Intel Might Kill Upgradable RAM in Future Laptops

Looks like Intel is doubling down on its on-package memory experiment, moving it from ultraportable laptops into high-performance mobile workstations. The company’s roadmap has revealed a new family of processors codenamed Razor Lake-AX. The key hardware feature of these chips is placing LPDDR6 memory directly on the same substrate as the CPU die.

Going the monolithic route over SO-DIMM isn’t completely new for Intel. Engineers have used a similar approach before in Lunar Lake (Core Ultra 200V), soldering up to 32GB of LPDDR5X-8533 right onto the package, but Razor Lake-AX is aimed at a different class of devices. While Lunar Lake saved ultrabook makers from extra millimeters of thickness and watts of power draw, the AX version is designed to solve memory bandwidth bottlenecks for powerful integrated graphics and AI accelerators.

According to early info, the Razor Lake platform in general is an optimized evolution of Nova Lake and is expected to hit the market in 2027. It will feature high-performance cores on the Griffin Cove architecture and power-efficient Golden Eagle cores. However, the AX version is a later product (pushed to a 2028 release) and will likely get next-gen graphics — either the final revision of Xe3P (Celestial) or an early version of Xe4 (Druid). To feed such a power-hungry GPU, the move to LPDDR6 is pretty much essential.

The leak also mentions the potential use of Intel’s own Z-Angle Memory (ZAM), though for now that’s just speculation. The main focus is on the industry-standard LPDDR6. Moving to on-package memory carries clear risks for both the vendor and consumers. First, repairability and configuration flexibility drop to zero. Buyers will have to choose their RAM capacity (likely 32, 64 GB or more) at the time of purchase, with no chance of upgrading later.

That said, the architectural benefits are obvious: lower latency, better power efficiency, and a significant reduction in motherboard real estate. For system integrators, this means they can build more compact and cooler-running systems without a discrete GPU in the lower and mid-range segments. Razor Lake-AX is expected to carve out a niche where not only AMD is actively competing, but also Apple with its M-series chips that have long used unified memory on package. Intel now has to prove that its solution scales better than the competition, delivering x86 applications in a familiar environment without compatibility issues or the stutter of instruction translation.

Can Intel deliver enough of a performance boost with fixed, non-upgradable memory to make users overlook the lack of upgrades, or does the mobile market still value config flexibility too much? Drop your thoughts in the comments.