Paul Alcorn over at Tom’s Hardware has reported on further layoffs and enhanced budgetary cuts going on within Intel. While the capital preservation angle is required as the business attempts to pivot to regaining a competitive fabrication capability for its foundry services, understanding the future trajectory of Intel becomes unclear with successive actions. Resources which were poached from the competition a few years ago have either resigned for personal reasons and moved on to better things or were demoted prior to departing for a different opportunity. Intel’s decision to exit the flash memory market may have been a wise decision if the assertions of Phison’s CEO are accurate. Additional exits include the prebuilt server business, divestiture of Mobileye, WWAN solutions, cryptocurrency mining solutions, and more. Even with the amazing progress achieved related to the driver stack for Intel’s Arc series of GPUs, confidence that Intel will continue to invest in their GPU portfolio must be appropriately tempered.
Waiting for the eventual release of subsequent products on the roadmap to manifest in a circa-2008 fashion, where Intel came back to take the performance lead with authority with the Core series line of CPUs, has become exhausting. Delays for new product releases are frustrating. Rebranding seven nanometer fabrication using a disingenuous schema further masks the issues that the company will face over the next few years. CPU power draw for higher-end products in the 13th generation can match or exceed the consumption of a GPU. The introduction of efficiency cores and the benefits of their performance-per-watt should be given a greater level of focus. The tandem of hardware and software engineering within a company as experienced and as large as Intel could produce something that’s competitive with Apple’s system-on-chip solutions in the M1 and M2 product lines from both the performance and power consumption perspective. A formal review is coming on a Beelink EQ12 which will highlight what Intel can achieve near the bottom of their product stack. While it’s not a benchmark-dominating solution by any stretch of the imagination, this low-powered chip punches above the performance implied by its spec sheet.
With hyperscalers and cloud providers opting to extend the life of existing hardware, hardware refreshes will occur at muted cadences when compared to the past few years. ARM-based offerings are available across all Tier 1 cloud providers at a lower cost than x86 alternatives. If customers can achieve a better return on investment using competitive alternatives, Intel’s foundry play will be dependent upon wrestling the fabrication lead from TSMC. Samsung remains as a viable alternative and is a fast follower of TSMC. “Intel 4” is still two nodes behind the competition. Focusing on performance-per-watt rather than force-feeding 200W+ to achieve a competitive performance target would be ideal. While the research for tackling insanely high-powered workloads is admirable, the need for such a solution (if it manifests) would greatly reduce the total addressable market for such high-performance products.