When Microsoft introduced the concept of Copilot+ PCs, it labeled it as a new era of AI-first Windows computing. After all, we have a machine with a dedicated AI chip that wants to locally perform a lot of AI-powered chores, thanks to a dedicated AI chip that can deliver at least 40 TOPS in terms of AI output.
For a while, Qualcomm was the only chipmaker appearing inside Copilot+ PCs, but AMD and Intel have finally started producing silicon that can match the AI performance criteria. Now, the situation has improved to an extent that you can pick an ultra-slim laptop, a workhorse, or even a gaming PC without losing the AI perks of a Copilot+ PC.
I recently had a chat with executives at Dell and Asus, and they mentioned how consumer demand for Cipolot+ PCs is higher at the moment compared to what they had initially expected. The enterprise uptake has also been fairly encouraging, though the whole Windows on Arm conundrum continues to linger. But the reality is that Copilot+ PCs are here to stay, and they offer enough meaningful benefits to consider them over legacy x86-based Windows machines.
They are fast (and getting faster)
Luke Larsen / Digital Trends
One of the first Copilot+ machines that we tried at Digital Trends was the Dell XPS, and it managed a feat that has eluded Intel and AMD for a while — beat the MacBook. At CineBench 2024, for example, it was faster than the MacBook Air M3 in multi-core (57%) tests.
The latest crop of Qualcomm silicon, especially the Snapdragon X2 Elite Extreme, is no different. The performance gulf is even higher with this 18-core monster, as it beats Apple’s M4 silicon and eclipses the M4 Pro in early tests. The onboard Adreno GPU is also ahead of Intel’s Lunar Lake and Apple’s M4 silicon, which is no mean feat.
Of course, these are benchmarks, but my real-world experience backs the claims of fluid performance. “My experience, however, has been better on the Dell laptop than the MacBook Air M3, which has been my daily driver since its launch. The XPS 13 does a noticeably better job at juggling multiple applications than the Apple laptop,” I wrote after testing the XPS 13 last year.
Microsoft
With the X2 series of processors, Qualcomm is touting up to 75% faster CPU performance, over twice the performance per watt and power efficiency for the GPU, and a massive 80TOPS AI output for the onboard NPU. Intel’s Panther Lake silicon is also on the horizon, and AMD’s current crop of Ryzen AI Max series processors is pretty good.
In a nutshell, you can now pick a Copilot+ PC without having to worry too much about the raw firepower, or even making a compromise between Arm or the x86 flavor of Windows.
Snappy and frugal
One of the biggest benefits of Copilot+ PCs, at least those with an onboard Qualcomm silicon and powering Windows on Arm, is the terrific battery life. The power draw by the Snapdragon processors is noticeably frugal, and even while running the PC in efficiency or lower power mode, they don’t stutter.
But it’s really the battery life that sets them apart. They’re not quite in the same league as the MacBooks, but getting pretty close. “The result is some of the best battery life I’ve seen from a Windows laptop,” said our review of the Asus ZenBook A14. “This is potentially multi-day battery life, depending on your workflow — something that, until recently, only the MacBook Air 13 with Apple’s incredibly efficient Apple Silicon processors could boast.”
With the upcoming slate of Snapdragon X2 Elite series processors, the battery mileage is only going up. Qualcomm is claiming 43% lower power draw for its next-gen silicon, and that means you can safely add a few extra hours to your usual per-charge mileage. I am also expecting similar gains from Intel’s 18A-based Panther Lake processors.
Based on the 2-nanometer processor, Intel is also claiming 40% higher power efficiency for the Panther Lake silicon while delivering a 50% jump in raw performance. Of course, they all meet the AI performance criteria set by Microsoft, so all the PCs powered by these chips will offer the full Copilot+ experience.
AI goes local
Talking about the Copilot+ experiences, this is the biggest draw. Local AI processing (and faster, while at it) means your system data never leaves the PC for cloud-based processing. In a nutshell, the whole AI experience is safer, as well. But most importantly, they open the doors for exclusive AI chops, too.
For example, Click to Do lets you select any item appearing on the screen and get more context about it, take appropriate action, launch a web search, and more. The AI, with vision capabilities, is aware of everything that’s appearing on the screen. Then we have Recall, which acts as a time machine and lets you revisit your PC usage history just the way you visit the screenshot gallery on your phone.
With Cocreator, you can create images without having to connect to the internet. Likewise, it also converts sketches or texts into polished visual assets. On a similar note, the Restyle feature applies AI-powered effects to photos using presets and text-based commands. I love the Super Resolution feature, in particular, as it lets you upscale and sharpen pictures easily.
On the more productive side of things, you get native live caption support with support for multiple languages. Then there’s the next-gen Semantic Search in the File Explorer and the system-wide Windows search, which lets you search files using natural language descriptions instead of specific names.
If you’re concerned about a security mishap
One of the biggest concerns with Copilot+ PCs is that in order to deliver their best AI experiences, they need deep system privileges, more so than a regular Windows PC. If your daily work involves handling sensitive data, this could be a risk if the features being used are not fully native and due to their compute-intensive nature and work on the edge, which means a portion of it is processed on the cloud.
“Copilot results do not inherit the security labels from the source files. This represents a serious risk for source files containing sensitive data,” says Concentric. In fields such as finance, research and development, and marketing, there is plenty of sensitive material that could use the advanced data processing capabilities of Copilot agents, but at the same time, it’s also pretty risky for multiple reasons.
AI processing for such material, aside from the obvious security alarms, also adds the burden of fact-checking and verification. Then there are the bigger risks, with features such as Recall. Following a botched introduction, furor, and subsequent relaunch with more security fortifications, the AI memory bank cum time machine is back.
But despite the tighter protocols, Kaspersky raises the risk of a scenario where the bad actor is someone who knows the PIN and can physically access your devices, or an instance where an attacker is trying remote access attacks by exploiting known vulnerabilities.
If you don’t want yet another subscription
It goes without saying that the biggest appeal of Copilot lies in its close integration with the Office suite of apps. Getting an agent to handle file edits, go through internal research material, generate takeaways, and more. But as good as all of it sounds, it’s not a free launch. At the very least, you will have to pay $20 per month for the Microsoft 365 Premium.
Microsoft
Arguably, it offers more value than what you would pay for a ChatGPT or Perplexity subscription. For enterprises seeking higher usage limits and security, the cost is even higher. A subscription is almost mandatory even for the most basic Copilot chores. There’s a cap of 60 AI credits per month for free users, even if you want to use Copilot for file analysis in Excel, or use perks such as drafting, summarization, and image creation in the Microsoft 365 apps.
In a nutshell, despite having all the local AI firepower at your disposal with a Copilot+ PC, it’s not a free lunch. Additionally, the new Copilot Vision and Voice capabilities for controlling a Windows 11 PC are limited to 30 minutes of usage per day, while the $20 monthly subscription raises it to only 60 minutes on a daily basis.
If your workflow has high-performance needs
Microsoft says Copilot+ PCs are all about AI conveniences. And that’s true to a large extent. But if your workflow entails the “real AI” work, such as model training, running large AI models locally, offline image generation, or anything with a heavy inference load, these machines are simply not up to the task. Or to put it the other way around, the hardware is not available out there.
Microsoft
You see, for demanding AI workflows, you need a laptop with a discrete graphics card and plenty of RAM. The developers I know almost universally suggest going with 128GB of RAM to be on the safe side. Then there’s the GPU bottleneck. Simply put, an NPU isn’t quite cut out for the task that a graphics engine can handle.
The only choice you are left with is picking up a gaming laptop with a discrete GPU and a processor that simultaneously meets the AI TOPS criteria for Copilot+ branding. And while at it, you also end up sacrificing a core selling point of Copilot+ PCs, which is long battery life. The most recent Copilot+ PC that I tried was the Asus ROG Zephyrus G14. It was stunning, but it could barely go beyond four hours on battery.
Even for non-AI tasks, such as video editing, a Copilot+ PC doesn’t quite deliver a definitive edge. A healthy few creative tools still rely on the CPU + GPU combination to get the job done, while the most advanced AI-powered tools in suites like those offered by Adobe still require cloud processing.
