AI has exploded in popularity over the last few years, and there are now many ways to use it. AI applications exist online for easy access in the cloud, lightweight AI features come built into many PCs and some AI tools can be set up to run on local hardware for custom setups. If you’re in the market for a new PC and want to tap into AI, it’s worth taking some time to consider what you want to do with AI and understand what different hardware can enable. MSI’s recent laptops offer a range of form factors and hardware capabilities, letting you find one that will fit your personal approach to AI.
For heavy local AI workloads, the MSI Raider 16 Max offers serious capabilities thanks to robust GPU hardware. The MSI Stealth 16 AI+ offers a middle ground with extra portability and battery life while coming equipped to run moderate local AI models. For lightweight portability and efficient local AI with the cloud for offloading heavier workflows, the MSI Prestige 16 AI+ is a strong option.
Model |
Platform |
Graphics |
GPU VRAM |
Practical AI tier |
Starting MSRP |
Best for: |
Raider 16 Max |
Up to Intel Core Ultra 9 290HX Plus |
Up to NVIDIA GeForce RTX 5090 Laptop GPU |
24GB GDDR7 |
Approx. 30B-class models at suitable quantization* |
Heavy local AI and professional use |
|
Stealth 16 AI+ |
Up to Intel Core Ultra 9 386H |
Up to NVIDIA GeForce RTX 5080 Laptop GPU |
16GB GDDR7 |
Approx. 7B–14B-class models at common quantization* |
Moderate local AI use and portability |
|
Prestige 16 AI+ |
Up to Intel Core Ultra X7 358H |
Up to Intel Arc B390 |
Shared system memory |
Smaller on-device models + efficient AI + cloud AI |
Windows AI features, cloud AI and premium design |
The rundown
Which MSI laptop is right for AI needs?
Choose the Raider 16 Max for heavy local AI, the Stealth 16 AI+ for moderate local AI with portability, or the Prestige 16 AI+ for efficient everyday AI and cloud-assisted workflows.
Which laptop is right for heavy local AI use?
The MSI Raider 16 Max is the strongest choice for heavy local AI, with up to an RTX 5090 Laptop GPU, 24GB of VRAM, and up to 300W of combined CPU and GPU power.
Which laptop can run moderate AI workloads?
The MSI Stealth 16 AI+ balances local AI performance and portability, with up to an RTX 5080 Laptop GPU and 16GB of VRAM for moderately demanding local models and creative AI workloads.
Is there a thin-and-light laptop that can tap into AI capabilities?
Yes. The Prestige 16 AI+ is designed for portable, power-efficient AI, combining an NPU and integrated Intel Arc graphics for everyday on-device tasks while using cloud AI for heavier workloads.
Which laptop is right for heavy local AI use? MSI Raider 16 Max
Running AI models and tools locally can call for some serious hardware. Often, the essential component is a powerful GPU. The NVIDIA GeForce RTX 5090 Laptop GPU combines extreme AI processing speeds of up to 1,824 TOPS, along with 24GB of high-bandwidth GDDR7 memory that can hold large AI models and context windows, supporting approximately 30B-parameter-class models with the appropriate quantization, though supported model size can vary depending on quantization level. This is critical for heavy AI models, advanced coding, RAG applications, and agentic workflows. When you need an AI that understands a deep, complex prompt and can take numerous actions to produce results, this VRAM capacity is crucial and the speed generates results quicker.
The MSI Raider 16 Max is well equipped for this heavier use. It combines up to an Intel Core Ultra 9 290HX Plus CPU and RTX 5090 Laptop GPU, giving you the kind of extreme performance you might need for engineering and creative workflows or gaming, as well as the ability to load heavy AI models.
As a heavyweight gaming laptop, the MSI Raider 16 Max includes high-end power delivery with up to 300W of combined CPU and GPU power. This helps ensure the components get plenty of power to run at their full potential. It also features MSI’s Cooler Boost Trinity Intra Flow Technology, which combines three fans, six heat pipes, and five exhaust vents to keep heat in check during sustained AI workloads.
Beyond AI work, the MSI Raider 16 Max includes a 2.5K+ OLED panel with a 240Hz refresh rate that works well for fast-paced gaming and content creation alike. It has quad speakers for powerful audio while you work or play. And it offers upgradability with extra storage slots and user-accessible memory slots.
When you need to run large AI models and generate tokens rapidly, the Raider 16 Max is MSI’s top offering for heavy local AI use.
Which laptop can run moderate AI workloads? MSI Stealth 16 AI+
The MSI Stealth 16 AI+ is a sharp laptop that blurs the line between a classic productivity machine and a gaming laptop. With that in mind, it also balances portability and convenience with some strong AI capabilities.
This laptop doesn’t go up to an RTX 5090 Laptop GPU for extremely large AI models and extremely fast token generation. It does go up to an RTX 5080 Laptop GPU, which provides 16GB of GDDR7 memory. This is still enough for a lot of capable AI models, letting you tap in 7B-14B-class models with typical quantization levels, and it can run at up to 1,334 TOPS. When you need AI to pore over some documents or generate a few images, this kind of muscle can take that on.
When you want to run smaller AI models for simpler tasks like voice-to-text transcription, the Stealth 16 AI+ also has an NPU that can handle these tasks while running on a smaller power budget. This goes alongside the Stealth 16 AI+’s better portability, giving you the option to run lightweight AI while away from power outlets and still tap into some heavier programs when anchored to a desk with wall power.
Creators and gamers can still enjoy a quad-speaker system and 240Hz QHD+ OLED panel display, as well. Though MSI is packing all of this into a thinner chassis than typical high-end gaming systems at 16.65 mm, the Stealth 16 AI+ is still fully featured with dual RAM slots, dual SSD slots, Thunderbolt 4 ports and full-size HDMI and Ethernet ports.
All this makes the MSI Stealth 16 AI+ the choice for moderate AI workloads.
Is there a thin-and-light laptop that can tap into AI capabilities? MSI Prestige 16 AI+
Portability is often as critical for laptop users as anything else. If you can’t easily bring your laptop with you, it may not prove very useful. The MSI Prestige 16 AI+ is optimized for portability with a chic aluminum chassis weighing just 3.66 pounds and measuring as thin as 0.47 inches while still packing in capable but efficient hardware for up to 24 hours of battery life.
You get an elegant, clean interface with the 16-inch OLED touchscreen that provides a wide DCI-P3 color gamut for creative work and a 120Hz refresh rate for strong motion clarity in work and play alike.
The system is built around highly efficient Intel hardware with an Intel Core Ultra X7 358H CPU. This package integrates an Intel Arc B390 GPU and an NPU. With this fully integrated package, you get some lightweight AI muscle that can tackle workloads efficiently. These components aren’t as fast at token generation as high-power GPUs, but they can use the greater capacity of system memory for larger models and context windows. Practical model size and performance still depend on available memory, quantization, software support, runtime, and workload. They also operate on lower power, so you can readily use them even without a constant power supply, making them a great match for MSI AI Engine and Copilot+. When more AI horsepower is needed, you can shift over to cloud-based AI services.
So, if you’re after a thin-and-light laptop that can tap into AI capabilities, the MSI Prestige 16 AI+ is ready to get the job done.
Frequently asked questions
Is RAM or VRAM more important for AI workloads?
Both types of memory are important for AI workloads, but which matters more at any given moment depends on the system component you’re using to run your AI model. VRAM is integrated into a discrete GPU, giving it high-speed access to that memory. To run an AI model on the GPU effectively, it needs to fit into the VRAM. If it can’t, the system will consistently have to bring pieces of the model from system memory over, which will create a serious bottleneck in token generation.
If you’re running AI on the CPU or an integrated GPU (like the Arc B390) or NPU, the system memory will hold the model and VRAM won’t be relevant.
How much VRAM do I need for local AI?
VRAM requirements depend on model size, quantization, context length, and software. More VRAM generally allows larger models to remain fully on the GPU for faster inference. As a practical reference, 16GB VRAM is well suited to many 7B–14B-class quantized models, while 24GB VRAM provides more room for roughly 30B-class models at suitable quantization.
Which is better for AI: CPU, GPU or NPU?
CPUs can run AI workloads, but they are generally not well suited to them. NPUs are good for efficient AI processing, requiring little power and getting access to system memory (which can be quite large). GPUs are also quite capable at running AI workloads, and often can run much faster than even NPUs, but they require more power and can be constrained by their often comparatively limited VRAM capacities. That said, when a model fits into VRAM, token generation can be extremely fast.
When should I run AI locally instead of in the cloud?
Running AI locally gives you easy control over models, letting you switch quickly to those that work best for your application. It can also let you maintain control over data privacy, since no information needs to be sent out to the third-party AI service provider. Local AI operation also doesn’t come with a token limit or cost per token beyond the cost of electricity to run the model. AI in the cloud can then be reserved for cases where you need access to very large models that can’t fit into your system’s memory or need extreme processing power to get a response within a reasonable timeframe.
Can I run AI completely offline on these laptops?
Yes, when using a fully local model and application. In that setup, prompts and data can remain on the device without being sent to a cloud AI provider. Some applications may still require internet access for model downloads, updates, or optional cloud features.
Why are gaming computers good for AI?
The parallel processing that GPUs are so great at also works well for AI applications, and since most gaming laptops include powerful GPUs (and many modern GPUs even include extra AI accelerators), gaming laptops have proven versatile as AI machines. The high-bandwidth VRAM linked to GPUs also gives AI models a dedicated pool of fast memory, helping the GPU access AI models faster.
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.