It has been reported that AMD has been slowly but steadily increasing its presence in the market for data center CPUs ever since it introduced its EPYC processor for servers back in 2017.
AMD is projected to hold more than 20% of the market share this year. It is anticipated that Arm-based System-on-Chip (SoC) manufacturers will hold around 8% of the market share, while Intel’s share will decrease substantially, although it will remain the market leader.
According to Frank Kung, a DigiTimes Research expert, AMD’s success in gaining market share from Intel is due to a few significant factors. One of these is that AMD’s latest EPYC processors have more cores than Intel’s Xeon Scalable processors, and EPYC CPUs are less expensive than Xeon Scalable CPUs with a comparable number of cores.
Additionally, it is expected that AMD’s latest EPYC CPUs manufactured by TSMC will have better availability than Intel’s Xeon Scalable CPUs, which are produced in-house. These factors are predicted to drive AMD’s market growth while simultaneously reducing Intel’s market share.
Kung has pointed out that pricing is one of the most important drivers behind shifting data center operators and server brands to AMD. The cost of AMD’s server CPUs is substantially lower than Intel’s, with some AMD products costing more than 40% less than their Intel counterparts, even when comparing products with similar specifications.
This significant price difference is particularly attractive to server companies who buy CPUs in large quantities, and there are no compatibility issues with either Intel or AMD processors since both are built on the x86 architecture. It means that servers can operate with either CPU without any concerns about compatibility.
According to Frank Kung, three main factors have led to AMD’s success in the server CPU market. The first factor is the lower prices of AMD CPUs compared to those of Intel, making them a more cost-effective choice for server companies. The second factor is AMD’s support from TSMC, where all their server CPUs are produced using the latest manufacturing process.
Unlike Intel’s in-house manufacturing process, it allows AMD to offer high-performance CPUs without delays in product launches, which has faced issues in recent years, causing delays in the production of its new server platform.
In the latter part of the previous year, AMD introduced its latest Zen 4 microarchitecture-based EPYC ‘Genoa’ processors, specifically created for mainstream servers. This platform will gradually increase in production throughout 2023.
In addition, AMD plans to release its EPYC ‘Bergamo’ CPUs later this year, which will feature up to 128 cores and use the Zen 4c architecture. AMD introduced its EPYC ‘Genoa’ processors based on Zen 4 microarchitecture towards the end of last year for mainstream servers. These processors are expected to be gradually adopted in 2023 and will help AMD maintain its competitive edge over Intel. AMD’s upcoming EPYC ‘Bergamo’ CPUs with up to 128 cores, based on Zen 4c architecture, will be released later this year for cloud data centers, which will further enhance its position in the market. In contrast, Intel’s Sierra Forest CPUs for cloud workloads are not expected to be available until 2024. Additionally, AMD plans to introduce its EPYC ‘Siena’ processors targeting the communications market later this year.
Arm-based SoCs are making progress in the server market, despite the challenge of software compatibility. Although these processors have comparable performance to x86 processors, their inability to run certain programs has impeded their adoption.
However, according to DigiTimes Research, companies like Amazon, Ampere, Alibaba, and Marvell have increased their share of the server CPU market from 3.5% in 2021 to 6.8% in 2022. With the entry of new players, such as Nvidia, and the launch of new products by established players, like Ampere, it is anticipated that Arm-based SoCs will continue to grow and capture 8.1% of the server CPU market.
According to McCarron, the server market experienced a strong growth period in 2021, with the peak of this cycle occurring in January 2022. However, he expects the market to decline significantly in the first quarter of 2023, reaching its lowest point. Even though both Intel and AMD are releasing new processors in 2022, analyst McCarron doesn’t expect the “Osborne Effect” to impact the server market, as the longer purchasing cycles in the industry mean that sales are not likely to drop due to anticipation of new models. Server purchases involve extensive planning, and customers do not typically postpone purchases in anticipation of new products like consumers might.
In contrast to PC demand, which may have inventory equivalent to a quarter or more, server inventory is likely only a few weeks’ worth. McCarron thinks that the server market will rebound after the first quarter, assuming there are no economic worries.
Nonetheless, he anticipates that economic concerns will take precedence over the ongoing competition between Intel and AMD in terms of influencing server buyers’ decisions. He predicts that the next cycle of server purchases may be less robust than previous ones, as purchasing behavior may be more cautious due to economic uncertainty.