IFA 2025: Copilot+ PCs Unleash On-Device AI on Windows 11 Laptops and Desktops

IFA 2025 delivered a clear message: Copilot+ PCs and Windows 11 are moving from concept to volume devices, and OEMs used Berlin’s stage to show how new silicon, on‑device AI acceleration, and platform-level security are reshaping what a modern laptop and desktop can do.

Overview​

The biggest theme at IFA 2025 was integration — not just faster processors and brighter screens, but hardware tuned to accelerate on‑device AI workloads, paired with Windows 11 features and partner services. Major announcements from Acer and Lenovo illustrated two complementary paths: ultraportable, business‑focused Copilot+ machines optimized for mobility and manageability, and high‑performance gaming/workstation hybrids that blur the line between desktop AI workstations and consumer laptops.
Key product highlights included Acer’s TravelMate X4 14 AI — a lightweight Copilot+ business laptop — alongside its Predator Helios 18P and refreshed Nitro line for gamers. Lenovo showcased the Legion Pro 7 for high‑end gaming and content work, the Legion Go 2 handheld Windows 11 gaming PC, and the ThinkPad X9 Aura Edition in a new Glacier/Glacial White finish. All of these devices are positioned around Windows 11 and Microsoft’s Copilot+ PC program, and they arrive as support for Windows 10 officially ends on October 14, 2025 — a hard date every buyer and IT pro must factor into upgrade plans.

---

Background: What is a Copilot+ PC?​

Copilot+ PCs are a product category centered on machines that pair Windows 11’s Copilot features with dedicated, locally accelerated AI capabilities. The design goal is to provide:

• Low‑latency, on‑device AI experiences for productivity and creativity.
• Hardware NPUs (neural processing units) or accelerator blocks inside SoCs for local model inference.
• Tight OS and firmware integration for manageability and security under the Windows 11 umbrella.

This category spans ultraportables for SMB and hybrid work, mainstream consumer laptops, and beefy desktop replacements geared for creators and developers who want local AI inferencing without always relying on cloud services.

---

IFA 2025: OEM Announcements and What They Mean​

Acer: From TravelMate business machines to Predator AI workstations​

Acer’s IFA lineup emphasized breadth — business Copilot+ PCs through to extreme gaming and workstation hybrids.

• TravelMate X4 14 AI:
• Positioned as a Copilot+ PC for SMBs and mobile professionals.
• Very light (around 1.27 kg) and thin (~15.9 mm), MIL‑STD durability, optional OLED display options, Wi‑Fi 7, and modern port sets.
• Intel Core Ultra Series CPUs (Series 2) power the device, and Acer advertises substantial on‑device AI performance figures. Buyers should expect meaningful NPU acceleration for tasks such as noise removal in calls, local summarization, and client‑side inferencing for productivity features.
• Note: publicly reported TOPS numbers vary between outlets and vendor materials — some communications list the TravelMate X4 at 40+ TOPS, while Acer product pages and press materials list up to 115 TOPS for certain Core Ultra configurations. This discrepancy underscores the need to read specific SKU specs carefully rather than relying on a single headline number.
• Predator Helios 18P:
• A bold move toward marketing a gaming laptop as a local AI workstation. Configurations include Intel Core Ultra 9 class processors (vPro options on some SKUs), NVIDIA GeForce RTX 5090 Laptop GPUs, and support for up to 192 GB ECC memory.
• The Helios 18P is notable for combining high GPU performance (RTX 50‑series laptop hardware), large memory footprints and enterprise features (vPro, ECC) — making it suitable for developers running large local models or creators with heavy 3D and video workloads.
• Thermal design and power delivery are critical here: these machines are heavy and thermally aggressive by design, and real‑world battery life is likely to be poor under sustained AI or GPU workloads.
• Nitro refresh and Orion desktops:
• Acer refreshed mainstream Nitro laptops and introduced high‑end Predator Orion desktops with RTX 50‑series GPUs. These are targeted at gamers who want AI features like accelerated upscaling, frame generation and local inferencing for streaming and creative tools.

Lenovo: Choice across gaming, handheld and professional lines​

Lenovo’s IFA presence reinforced a multi‑segment strategy — premium ThinkPad Copilot+ devices for enterprise plus Legion systems for gamers and creators.

• Legion Pro 7:
• Offered with AMD Ryzen 9 9955HX3D (Ryzen 9000 HX series) and up to NVIDIA GeForce RTX 5080 Laptop GPU.
• Lenovo highlights a combined CPU+GPU power ceiling — up to 275W TDP under Legion Coldfront vapor + hyperchamber thermal designs — to sustain high frame rates and AI workloads.
• The inclusion of high TDP envelopes and advanced cooling suggests Lenovo expects Legion devices to be used for local AI training/inference, real‑time content creation, and professional gaming.
• Legion Go 2:
• A Windows 11 handheld gaming PC with an 8.8‑inch OLED VRR display, up to 32 GB RAM and up to 2 TB PCIe storage.
• This iteration pushes the handheld format further into premium territory — higher specs, larger batteries, and detachable controls — but with price points that have drawn community scrutiny. It represents an ecosystem play: Windows 11, PC gaming services, and local performance for on‑the‑go AI enhancements (e.g., companion apps that use local models for chat or in‑game assistance).
• ThinkPad X9 Aura Edition (Glacier/Glacial White):
• The ThinkPad X9 continues to wear the Copilot+ PC banner, integrating Intel Core Ultra processors and 40+ TOPS of inference capability in a premium convertible/ultraportable design.
• Additional focus on security: Secured‑Core PC features, ToF human presence detection, dTPM, fingerprint sensors, and commercial manageability tools remain central to the ThinkPad positioning.

---

The Technical Reality: NPUs, TOPS, TDPs and What Buyers Should Actually Compare​

Marketing for AI laptops leans heavily on headline numbers: TOPS (tera‑operations per second) for NPUs, AI TOPS for GPUs, and TDP (thermal design/power) for sustained performance. These figures are useful, but they need context.

• TOPS is a raw throughput metric. It does not directly equate to real application performance unless you know the model architecture, operator support, memory bandwidth, and software stack being used. Comparing 40 TOPS to 115 TOPS is not apples‑to‑apples unless the exact workload and accelerator microarchitecture are specified.
• GPU AI TOPS (e.g., NVIDIA laptop RTX 50‑series AI TOPS numbers) are often reported separately from CPU/SoC NPU TOPS. A device with strong GPU TOPS but a weak NPU may still be superior for GPU‑accelerated model inference but worse for ultra‑low‑latency, low‑power background features.
• TDP is the other side of the coin. High TDP (e.g., combined 275W CPU+GPU budgets) enables desktop‑class performance but affects battery life, thermals, acoustic profile, and chassis weight. For ultraportables, a modest NPU with efficient performance per watt is often more valuable than a headline TOPS count.
• Memory type and capacity matter: LPDDR5x vs DDR5, capacity and whether memory is soldered vs user‑upgradeable, and the presence of ECC can materially affect workloads in content creation or data‑sensitive AI tasks.

What buyers should compare directly

• Effective TOPS for the target workload (inference vs inference+training).
• Sustained performance: is the quoted TDP sustainable or burst only?
• Memory capacity and bandwidth (LPDDR5x 8533 MT/s vs DDR5x), and whether ECC is available for data integrity.
• Thermal design and cooling — do third‑party reviews show thermal throttling under extended loads?
• Battery capacity and realistic battery life under mixed and heavy workloads.
• Software stack and driver maturity — does the device ship with vendor‑validated ML runtimes and frameworks for on‑device models?

---

Security, Manageability and Enterprise Implications​

Windows 11 remains the central platform element. For enterprises, Copilot+ PCs bring a mix of opportunities and obligations.
Security and manageability advantages

• Secured‑Core PC features, dTPM and modern firmware flows provide stronger protection against firmware‑level attacks.
• Intel vPro and similar manageability suites allow IT teams to remotely manage and secure devices — useful for hybrid workforces.
• Local model inferencing reduces cloud exposure for sensitive data if set up correctly, because private data can be processed on the device rather than being uploaded to a service.

Risks and operational caveats

• On‑device AI introduces new attack surfaces: model files, inference runtimes, and accelerators must be secured and patched. Supply chain or firmware vulnerabilities affecting NPUs or GPU microcode could present a systemic risk.
• Licensing and telemetry: some Copilot features depend on cloud services and subscriptions; enterprises must audit data flows and licensing to ensure compliance.
• Windows 10 end‑of‑support (October 14, 2025) forces a migration decision. Organizations with legacy hardware will need to evaluate options: hardware upgrades to Windows 11‑capable devices, compatibility testing, or enrolling eligible systems in the consumer Extended Security Updates (ESU) program for limited extra time.

---

Windows 10 End of Support: Practical Implications (October 14, 2025)​

This is a fixed, non‑negotiable calendar event: Windows 10 will reach end of support on October 14, 2025. After that date, Microsoft will stop providing feature updates, security updates, and technical support for the OS.
Practical consequences

• Devices remaining on Windows 10 will no longer receive security patches from Microsoft — increasing exposure to vulnerabilities over time.
• Microsoft offers an Extended Security Updates (ESU) program for eligible devices for a limited period; enterprises and consumers should verify eligibility and costs.
• For many users and organizations, the most practical path is to evaluate Windows 11 upgrade paths or budget for hardware refreshes to Windows 11‑capable Copilot+ PCs.

Recommended steps for IT teams and individuals

• Inventory: identify all Windows 10 devices and collect hardware details (CPU, TPM, storage, RAM).
• Evaluate: use tools and vendor guidance to determine Windows 11 eligibility for each device.
• Prioritize: classify devices by business criticality and upgrade window.
• Migrate: plan pilot migrations to Windows 11, test line‑of‑business apps and Copilot features where relevant.
• If needed, enroll eligible systems in ESU as a stopgap while migration proceeds.

---

Real‑World UX: What On‑Device AI Will Actually Change​

Expect incremental but tangible improvements in daily workflows rather than overnight revolutions. Practical examples include:

• Faster, local meeting enhancements: real‑time transcription cleanup, background noise removal, and local summarization for calls — with less latency and less reliance on the cloud.
• Creative workflows: local accelerators enable faster generative tasks (e.g., image or video upscaling, preview generation) in apps that leverage native runtimes.
• Code and developer tools: local AI assistants that index codebases and provide suggestions with low latency, beneficial in environments where source code privacy is important.
• Gaming and streaming: AI‑enhanced upscaling, frame generation and latency‑reducing inferencing for stream enhancements and encoder assist.

The catch: application support matters. Hardware capability alone does not deliver a superior experience; software vendors need to optimize for the available NPUs, GPU features and Windows‑level APIs to produce smooth, reliable user experiences.

---

Pricing, Availability and the Upside/Downside Economics​

IFA 2025 shows a split market:

• Premium desktop‑class machines (Predator Helios 18P, Legion Pro 7) command high price points and aim at creators, AI developers, and prosumers. These systems provide near‑desktop capabilities in mobile form factors but often at the cost of portability, battery life and price.
• Ultraportables (TravelMate X4, ThinkPad X9) bring AI features to business users with better battery life and portability, but the most aggressive performance numbers may be reserved for larger, more expensive SKUs.

Economic tradeoffs to consider

• Long‑term value: devices with enterprise features (vPro, ECC, robust manageability) can reduce downtime and TCO for businesses even if upfront costs are higher.
• Upgrade cycles: the arrival of Copilot+ PCs may accelerate refresh cycles as organizations seek Windows 11 compatibility and local AI capabilities.
• Subscription ecosystems: bundled services (three months of PC Game Pass or Copilot subscriptions) can sweeten consumer purchases but add recurring cost considerations for power users and enterprises.

---

Critical Analysis: Strengths, Weaknesses and What to Watch​

Strengths​

• On‑device AI reduces latency and can preserve privacy for sensitive tasks.
• The integration of modern silicon (Intel Core Ultra, AMD Ryzen 9000 series) and discrete GPUs (RTX 50‑series) creates genuine capacity for local AI workloads.
• OEMs are blending enterprise features and gamer‑level performance, widening the use cases for powerful mobile machines.
• Windows 11’s security and manageability features are maturing alongside hardware innovations, making targeted enterprise deployments more feasible.

Weaknesses and risks​

• Marketing noise: inconsistent TOPS figures across channels and SKU complexity make meaningful comparisons difficult for consumers.
• Battery and thermals: desktop‑level performance in a laptop chassis will almost always demand compromises in weight, noise and battery life.
• Software readiness: applications must be rewritten or optimized for local accelerators and vendor SDKs to exploit hardware; until then, the practical gains will be mixed.
• Privacy and governance: local models reduce cloud exposure but introduce new administrative burdens for securing model files and runtimes.
• Cost and upgrade pressure: with Windows 10 end of support looming, some buyers may feel forced into early hardware refreshes, raising concerns about e‑waste and budget impact.

What to watch​

• Independent benchmarks that show sustained AI inference performance on real workloads (e.g., local summarization, image generation latency, model accuracy tradeoffs).
• Reviews that measure battery life under representative AI workloads, not just synthetic peak numbers.
• OEM and Microsoft clarity on the software stack: runtime updates, security patch cadence for NPUs, and enterprise deployment tooling for Copilot features.

---

Buying Guide: How to Choose a Copilot+ PC in the Near Term​

When comparing Copilot+ PCs, use this checklist to move beyond marketing claims:

• Confirm the exact CPU/GPU/NPU configuration for the SKU you will buy. Headline models may not represent all SKUs.
• Ask for sustained performance numbers or independent benchmarks, not just peak TOPS.
• Check memory type and capacity; for serious creative and AI workloads, more RAM and higher bandwidth matter.
• Validate upgradeability — can storage and RAM be upgraded later?
• For enterprise purchases, insist on vPro or equivalent manageability and Secured‑Core PC features.
• Evaluate battery capacity in watt‑hours and look for real‑world battery tests that include AI/ML workloads.
• Review thermal design: higher TDP claims must be supported by robust cooling to avoid throttling.
• Consider support and lifecycle: warranty terms, long‑term driver support for accelerators, and upgrade pathways as Windows evolves.

---

Conclusion​

IFA 2025 marks an inflection point where AI‑centric hardware and the Windows 11 platform converge into a tangible product category: the Copilot+ PC. The shift is pragmatic — local AI acceleration promises lower latency, new on‑device capabilities and enterprise controls — but the leap from impressive specs to durable, everyday improvements depends on software maturity, real‑world thermals, battery tradeoffs and clear vendor messaging.
Buyers and IT leaders must be discerning: read SKU details, look for independent sustained performance tests, and plan Windows migrations around the immovable calendar of Windows 10’s end of support on October 14, 2025. The new generation of laptops and desktops unveiled at IFA shows significant potential, but the real test will be how quickly software ecosystems and enterprise processes adapt to make on‑device AI both useful and secure in day‑to‑day computing.

---

Source: Windows Blog IFA 2025: Accelerating innovation with Copilot+ PCs and Windows 11