Google makes an AI play with ChromeOS PCs

Google is making a strategic play in the AI PC space in a bid to slow Microsoft’s generative AI (genAI) technologies in Windows 11 from gobbling up every desktop.

The company is promoting its brand of AI PCs with Gemini AI services integrated into ChromeOS, which is a Linux-based OS in Chromebooks. The goal: make Gemini a viable alternative to Microsoft’s Copilot, which is prepackaged in Windows 11.

Microsoft has classified AI PCs as laptops with specialized neural chips that can run genAI tools without an internet connection. The company introduced the first Copilot+ PCs last year with chips from Intel, AMD, and Qualcomm.

Google’s alternative are Chromebook AI PCs, which also have specialized AI chips that can run genAI tasks on the device.

“When people generally talk about AI PCs, they’re referring to Copilot+ on Windows, Apple Intelligence for Apple, and then more and more there’s Chromebooks that have an NPU [neural processing unit],” said Jitesh Ubrani, research manager at IDC. “They’re trying to take advantage of them.”

Lenovo last month introduced the Chromebook Plus 14, which Google called its “most advanced Chromebook Plus yet.” It is the first Chromebook with specialized silicon that includes a MediaTek AI chip that runs at 50 TOPS (tera operations per second).

Lenovo Chromebook Plus 14
Lenovo

Google’s move comes as IT buyers face the need to upgrade PCs to Windows 11 because Windows 10 will reach the end of support in October.

Chromebooks are prevalent in the education markets in the US and Japan. While they  don’t pose a major threat to Windows, they do provide an alternative, Ubrani said.

Analysts have argued that many IT buyers are upgrading to AI PCs without any clarity on how to use those specialized chips for on-device AI. Instead, many companies see the upgrades as future-proofing their hardware or getting ahead of economic tariffs that could  raise PC prices.

Because Chromebooks rely heavily on the cloud and access to Gemini online, it’s not clear how NPUs would change things, Ubrani said.

The NPUs in Chromebooks are often tied to the cost of the devices, Ubrani said. IT buyers figure that if they’re investing in top-line Chromebooks — such as the Chromebook Plus segment — they might as well get an AI chip, Ubrani said.

“You essentially can’t get a processor without an NPU at certain price points — so, take it or leave it. That’s where we’re seeing NPUs in Chromebooks more and more,” Ubrani said.

Google and Microsoft are creating their own AI moats for PCs with NPUs. Microsoft, for example, is developing small language models (SLMs) for Microsoft 365 that can run offline on AI chips.

Google has no immediate plans for offline AI capabilities for Google Workspace. But client-side acceleration in browsers through AI chips could make large language models run faster on Chromebooks, said Jack Gold, principal analyst at J. Gold Associates. (Chromebooks use a browser interface.)

“Increasingly, like browsers today, AI in the browser will use all available device resources, just like browsers now use GPUs, if available. This is a play for making AI run faster and better on Chrome devices,” he said.

Google isn’t alone in trying to shake up the Windows ecosphere; even as it is integrating a heavy dose of Gemini into its Workspace apps, rival OpenAI — the company behind ChatGPT — is planning to launch a browser-based productivity app as an Office alternative.

The AI infrastructure adopted by companies could also drive decisions on AI PC purchases, Gold said. Chromebook AI PCs might be attractive to small-and-medium businesses and enterprise users heavily invested in Gemini.

“The issue of Windows 11 vs. Chrome devices is a broader competition than just Copilot/AI, but increasingly AI will be an important capability,” Gold said.

Other factors, including security and compliance, will play a role in PC purchases and adoption of models, he said. In particular, the idea of leaky AI models isn’t attractive to enterprises. “There is some risk…, especially if workers individually download things to their machine.”

Having better AI performance on laptops alone won’t guarantee sales, said Ranjit Atwal, research director at Gartner. “The development of the small language models and how efficiently they run on ‘more’ TOPS will be the key,” Atwal said.

In the end, the PC should be seen just one more front for the ongoing AI battle between Google and Microsoft, Atwal said. “I think Copilot vs. Gemini is the battle — one is enterprise and the other consumer,” Atwal said.

The key will be SLMs that drive specific tasks, applications and multi-modal interfaces, Atwal said.