Lenovo’s IFA keynote and hands-on demos in Berlin crystallized a simple, audacious claim: within four to five years every personal computer will be an “AI PC” — a device with a built‑in Neural Processing Unit (NPU) and the on‑device intelligence to run many AI tasks locally. That declaration, voiced by Luca Rossi, Lenovo’s President of the Intelligent Devices Group, is more than marketing bravado; it maps to observable shifts in silicon, software, and procurement incentives today — but it also leans heavily on assumptions about software maturity, buyer behavior, and ecosystem integration that deserve careful scrutiny. (repubblica.it)
Lenovo used IFA 2025 to showcase a broad AI‑first portfolio — from updated ThinkPads and Concept devices to AR glasses, handheld gaming hardware, and AI‑aware docks and accessories. The company framed 2024 as the year hardware laid the groundwork, 2025 as the year hardware matured, and the next several years as the period when software and agent ecosystems turn NPUs and local inference into mainstream value. Lenovo’s expectations are quantified in two headline predictions: that AI PC global penetration is already measurable and that Lenovo expects to own an outsized share of that growing market, with rapid growth projected over the next 12–18 months. (news.lenovo.com)
This feature unpacks Lenovo’s claims, verifies the hard facts, assesses where the industry actually stands on NPUs and Copilot+ PCs, and evaluates the strengths and risks of a rapid AI PC transition for consumers, enterprises, and the PC ecosystem.
Industry reports and product briefings now commonly reference two practical NPU tiers:
Verification:
That said, the road from hardware capability to habitual, cross‑platform, super‑agent‑driven computing runs through software, standards, and economic realities. The main danger for OEMs and platform players is mistaking hardware capability for user value. If ISVs and agents fail to produce discoverable, trustworthy features that clearly improve daily workflows, then AI PC penetration will stall long before the “every PC” point.
Lenovo’s vision is bold and technically grounded; its realization depends on a complex sequence of software adoption, interoperability, and sensible marketing that converts buyer curiosity into sustained usage. The next 18 months will be decisive: expect shipping volumes, ISV announcements, and enterprise procurement language to either validate Rossi’s optimism — or to reveal the gaps that still need closing.
Source: Windows Central At IFA, Lenovo’s Luca Rossi predicted every PC will have AI in five years. Are you ready?
Overview
Lenovo used IFA 2025 to showcase a broad AI‑first portfolio — from updated ThinkPads and Concept devices to AR glasses, handheld gaming hardware, and AI‑aware docks and accessories. The company framed 2024 as the year hardware laid the groundwork, 2025 as the year hardware matured, and the next several years as the period when software and agent ecosystems turn NPUs and local inference into mainstream value. Lenovo’s expectations are quantified in two headline predictions: that AI PC global penetration is already measurable and that Lenovo expects to own an outsized share of that growing market, with rapid growth projected over the next 12–18 months. (news.lenovo.com)This feature unpacks Lenovo’s claims, verifies the hard facts, assesses where the industry actually stands on NPUs and Copilot+ PCs, and evaluates the strengths and risks of a rapid AI PC transition for consumers, enterprises, and the PC ecosystem.
Background: why the AI PC conversation matters now
The hardware inflection happened fast
The defining technical change underpinning the AI PC narrative is the arrival of dedicated NPUs in mainstream laptops and desktops. Microsoft’s Copilot+ program crystallized a minimum‑capability vision — devices with NPUs capable of 40+ TOPS (trillions of operations per second) for the highest tier of on‑device AI experiences — and OEMs responded by shipping machines across ARM and x86 platforms that embed NPUs of varying throughput. Microsoft documents Copilot+ as a class of Windows 11 PCs that includes NPUs in this performance band, explicitly highlighting on‑device inference and low‑latency features as differentiators. (microsoft.com)Industry reports and product briefings now commonly reference two practical NPU tiers:
- Entry/commodity features: NPUs in the ~10 TOPS range that enable local noise suppression, live transcription, and camera effects.
- Advanced Copilot‑class features: NPUs in the 40+ TOPS range aimed at heavier local generative or multi‑modal inference workloads.
A commercial tailwind: Windows 10 end of support
Microsoft’s announcement that Windows 10 will reach end of support on October 14, 2025, creates a built‑in replacement cycle for many users and organizations. Microsoft’s official lifecycle pages and support guidance underscore that after the cutoff date, security updates and technical assistance will end for Windows 10 — an important procurement lever for enterprises and consumers that may accelerate device refreshes. This timing gives OEMs and channel partners a natural policy reason to push Windows 11 Copilot+ PCs and AI‑capable devices as the recommended upgrade path. (support.microsoft.com, learn.microsoft.com)What Lenovo actually claimed at IFA — and what’s verified
The headline: “Every PC will be an AI PC in four, maximum five years”
Luca Rossi told attendees and interviewers that the industry is on a trajectory where AI functionality becomes standard across personal computing devices, predicting universal AI PC penetration in “four, maximum five years.” That statement is framed as a forecast based on current momentum in hardware availability, enterprise procurement, and software partners working to utilize local NPUs. Rossi also provided intermediate metrics: about 5% global AI PC penetration in early 2025, with Lenovo’s share of that segment near 30%, and the company forecasting to hit 50% share within 12–18 months. Those numbers, as presented at IFA, reflect internal sales proportions and market estimates rather than an external independent audit. (fudzilla.com)Verification:
- Multiple Lenovo press briefings and interviews with Rossi echo the ambition and similar numerical framing about growth and pacing. Those primary company statements are on record in Lenovo’s press materials and regional media coverage. (news.lenovo.com)
- Independent market trackers report strong AI‑capable device shipments and growth forecasts — for example, analyst commentary during 2025 projected significant AI laptop shipment growth from a small base (Gartner and others made multi‑year forecasts), but absolute percentages and OEM share figures are variable across tracker methodologies. Industry commentary supports a fast growth trajectory but not a universal consensus that “every PC” will be AI‑capable by 2029–2030. (windowscentral.com)
Hardware-first adoption, software lag
Rossi explicitly acknowledged that much of the current AI PC demand is driven by hardware desirability (thinness, battery life, premium designs) rather than killer AI apps. He compared the software maturity curve to the early days of mobile app stores — a long, multi‑year process before ubiquitous, indispensable applications emerged. Lenovo is optimistic about ISVs (Independent Software Vendors) porting applications to use local NPUs; the company cites several dozen to a hundred ISVs in early porting activities. That ISV engagement is real and visible in partner programs, but the scale and depth of porting — especially for complex generative tasks — will determine the timeframe for mass consumer behavioral change.The technical realities: NPUs, TOPS, and what they deliver
NPUs are necessary but not sufficient
NPUs provide specialized acceleration for matrix and tensor operations used in inference. In practice:- NPUs reduce latency for local inference and can reduce cloud calls for common features like live captions, background removal, and transcription.
- NPUs contribute to battery efficiency by moving workloads to lower‑power accelerators.
- NPUs’ usefulness depends on software that routes optimal work to them, and on model compression, quantization, and other inference optimizations. (microsoft.com)
TOPS is a marketing shorthand — use it with caution
TOPS communicates theoretical throughput but tells only part of the story. Real‑world AI feature speed, quality, and responsiveness depend on:- Model architecture and how well it’s optimized for the device’s NPU
- Memory subsystem bandwidth and latency
- Thermal headroom and power management
- Software stack (drivers, runtimes, ONNX‑style bridges)
Multiple OEM briefings and industry analysis explicitly warn that TOPS must be interpreted in context and tested on real workloads. Vendors sometimes use TOPS thresholds in marketing (Microsoft’s Copilot+ 40+ TOPS is an example), but claims about trillion‑parameter models on laptops or “support” for massive models often require technical caveats.
Software and the “super agent” thesis
What Lenovo calls a “super agent”
Rossi described a future of intent‑based computing driven by “super agents” — advanced assistants that transcend single apps and ecosystems, anticipate user needs, and autonomously trigger tasks and workflows across devices and contexts. In Lenovo’s view, that super agent may blend local AI (for privacy and latency) with cloud models (for scale) and coordinate across phones, PCs, wearables, and AR glasses. This model maps to broader industry language about agents and assistant ecosystems but remains largely conceptual at present. Lenovo has demonstrated on‑device agents (Lenovo AI Now, LLMs running locally with PKBs / personal knowledge bases), yet the “super agent” vision requires far deeper cross‑vendor standards and interoperability than currently exists. (news.lenovo.com)Software maturity is the main gating factor
Even if high‑TOPS NPUs become ubiquitous, useful mass adoption depends on three software conditions:- Real ISV adoption — mainstream productivity and creative apps must integrate local AI paths that materially improve workflows.
- Agent interoperability — assistants must be able to access and act across apps, OS services, and device contexts with predictable privacy and security boundaries.
- UX discoverability — average consumers must understand and trust these capabilities enough to change behavior. Lenovo and others expect a long tail between hardware readiness and software that creates new behavioral norms.
Business and procurement dynamics accelerating adoption
Windows 10 end‑of‑life is a real procurement signal
Enterprises and many consumers face a calendar deadline: October 14, 2025, after which Windows 10 loses mainstream support. Microsoft’s guidance encourages upgrade to Windows 11 or paid Extended Security Updates (ESU) for those who cannot upgrade immediately. That deadline is already shaping RFPs and procurement conversations; many organizations will use refresh cycles to swap older hardware for Windows 11 devices that can support Copilot and on‑device AI features. This is a pragmatic, verifiable lever accelerating Copilot+ and AI PC adoption in enterprise settings. (support.microsoft.com, learn.microsoft.com)Enterprise economics favor on‑device AI in some verticals
On‑device AI delivers operational benefits that resonate with certain verticals: lower latency for clinical transcription in healthcare, local privacy for finance and legal workflows, and resilience for field operations in manufacturing and logistics. These are tangible justification points procurement teams will use to specify AI‑capable hardware by default in many enterprise fleets. However, the ROI equation varies widely by use case and scale. Analysts and vendor briefings stress the need for measured proof‑of‑value before blanket deployments.Strengths of Lenovo’s vision
- Real product momentum: Lenovo’s device portfolio — from Copilot+ ThinkPads to novel form factors and AR glasses — demonstrates that the company is executing hardware that maps to the AI PC definition. This breadth gives Lenovo credibility when discussing device ecosystems. (news.lenovo.com)
- Channel and OEM scale: As a top global PC OEM, Lenovo can push inventory, enterprise deals, and partner programs to accelerate adoption. Their market share in AI PC shipments is material today, which supports Rossi’s optimism about faster near‑term growth.
- Ecosystem leverage: Lenovo’s partnerships with Microsoft, Intel, AMD, and Qualcomm — coupled with its own software experiments (Lenovo AI Now) — position it well to drive integrated experiences that combine hardware, device mgmt, and local AI capabilities. (news.lenovo.com)
Risks and caveats
- Software is the uncertain variable: Without widely adopted, differentiated, and proven AI applications that change workflows, hardware alone won’t convert mainstream users. Rossi acknowledged this; history shows platform shifts need killer apps and discoverable value.
- Overreliance on TOPS marketing: Users and IT buyers may conflate TOPS with user experience. Independent benchmarking and real workload tests will be essential to separate marketing from measurable gains. Analysts explicitly warn against treating TOPS as the whole performance story.
- Privacy, security, and regulatory friction: Local AI reduces cloud egress for sensitive data, but agent‑style autonomy raises new questions about consent, data flow, and corporate governance. Enterprises will demand robust controls and verifiable safeguards before allowing agents to act autonomously on business data.
- Fragmentation risk: If OEMs, cloud providers, and OS vendors implement agents and AI features in incompatible ways, the “super agent” dream of cross‑device intent orchestration may become a set of siloed assistants rather than a unified layer. That fragmentation would inhibit universal adoption. (news.lenovo.com)
What to watch next — practical indicators that Rossi’s timeline is realistic
- ISV porting velocity: Look for major cross‑platform integrations from Microsoft 365 peers, Adobe, and other productivity suites announcing optimized local inference paths.
- Benchmarks that move beyond TOPS: Real‑world user‑centric performance tests for transcription latency, generative image speed, and multi‑modal tasks on representative devices.
- Enterprise RFP language shifts: Procurement templates that specify baseline NPU TOPS ranges or Copilot+ feature support as minimums.
- Agent interoperability standards: Agreements or public APIs that let assistants access app contexts and device services across vendors securely.
- Price compression: AI PC price bands expand into mainstream segments (sub‑$700) with usable NPUs, not just premium tiers. (windowscentral.com)
Practical advice for buyers and IT leaders
- For consumers: If your current device meets needs and you’re not cloud‑sensitive, a mid‑cycle upgrade isn’t strictly necessary — but if you want future‑proofing for Copilot features and local AI, plan upgrades around OEM models that explicitly list NPU capability and Copilot+ support.
- For IT decision makers: Treat Windows 10 end of support as a hard deadline for risk mitigation; build pilot programs that evaluate on‑device AI features against actual workflows before full fleet upgrades.
- For developers and ISVs: Prioritize portable inference workflows and quantization/compilation tools that target the major NPU runtimes and integrate with OS‑level assistant APIs to ensure discoverability. (support.microsoft.com, microsoft.com)
Conclusion
Lenovo’s IFA 2025 messaging is a credible, well‑capitalized version of a broader industry narrative: NPUs are mainstreaming, Windows 11 Copilot+ has codified performance thresholds, and procurement incentives (notably Windows 10 end of life) create a tangible upgrade cycle. Those forces make Rossi’s timeline — a world where AI PC characteristics are normative within four to five years — plausible, particularly in enterprise and premium consumer segments.That said, the road from hardware capability to habitual, cross‑platform, super‑agent‑driven computing runs through software, standards, and economic realities. The main danger for OEMs and platform players is mistaking hardware capability for user value. If ISVs and agents fail to produce discoverable, trustworthy features that clearly improve daily workflows, then AI PC penetration will stall long before the “every PC” point.
Lenovo’s vision is bold and technically grounded; its realization depends on a complex sequence of software adoption, interoperability, and sensible marketing that converts buyer curiosity into sustained usage. The next 18 months will be decisive: expect shipping volumes, ISV announcements, and enterprise procurement language to either validate Rossi’s optimism — or to reveal the gaps that still need closing.
Source: Windows Central At IFA, Lenovo’s Luca Rossi predicted every PC will have AI in five years. Are you ready?