The new Xeon 6 parts has up to 128 Performance-cores and built-in accelerators aimed at everything from virtualised RANs to machine learning and security tasks. The chip  includes Priority Core Turbo and Intel Speed Select, two turbo tools that favour heavy lifting cores while the rest tick over on lower power.

That kind of fine-grain control is exactly what hyperscalers and HPC users want for squeezing every last instruction cycle out of demanding AI inferencing and training workloads.

Intel claims its memory handling and data throughput at high capacity beat AMD’s EPYC chips, though those 9005-series monsters are holding their own with Google and Oracle both rolling them out at scale.

Nvidia’s endorsement of the Xeon 6 in its DGX B300 AI systems gives it credibility. Imperial College London has selected the chip to power its HX2 supercomputer for research across engineering, medicine and other scientific fields.

Chipzilla is resisting the surge in Arm-based processors. Apple, Ampere and Qualcomm which are cranking out efficient silicon for the data centre, and Google has already built Axion chips based on Arm to run its next wave of server gear.

Intel insists it is working closely with developers to fine-tune AI and ML workloads for its CPUs. That’s vital, because the competition has already moved aggressively.

But none of this is helping Intel's share price. Intel is down 32.4 per cent over the past year, while the broader chip industry has grown by 7.3 per cent.

Earnings estimates for 2025 and 2026 have both taken a beating, falling more than 30 per cent over the past 12 months. Zacks Investment Research ranks the Intel shares at a lukewarm #3, which translates to “hold and hope.”