Intel (INTC 2.96%) is still the leader in the data center CPU (central processing unit) market, but it has undoubtedly fallen upon tough times. A severe delay getting its innovative Sapphire Rapids chips to market opened the door for rival Advanced Micro Devices (AMD -0.18%) to overtake Intel in terms of performance. AMD's EPYC Genoa chips are a home run for the company, and market share will likely shift toward AMD this year.

Beyond Intel's delays, the company is at a disadvantage in terms of core counts. Sapphire Rapids tops out at 60 cores, while AMD's Genoa goes up to 96 cores. AMD is also planning a cloud-optimized server chip with a whopping 128 cores, set to arrive sometime in the first half of this year. For workloads where core counts and core density matter, AMD simply can't be beat.

Intel ups the core-count ante

2024 is going to be an important year for Intel's server CPU business. The semiconductor giant expects to launch Emerald Rapids, the successor to Sapphire Rapids, later this year, followed by Granite Rapids sometime in 2024.

While those product lines will focus on raw performance, Intel is creating a brand-new family of chips aimed at core density and cloud workloads. Sierra Forest, set to arrive in the first half of 2024, will be powered by Intel's efficiency-minded E-cores. These cores are smaller, less powerful, and lack some of the features of Intel's standard P-cores but allow for much higher densities.

Sierra Forest will top out at 144 cores. Assuming AMD's plans don't change, Intel will regain the core-count advantage when Sierra Forest launches in 2024. The company will then follow up with Clearwater Forest sometime in 2025, which will be built on Intel's upcoming 18A process node.

In the PC CPU market, Intel's inclusion of E-cores is a big reason the company was able to stage a comeback with its Alder Lake and Raptor Lake chips. Intel's newer PC CPUs generally feature a mix of P-cores and E-cores, which deliver a potent combination of strong single-thread performance and exceptional multithreaded performance.

Intel is taking a slightly different approach in the server CPU market. The company's server chips won't mix the types of cores. Instead, the main lineup will feature only P-cores and will be aimed at workloads like artificial intelligence (AI) that need raw power, while the secondary lineup will feature only E-cores and be aimed at cloud customers who benefit from as many cores as possible.

No more delays (hopefully)

Intel's biggest issue in recent years has been chronic delays. Manufacturing issues have enabled AMD, which uses TSMC for manufacturing, to steal market share in both the PC and server CPU markets. Sapphire Rapids will, hopefully, mark the end of this delay-ridden era for Intel.

By all indications, Intel's plan to launch five new process nodes in four years, an unprecedented pace, is on track. This type of pace will be necessary for Intel to catch up to TSMC in terms of manufacturing and also help Intel's burgeoning foundry business. If Intel can successfully follow through, any disadvantage to AMD should evaporate by 2024 or 2025.

Unfortunately, this aggressive roadmap is aligned with the worst PC market in decades and a period of sluggish demand for server chips. Intel has been forced to slash costs and cut its dividend as it grapples with significant losses. It will be a delicate dance getting its new process nodes up and running while keeping a lid on costs.

Intel's comeback in the server-chip market will take a couple of years to fully play out, but its plan looks like a winner. If Intel can execute, it could erase AMD's advantages and reclaim its place at the top of the server CPU market.