BERLIN — The AI PC revolution roared into IFA 2025, where Acer and Lenovo planted their flags with a new wave of Copilot+ systems that promise to make on-device artificial intelligence a daily reality for Windows 11 users. These aren't concept devices; they're shipping hardware with dedicated neural processing units (NPUs), high-end GPUs, and Windows 11 optimizations designed to keep sensitive data local, slash latency, and unlock new creative and productivity workflows.

The timing is no accident. With Windows 10 end of support locked on October 14, 2025, OEMs are positioning Copilot+ PCs as the natural upgrade destination for businesses and consumers seeking hardware that runs Windows 11 natively and leverages its AI features. IFA proved that the category is splitting into two distinct lanes: ultraportable business machines that emphasize efficiency and manageability, and desktop-class mobile workstations that target gamers, creators, and AI developers.

What Exactly Is a Copilot+ PC?

Copilot+ PCs are a Microsoft-defined category that pairs Windows 11 Copilot features with locally accelerated AI capabilities. The formula demands a dedicated NPU inside the SoC, tight OS and firmware integration, and hardware that can run AI models without constantly phoning home to the cloud. The payoff includes low-latency inference for meeting transcription, background noise removal, real-time summarization, and creative AI tasks—all processed on-device for better privacy. These machines span a broad spectrum: lightweight 14-inch laptops for SMBs, mainstream consumer notebooks, and hulking desktop replacements with workstation-grade memory and graphics.

Acer: From Boardroom to AI Lab

Acer’s IFA lineup showcased ambitious breadth, bookended by a business-first Copilot+ ultraportable and a desktop-replacement gaming rig that the company explicitly markets as a local AI workstation.

TravelMate X4 14 AI

The TravelMate X4 14 AI targets SMBs and mobile professionals with a 1.27 kg chassis, MIL-STD durability, optional OLED display, Wi-Fi 7, and modern ports. Inside, it packs Intel Core Ultra Series 2 processors with an integrated NPU. Acer advertises substantial on-device AI performance, but here lies a critical caution: different press materials and product pages cite figures ranging from 40+ TOPS to 115 TOPS for certain configurations. This gap is not an error—it reflects the complexity of modern SoC designs where core count, power limits, and memory bandwidth can dramatically alter sustained AI throughput. Buyers must scrutinize individual SKU specs rather than relying on a single headline TOPS number. In practice, the NPU accelerates tasks like noise cancellation, local meeting summarization, and client-side inferencing for productivity apps.

Predator Helios 18P

Acer’s boldest move is the Predator Helios 18P, a 18-inch gaming laptop rebranded as a local AI powerhouse. Configurations top out with Intel Core Ultra 9 processors (vPro options on some SKUs), NVIDIA GeForce RTX 5090 Laptop GPU, and up to 192 GB of ECC memory. That memory capacity and error correction are usually reserved for server-class hardware, making the Helios 18P a compelling option for developers training or running large models locally, or for 3D artists and video editors handling massive datasets. Thermal design becomes the limiting factor: these machines are thick, heavy, and thermally aggressive. Expect poor battery life under sustained AI workloads—this is a machine meant to be plugged in. The combination of RTX 50-series GPU AI TOPS (often reported separately from NPU TOPS) and a high-wattage CPU means the Helios 18P can likely muscle through large batch inferencing or real-time ray tracing and AI upscaling simultaneously, but it also demands robust cooling that generates significant fan noise.

Nitro Refresh and Orion Desktops

Acer also refreshed its mainstream Nitro gaming laptops and launched high-end Predator Orion desktops equipped with RTX 50-series GPUs. These machines target gamers who want AI-accelerated upscaling, frame generation, and local inferencing for streaming and creative tools, but they lack the enterprise features and extreme memory ceilings of the Helios 18P.

Lenovo: From Handhelds to Enterprise

Lenovo’s IFA presence reinforced a multi-segment strategy, with premium ThinkPad Copilot+ devices for enterprise and Legion systems that blur the line between gaming and professional creation.

Legion Pro 7

Configured with AMD Ryzen 9 9955HX3D (part of the Ryzen 9000 HX series) and up to an NVIDIA GeForce RTX 5080 Laptop GPU, the Legion Pro 7 can pull a combined CPU+GPU power envelope of up to 275W TDP, sustained by Lenovo’s Coldfront vapor chamber and hyperchamber thermal design. That industrial cooling capacity suggests Lenovo expects users to push sustained AI training or inference loads, not just bursty gaming. The 3D V-Cache on the CPU offers benefits for simulation and compute-heavy tasks, while the high TDP ceiling ensures the RTX 5080 can run without severe throttling. This is a creator’s machine dressed in gamer aesthetics, and it signals that AMD is serious about placing AI-capable silicon in high-performance laptops.

Legion Go 2

The Legion Go 2 reimagines the handheld gaming PC with an 8.8-inch OLED VRR display, up to 32 GB of RAM, and up to 2 TB PCIe storage. Detachable controllers and a larger battery push the format toward premium territory, but community reaction at IFA focused on likely high price points. As a Windows 11 device, it inherits Copilot capabilities and can run local AI enhancements—think companion apps that use on-device models for in-game hints or real-time translation—but the handheld form factor imposes thermal and battery constraints that limit sustained AI workloads. It’s an ecosystem play: Lenovo wants to capture the intersection of PC gaming portability and emerging AI-enhanced experiences.

ThinkPad X9 Aura Edition

Now available in a new Glacier White finish, the ThinkPad X9 Aura Edition carries the Copilot+ PC banner with Intel Core Ultra processors delivering 40+ TOPS of inference capability. Lenovo emphasizes Secured-Core PC features, ToF human presence detection, dTPM, fingerprint sensors, and commercial manageability tools. For enterprise buyers, this machine represents the safe, manageable on-ramp to on-device AI without sacrificing the ThinkPad hallmarks of keyboard quality and durability.

The Technical Reality Behind the Marketing

Headline numbers like TOPS (tera-operations per second) and TDP (thermal design power) dominate AI laptop marketing, but they can mislead. TOPS is a raw throughput metric that tells you nothing about real application performance unless you know the model architecture, operator support, memory bandwidth, and software stack. Comparing 40 NPU TOPS from one SoC to 115 NPU TOPS from another is meaningless without workload context. Meanwhile, GPU AI TOPS (often advertised separately for NVIDIA RTX cards) may be far higher than NPU TOPS, but that brute force comes at the cost of power efficiency. A laptop with a strong GPU but weak NPU might excel at large model inference but fall short on low-power background tasks like temporal noise reduction.

TDP is equally critical. A combined 275W ceiling enables desktop-class performance but punishes battery life, thermals, and acoustics. For ultraportables, a modest NPU with high performance-per-watt often delivers more practical value. Memory configuration also matters deeply: LPDDR5x with high bandwidth improves NPU throughput, while soldered versus socketed RAM affects upgradeability. ECC support, rare outside workstations, becomes crucial for data integrity in professional AI and content creation pipelines.

Discerning buyers should compare:
- Effective TOPS for their specific workload (inference only vs. training)
- Sustained performance—can the machine hold peak TOPS or TDP, or does it throttle?
- Memory capacity, bandwidth, and whether ECC is available
- Thermal design and third-party review evidence of throttling
- Battery capacity in watt-hours and realistic battery life under heavy loads
- Software stack maturity: validated ML runtimes and framework support for the NPU and GPU

Windows 10 End of Support Adds Urgency

Microsoft will stop supporting Windows 10 on October 14, 2025—an immovable deadline. After that date, no more security patches or feature updates. Organizations and individuals must inventory hardware, assess Windows 11 eligibility, and plan migrations. The Extended Security Updates (ESU) program offers a temporary lifeline for eligible systems, but costs and limited duration make it a stopgap. For many, the arrival of Copilot+ PCs creates an opportunity to refresh aging hardware while gaining local AI capabilities and modern security features like Secured-Core PC and dTPM.

What On-Device AI Actually Changes Today

For all the hype, the immediate benefits are incremental but tangible: real-time meeting transcription cleanup that works even offline; background noise removal that doesn’t ship audio to a data center; local meeting summaries generated by a small language model running on the NPU; and faster generative AI previews in creative software. Developers can run local code assistants that respect source code privacy. Gamers get AI-enhanced upscaling and frame generation rendered by dedicated GPU cores. None of this replaces the cloud, but it reduces dependency and improves responsiveness.

The catch is software maturity. Applications must be written or optimized to tap these accelerators via Windows Copilot APIs or vendor SDKs. Until then, the hardware often sits underutilized. Microsoft and OEMs must clarify the software stack—runtime updates, security patches for NPU microcode, and enterprise deployment tooling—to turn potential into day-to-day value.

How to Choose a Copilot+ PC Right Now

  • Confirm the exact CPU/GPU/NPU configuration for the SKU you will buy; headline specs often apply only to top-tier configs.
  • Look for independent benchmarks measuring sustained AI inference, not just peak TOPS.
  • Check memory type and capacity; for serious AI workloads, 32 GB or more with high bandwidth is advisable; ECC matters for data-sensitive tasks.
  • Validate upgradeability—can you add RAM or storage later?
  • For enterprise, insist on vPro manageability and Secured-Core PC features.
  • Evaluate battery capacity and search for real-world battery tests that include AI workloads.
  • Review thermal design; high TDP claims need robust cooling to avoid throttling.
  • Consider support and lifecycle: warranty terms, long-term driver commitments for NPU and GPU, and upgrade pathways as Windows evolves.

The Road Ahead

IFA 2025 proves that AI-accelerated hardware is no longer a prototype. Acer and Lenovo are shipping machines with genuine on-device AI muscles, tied to a Windows 11 platform that increasingly bakes AI into the OS. But the gap between spec sheets and daily experience remains wide. Software ecosystems must catch up, and buyers must learn to evaluate AI performance beyond marketing-friendly TOPS numbers. The looming end of Windows 10 support provides a practical nudge, but the real test will be whether these Copilot+ PCs deliver reliable, secure, and useful AI experiences that go beyond the booth demos and actually improve the way we work, create, and play.