Components > Handset Components Blog

NVIDIA-Arm: Now What?

by Sravan Kundojjala | 2月 08, 2022

The $60+ billion mega semiconductor deal fell apart after 18 months of its original announcement. NVIDIA embarked on seemingly impossible task in September 2020 and gave itself 18 months to close the Arm acquisition. However, slightly ahead of its 18-month deadline, NVIDIA dropped its pursuit. Despite gaining some support from industry players such as Broadcom, Marvell and MediaTek, lukewarm response from regulators (CMA, EU, FTC and China) and vehement opposition from Qualcomm, Intel and Microsoft sealed the fate of this mega-deal well before the deadline. Unsurprisingly, SoftBank and NVIDIA decided to part ways.

Now what?

NVIDIA will be fine with or without Arm. The company has established an unassailable lead in data centre AI by leveraging its GPU technology. NVIDIA is planning to enter the CPU market to complement its data center GPU business with the help of an Arm CPU license. NVIDIA appears fully committed to Arm with or without the deal. As part of its deal sign-on, NVIDIA prepaid $2 billion to SoftBank, of which $640 m ($170 million for licensing and $470 million royalties) allocated for Arm license and royalty for up to 20 years.

Will NVIDIA execute its Arm server CPU Grace roadmap without Arm? We don’t see any big issue as NVIDIA can still work closely with Arm like other licensees like Amazon and Ampere. In addition, as one of the lead licensees, NVIDIA can get early access to Arm’s IP. The question is can NVIDIA build an entire server CPU ecosystem on its own? We think it will take at least 5-10 years for NVIDIA to establish its Arm CPUs. Currently, less than 15% of NVIDIA revenue comes from Arm-based chips.

SoftBank will miss NVIDIA’s share bounty, though. SoftBank hoped to get $12 billion in cash and up to an 8% stake in NVIDIA by selling Arm to NVIDIA. SoftBank originally spent $32 billion to acquire Arm, and the NVIDIA deal would have given it a $30+ b return on that investment. Today, SoftBank indicated that it will take Arm to IPO within FY 2023 (ends in March 2023).

With semiconductor companies posting all-time high revenues and profits, Arm has no better time to go public. The semiconductor shortage has precipitated strong demand and, in turn, massive CapEx investments across the industry. Moreover, the demand outlook looks good, with many companies still reporting huge backlogs for 2023.

Between 2015 and 2019, Arm’s revenue grew just 5 percent CAGR. After that, however, Arm’s revenue started accelerating again with increased semiconductor demand across smartphones, automotive, embedded etc., from 2020. Between 2019 and 2021, Arm’s revenue grew 16.3 percent. In CY 2021, Arm’s revenue grew 16.5% to $2.5 billion. Armv9 revenue kicked off from late 2021 and will contribute materially in 2022.

Arm has become an R&D intensive asset from the last six years, affecting its profitability. As an IP company, Arm invests ahead of revenue. Under SoftBank, Arm’s R&D spent outpaced that of revenue. This trend will continue for the next five years as Arm invests in data center CPUs to catch-up with x86.

The critical issue that Arm contends is who will fund its increased R&D?

The need for increased funding was the rationale that Arm cited when SoftBank acquired it in 2016 and when NVIDIA announced its intention to acquire Arm in 2020. Now that NVIDIA is out of the picture and SoftBank wants to exit with an IPO, who will fund increased R&D?

In its recent regulatory filing, Arm highlighted the competition from architecture licensees such as Apple and Qualcomm, who keep proprietary designs to themselves without contributing to the broader ecosystem. Arm thinks its standard off-the-shelf licenses may not compete with architectural licensees, forcing Arm to spend more on R&D to level up with architectural licensees whose designs are far more advanced than Arm’s. In addition, the performance gap between the Arm’s standard and custom cores continues to widen, further putting pressure on Arm. Arm also contends with increased competition from RISC-V (Intel, Imagination, Si-Five etc.) and IFS (Intel’s foundry ecosystem).

The good news for Arm is it takes years to recreate a new ecosystem and all of Arm’s leading licensees are still committed to it. Arm gets 80% of its revenue from top-20 customers despite having 150+ active licensees. Qualcomm indicated that it is ready to take a minor stake in Arm if it goes IPO. Arm’s customers posted massive revenue growth in the last two years and are on track to post another record year in 2022.

Price increases remain a practical option for Arm to fund its R&D. Currently, Arm gets 4-5 cents royalty per chip and 20 cents or more on advanced Armv8/v9 designs. However, with increasing competition, can Arm afford to raise prices? Arm is highly dependent on smartphone and embedded markets. If the company cannot spend R&D on data center, it may have to cut down on some of those projects and just focus on the core mobile market but that would limit its growth.

As a public company, Arm will have to endure a difficult transformation and the company’s leadership needs to make sure that it is focused on the future.

Follow me on Twitter or connect with me on LinkedIn. You can share your thoughts and comments directly if you would like to.

Sravan Kundojjala
@SKundojjala

Previous Post: Qualcomm fast forwards analyst day target to 2022 | Next Post: Intel's IFS Holds Promise but Challenges Galore

Let's talk

Now you know a little about us, get in touch and tell us what your business problem is.
Name:
Email:
Telephone:
Country:
Inquiry / Message:

please enter captcha from left