Samsung Electronics has begun testing a standalone neural processing unit (NPU) for Windows PCs with HP and Lenovo, according to a report from South Korean outlet ChosunBiz. The chip, codenamed GAIA, is designed as a discrete accelerator for AI workloads, rather than a full processor that would replace a laptop’s CPU or GPU. If the ongoing evaluations succeed, production PCs with GAIA could land in late 2027 or early 2028 — potentially reshaping how Windows laptops handle generative AI.

What GAIA Actually Is (and Isn't)

GAIA is being developed by Samsung's System LSI division on a 4nm manufacturing process. It is not a system-on-chip (SoC) that integrates general-purpose CPU cores or graphics. Instead, it functions as a dedicated AI accelerator — a bolt-on component that can be paired with an existing Intel, AMD, or even Qualcomm main processor to handle AI tasks like language models, image generation, and real-time translation.

This distinction is critical. Unlike Qualcomm's Snapdragon X, Intel's Core Ultra, or AMD's Ryzen AI, which all bake an NPU directly into the main chip package, GAIA is an add-on. Samsung reportedly aims to fill a gap: laptops built on platforms that lack a sufficiently powerful integrated NPU could gain AI processing muscle without a complete system redesign.

Samsung is also pursuing processing-in-memory (PIM) technology and specialized DRAM for GAIA, but concrete details about performance, power consumption, data formats, memory bandwidth, or software interfaces are absent. The chip is still an early prototype, and Samsung has made no public announcement about its specifications, pricing, or launch timeline.

What This Means for Windows Users

Home users and students: If GAIA reaches retail systems by 2028, it could enable more affordable AI-capable laptops. Today, a Copilot+ certified PC must have a processor with an integrated NPU hitting 40 trillion operations per second (TOPS). That requirement funnels buyers toward the latest generation of chips from Qualcomm, Intel, or AMD. A discrete NPU like GAIA could let laptop makers pair an older or cheaper main processor with Samsung's chip, potentially lowering prices while still running local AI features like Windows Studio Effects, Recall, or Cocreator. However, there's a catch: Microsoft's current Copilot+ certification explicitly ties the NPU to the main SoC. A laptop with GAIA alongside, say, a 12th Gen Intel Core might have the raw AI power but still not get the Copilot+ badge — and the software experiences that come with it.

Power users and creators: Anyone editing video, training local models, or experimenting with AI image generation stands to benefit from more AI compute that doesn't eat into the main GPU or CPU resources. GAIA could offload those tasks entirely, potentially improving battery life. But without knowing the chip's throughput, precision (INT8, FP16, etc.), or framework support, it's impossible to gauge real-world performance. Early adopters should be cautious: discrete accelerators have historically suffered from spotty software optimization.

IT administrators and enterprise buyers: The Copilot+ certification question is the elephant in the room. Microsoft has invested heavily in defining Copilot+ PCs as devices with integrated NPUs, tightly coupled to the main processor. A discrete GAIA chip would not automatically qualify a laptop for Copilot+ features like enhanced security, system-wide AI indexing, or third-party app acceleration. Microsoft has been exploring alternative pathways — recent leaks suggest the company is testing Copilot+ workloads on discrete GPUs — but no official policy change exists. For organizations planning large hardware refreshes around Copilot+, GAIA introduces uncertainty. A laptop purchased today with an integrated NPU is a safe bet. One that relies on a future add-on NPU is a gamble.

How We Got Here

Microsoft's Copilot+ PC framework, introduced in 2024, marked a sharp turn in Windows computing. For the first time, the operating system demanded a minimum NPU capability — 40 TOPS or more — baked into the main processor. This requirement immediately split the market: only laptops with Qualcomm Snapdragon X, Intel Core Ultra (Series 2), or AMD Ryzen AI 300 series chips qualified for the full Copilot+ experience, including features like Recall and advanced Windows Studio Effects.

Samsung, meanwhile, has been absent from the Windows PC silicon scene since its 2012 Exynos 5 Dual, which powered a handful of Chromebooks. The GAIA project, led by its System LSI arm, signals a return — but not in the way many assumed. Rather than competing head-to-head with Intel or Qualcomm, Samsung appears to be positioning GAIA as an AI co-processor, akin to how some smartphones pair a baseband modem with an external AI chip.

The broader industry context is also shifting. NVIDIA's recently announced RTX Spark SoC takes the opposite approach: it integrates CPU, GPU, and NPU onto one chip, targeting high-end Copilot+ laptops. Samsung's bolt-on method is cheaper and more flexible, but it sidesteps the tight integration that Microsoft's certification currently demands. This divergence hints at a future where Microsoft may be forced to recognize raw AI capability regardless of how it's packaged — a development that could be accelerated if GAIA passes HP and Lenovo's validation.

What to Do Now

For most Windows users, the immediate advice is simple: don't wait for GAIA. If you're in the market for a new laptop and care about future-proof AI features, buy a current Copilot+ certified PC. Devices with Qualcomm, Intel, or AMD processors that meet the 40 TOPS threshold are widely available, and the software stack is mature enough for everyday use.

For IT managers: monitor Microsoft's developer communications closely. The Windows AI platform is evolving rapidly, and a relaxation of the Copilot+ SoC requirement would open the door for discrete accelerators. If your organization's timeline stretches into 2028, you may want to evaluate GAIA-equipped systems if and when they materialize, but don't delay current deployments on the promise of future hardware flexibility.

For enthusiasts and developers: keep an eye on HP and Lenovo's evaluations. If either company decides to incorporate GAIA into retail products, Samsung will need to release an SDK and driver support for Windows. That would be your signal to start experimenting. Until then, treat GAIA as a promising but unproven prototype.

Outlook

GAIA’s true impact hinges on two unknowns: its technical merit and Microsoft's willingness to adapt. If GAIA delivers competitive TOPS and strong framework support, it could fragment the Copilot+ ecosystem — or force Microsoft to revise its definition. The company has already shown flexibility by testing discrete GPU paths; a dedicated NPU might be an even cleaner solution. For now, Samsung's chip is a fascinating prototype that underscores the growing appetite for flexible AI hardware in Windows PCs. The next milestone will be any public word from HP, Lenovo, or Samsung about the evaluation results. Expect those signals well before 2027.