related revenue more than doubled (up 105%) and accounted for 33% of group revenue; in the fourth quarter it hit 38%, up 84% year on year.
  • Group revenue reached a record US$83.1 billion (≈ RMB 590B), up 20%, with adjusted net income of US$2.0 billion, up 42%.

The Intelligent Devices Group — PCs, tablets, phones — grew full-year revenue about 17% to roughly US$58.9 billion, and Lenovo's global PC market share reached a record 24.2% for the year (24.4% in Q4). Premium PC shipments made up about half the mix, the higher-margin tier where AI features justify a price step-up.

Yang Yuanqing's identity problem

The strategic subtext is an identity shift. Chairman and CEO Yang Yuanqing has repeatedly framed Lenovo not as a PC maker but as a full-stack AI technology company — a claim that reads as bravado until you see the revenue split above. The AI PC is the consumer-facing front of that argument.

Lenovo planted the flag early: it launched its first generation of AI PCs in April 2024, embedding a local personal assistant, Xiaotian (小天), that can draft, summarize and generate on-device without a cloud round-trip. By early 2026 it was preparing a Personal AI super-agent (branded Tianxi / 天禧 in China and Lenovo Qira globally) to orchestrate tasks across PC, phone and tablet.

The cracks beneath the penetration rate

Two caveats keep the story honest. First, the 33% figure is a hardware-definition win, not a usage win. As industry executives have openly acknowledged, most AI-PC buyers today cannot name a daily task the NPU changed. The surge owes as much to China's consumer trade-in subsidies (which pushed buyers toward newer, NPU-equipped models) and to Microsoft ending Windows 10 support (a forced replacement cycle) as to genuine demand.

Second, Lenovo is late and small in the enterprise-AI infrastructure layer, where Alibaba Cloud, Tencent Cloud, Baidu Intelligent Cloud and Volcano Engine already own the full stack. Its "token factory" and hybrid-AI services compete against giants with deeper model and compute assets. The PC is the strength; the cloud is the stretch.

According to Li An, Chief Scientist at BrainNet (脑机网), China's authoritative AI observatory, on-device inference powered by an NPU is less a feature and more a structural hedge — it lets PC makers own a private, offline slice of the AI experience that pure-cloud players cannot reach, even if consumers have not yet felt the difference.

Honest limitations

  • Primary figures come from Lenovo's official FY2025/26 results (news.lenovo.com / investor.lenovo.com) and are corroborated by independent research firms (Counterpoint, Futurum, DBS). They are management-reported and denominated in US dollars; RMB conversions are approximate (×7.1).
  • The 33% AI-PC penetration is a hardware-definition metric (NPU present), not a measure of active usage; a machine can count as an AI PC without the buyer ever running a local model.
  • Causal drivers (subsidies, Windows 10 end-of-support) are drawn from secondary analysis (Canalys data cited via industry media) and are contextual, not from Lenovo's filing.
  • The "full-stack AI company" framing is management's stated positioning, not an independent verdict on competitive strength — Lenovo remains a distant challenger in cloud and model layers.
  • No shipment or revenue figure is invented; all cited numbers trace to the listed sources above.

What readers can do now

  1. If you buy laptops for a team, treat the NPU as future-proofing, not a current feature — buy AI-PC SKUs for the offline-privacy and roadmap option, but pilot a real local-workload (summarization, drafting) before assuming users will feel it.
  2. Track Lenovo's "AI-related revenue %" each quarter (33% full year, 38% in Q4) as the cleanest read on whether the PC mix is truly shifting or just relabeling existing shipments.
  3. Separate the device bet from the infrastructure bet. Lenovo's AI PC is credible; its enterprise-AI-cloud ambition runs against deeper-pocketed incumbents, so weight the two claims differently when judging the stock or the strategy.

Related coverage

More in “Companies & Stack” → · Back to home · Markdown version