A Shot in the Arm for RISC-V?

Intel eyes SiFive acquisition, hedges on Nvidia–Arm deal

Last week, Bloomberg reported that Intel is exploring an acquisition of SiFive for more than $2 billion, effectively quadrupling SiFive’s last valuation at a funding round in 2020. SiFive is a fabless start-up providing commercial processor designs based on the RISC-V open-standard instructure set architecture, as well as intellectual property (IP) and silicon solutions.

The pending acquisition of Arm by Nvidia has drawn sharp criticism from all corners of the industry, owing to fears that it would squash Arm’s independence and place existing licensees in the precarious situation of a rival owning a central piece of their technologies. Tactically, acquiring SiFive would allow Intel to gain RISC-V IP and silicon solutions to compete more effectively in edge computing and artificial intelligence applications as customers may turn away from Arm, as well as gaining a direct stake in an instruction set that will prove valuable down the line.

RISC-V’s academic origins came from a need for a practical platform that’s open-sourced and deployable in any hardware or software designs — without incurring royalty liabilities and licensing fees. As an open standard, RISC-V is used in wide array of applications, being particularly useful for embedded low-power solutions. Interest in RISC-V has soared since September 2020 when Nvidia announced its intent to acquire Arm from SoftBank for $40 billion (see Arm and Nvidia. Heads or Tails).

SiFive was founded in 2015 by three researchers from the University of California, Berkeley. Its DesignShare platform allows hardware-makers to incorporate IP into prototypes without having to pay royalties straightaway. This freedom gives customers the opportunity to explore multiple designs before committing to a production version that can then be licensed. This business model is a stark difference to the current Arm licensing deals that require customers to pay upfront to access the company’s chip designs. Indeed, the growing interest in RISC-V has prompted Arm to respond by making changes to its licensing model (see Arm’s New Programme for Custom Chips).

Shortly after the news of the potential acquisition broke, SiFive said that Intel’s Foundry Services selected SiFive’s Performance P550 core as a development platform to be fabricated on Intel’s 7 nm process technology. This platform, codenamed Horse Creek, joins announcements from Tenstorrent, Samsung and Renesas Electronics, which all recently picked SiFive IP to be used in upcoming products.

After Nvidia’s proposed acquisition was announced, we commented that such a move could significantly damage Arm’s ecosystem and create another catalyst to the growth of RISC-V. Intel clearly sees a similar scenario, and has set its sights on attracting Arm licensees that fear the prospect of Nvidia ownership. The move would also give Intel access to an instruction set suited to low-power applications that are growing in industrial and edge computing areas, both of which are vital sources of growth and differentiation for Intel.

RISC-V is a long way from being an alternative to x86 or even Arm designs for data centre chips, but we believe it will play a large role in the explosive growth of edge computing. Nvidia’s announcement of its new Arm-based, Grace CPU for data centre silicon reignited the debate about why Nvidia needs to acquire Arm rather than just license the IP. Nvidia argues it will strengthen the Arm ecosystem by increasing investment and adding to the IP available to Arm licensees. This is a fair and much-needed pledge, but the commercial reality of offering shareholders a return on their $40 billion investment is hard to ignore. This increases competitive concerns that Nvidia will selectively control access to crucial technologies, gaining an advantage over licensees.

With Nvidia’s proposed acquisition of Arm still under regulatory review, the news of Intel’s interest in SiFive will raise eyebrows. Should it materialize into a formal approach, many will view it as an opportunistic move to embrace an architecture that stands to benefit if licensees accelerate their search for alternatives to Arm.

Qualcomm president Cristiano Amon recently suggested that, if the Nvidia bid doesn’t go ahead, Qualcomm and others could invest under a consortium model in which they work together to preserve the long-term independence of Arm. This is undoubtedly the better option for Arm’s independence, but the reality remains that only Nvidia is prepared to pay SoftBank more than $40 billion. If Nvidia navigates the political and regulatory obstacles ahead, Intel’s move for SiFive could be a well-calculated gamble that will be a shot in the arm for RISC-V.