What this is
Intel's next-gen mobile processor, Razor Lake AX, has surfaced in the update logs of hardware detection tool HWiNFO, with a 2027 launch window expected. Leaked specs point to 6-channel memory and bandwidth approaching 512 GB/s — roughly twice today's flagship Strix Halo (256 GB/s). Its direct rival is AMD's still-unannounced Medusa Halo; both target high-performance notebooks and small workstations in the APU (CPU with strong integrated graphics) category.
On the surface, this looks like routine silicon news. But that bandwidth figure carries an implication we want to flag: it's the hardest bottleneck for running large AI models locally. Doubling it means the same class of device could comfortably run a tier larger model.
Industry view
The local AI community — think r/LocalLLaMA-style forums — is broadly excited. Strix Halo machines already handle 70B-parameter models locally today. With bandwidth doubled, 2027 consumer notebooks could plausibly run larger models at inference speeds approaching mid-tier cloud APIs.
The pushback is equally sharp. Veteran players raise three concerns: first, Intel routinely discounts paper specs in real-world performance, so whether 512 GB/s is peak or sustained throughput remains unconfirmed; second, a 2027 launch plus OEM rollout and software stack maturation means genuinely usable products likely won't arrive until 2028; third, AMD and Qualcomm are pushing comparable silicon in parallel — the resulting hardware arms race could actually lengthen the software ecosystem's maturation cycle, delaying benefits to consumers.
Impact on regular people
For enterprise IT: Post-2027, the long-standing compliance objection that "sensitive data can't leave the building" becomes, for the first time, indefensible at the hardware level. Local AI workstations can start entering three-year procurement plans.
For working professionals: When you replace your machine over the next two to three years, "can it run large models locally" will sit alongside RAM and screen size on your checklist. A third path is forming beyond Mac Studio and Strix Halo workstations.
For the consumer market: No short-term impact — it doesn't ship until 2027. But the signal is clear: chip vendors have made "running AI locally" the central pitch for their next generation, and PC buying logic will shift accordingly.