Intel is being talked about as a name that could shake up premium AI laptops over the next few years. By putting on-package LPDDR6 on Razor Lake-AX, the company could increase memory bandwidth, cut power consumption and free up more room for thin and light designs. If it plays out the way the leaks suggest, the change would also improve integrated graphics performance and local AI processing. It is also a sign that Intel is preparing to compete more directly with the hardware approach Medusa Halo is pursuing. For anyone who wants an AI laptop that is both compact and powerful, this is a development worth watching closely.
How does Intel Razor Lake-AX's new RAM placement benefit AI PC laptops?
According to a leak about Intel Razor Lake-AX reported by Wccftech, Intel may bring on-package memory back to high-performance laptops after Lunar Lake, with the timeframe mentioned being around 2028. What stands out is that this approach is not aimed at ultra-low-power machines. Instead, it targets a larger SoC with more CPU cores, a stronger iGPU and greater bandwidth demands. Paired with LPDDR6, the chip would have more headroom to feed integrated graphics, the NPU and shared-memory workloads without making the device much bigger.
The biggest difference between this design and the familiar RAM layout is the distance data has to travel. In many of today's laptops, DDR5 or LPDDR5X RAM still sits on the mainboard, which means signals have to cross longer traces and the board has to set aside space for the power circuitry around the memory chips. With on-package memory, Intel Razor Lake-AX could shorten the signal path, waste less power and open the door to higher bandwidth at the same power level. That makes LPDDR6 more than just a newer generation of RAM: it is the piece that lets a slim design still run a large iGPU and sustain local AI workloads more reliably.
Where could LPDDR6 help Intel Razor Lake-AX take on Medusa Halo?
The rumor is drawing attention because of the rival Intel is aiming at. The AX segment is described as targeting premium laptops with powerful integrated GPUs, which is exactly where AMD is applying pressure with its Halo APUs. With Panther Lake stopping at LPDDR5X 9600 MT/s, a jump to LPDDR6 could give Intel Razor Lake-AX a noticeable bandwidth boost, enough to feed a larger iGPU, handle heavier textures and share memory more effectively with the NPU. For AI PC laptops, that is a very practical benefit, because with the same amount of RAM the system can serve the CPU, integrated graphics and AI block more efficiently.
The table below shows why this change matters more than simply raising clock speeds.
| Memory layout | Strengths | What users notice |
|---|---|---|
| RAM on the mainboard, as in many laptops today | Easy to implement, flexible configurations | Stable, but bandwidth and power efficiency are limited by longer traces |
| On-package LPDDR6 next to the SoC | Higher bandwidth, better power efficiency, more compact board | The iGPU and NPU get more headroom for AI, graphics and heavy multitasking in a thin chassis |
| High-bandwidth APU design like the Halo rival | Suits a powerful integrated GPU and unified memory | Smaller than a traditional workstation, yet still capable of content creation and local AI |
This is also why the race with the Ryzen AI MAX 495 keeps coming up. Now that AMD has shown that a high-end APU can be compelling thanks to a strong integrated GPU and a large pool of shared memory, Intel Razor Lake-AX needs an answer that goes beyond CPU core counts to how data is fed to the entire system. LPDDR6 is valuable because it directly affects preview rendering speed, the ability to run local AI models, smoothness on high-resolution displays and battery life.
Of course, this is still a rumor, and Intel has not announced the final configuration of this chip family. On-package designs usually come with trade-offs in upgradeability and repairability, so not everyone will like RAM being more tightly bound to the SoC. Even so, if Intel really does choose this new RAM standard, the message is fairly clear: the race for high-performance laptops is no longer just about adding cores or raising clocks, but about optimizing the whole memory architecture so that thinner AI PCs can still be fast enough, cool enough and efficient enough.
