• New processor chip design platforms from Arm will strengthen and deepen Arm’s importance within the data center technology industry.
• These developments increase the competitive and other pressures currently facing Intel, and could also jeopardize NVIDIA’s proposed acquisition of Arm.
Arm’s latest chip design platforms will strengthen and deepen the UK’s firm’s importance within the data center technology industry. This will increase the competitive and other pressures currently facing Intel, which are chipping away at Intel’s dominance in the data center processor space. It could also jeopardize NVIDIA’s proposed $40 billion (£29 billion) acquisition of Arm, which is currently under investigation by the UK government.
Arm has announced two new chip design platforms, the Neoverse N2 and the Neoverse V1, along with a new mesh interconnect to support them. The N2 is the latest version of Arm’s general compute platform and will target use cases ranging from hyperscale data centers to edge computing environments.
Arm does not sell or produce data center processing chips itself, but instead licenses and advances Arm chip designs among chip manufacturers that include NVIDIA, Qualcomm, Ampere Computing, and Fujitsu.
Recently, Arm has also made inroads into the market for hyperscale data centers where its N1 core is used to power AWS Graviton2 processors within AWS EC2 instances. Arm has also said that Oracle will use Arm-based chips made by Ampere Computing to support its expanding cloud business, and that Alibaba will offer Arm-based cloud computing services, although it has not disclosed the chip manufacturer.
Arm’s new V1 platform on the other hand is a completely new product that will support HPC and machine learning use cases. The market for HPC is heating up, with Arm’s platform infrastructure already being used within supercomputers from Japan to India to Switzerland to the United States. This is a market where Arm has spotted clear future growth opportunities.
Arm’s increasing focus on the data center industry has amplified the troubles facing Intel, which last week reported a 20% decline in sales and a 63.5% fall in operating profit for its data center business in Q1 2021. Although Intel still commands a dominant position in the data center processor market, it faces intense competition from its primary CPU rival AMD, while also losing business from hyperscale data center operators that have switched to building their own processors. In addition, Intel is seeing increased competition from GPU specialist NVIDIA, which recently unveiled a new Arm-based data center CPU for giant-scale HPC and AI use cases. The first customers will include the Swiss National Supercomputing Center and the Los Alamos National Laboratory in the U.S.
Meanwhile, as more organizations leverage Arm-based processors, and as Arm targets a broader range of data center compute use cases, this could attract even greater scrutiny of its planned takeover by NVIDIA. NVIDIA has also set its sights on capturing new opportunities in the data center industry, especially HPC ones, and sees its acquisition of Arm as a key milestone that will help it realise those ambitions. However, for the deal to go ahead it must also be approved by national regulators. Last week, the UK government launched an investigation into the proposed takeover, citing concerns that it could undermine UK national security. Arm’s technology is used in critical infrastructure around the UK, including national defense applications. The UK government is also being lobbied by those who believe the takeover could destroy Arm’s role as the Switzerland of the semiconductor market that the UK should focus more on technological sovereignty – especially with current chip shortages – or that NVIDIA could gain an unfair commercial advantage by limiting competitor access to Arm technology.
While these concerns have been expressed in the UK, NVIDIA’s acquisition of Arm also requires approval from regulators in the U.S., the European Union, and China. As Arm technologies look set to play an even greater role in big data, HPC, supercomputer, and hyperscale data center environments – the regulatory approval process will be protracted at best and could even see moves to stop it.