SoftBank & Arm – All the stops pt. II

More like Qualcomm than Nvidia. 

• SoftBank has made public the F-1 for the IPO or Arm which is a fair appraisal of the company, but it does not read like the filing of an AI company which may affect SoftBank’s ability to convince the market to pay $64bn for the shares.
• Highlights of the filing include:
  • First, Financials where Arm has largely returned to the form it enjoyed prior to its acquisition by SoftBank, but it is suffering from the downturn in the smartphone market.
  • This, combined with the difficulties in the Chinese market, caused FY 2024 (March 2023) revenues to decline by 0.8% which has continued into FY 2024 where revenues declined by 3.4% in FQ1 2024 (June 2024).
  • This is not as bad as SoftBank’s FQ1 24 results had led us to believe, but it is still going to be an issue when investors are being asked to pay a 2023 and 2024 EV/Sales multiple of 23x on declining revenues even if the declines are temporary.
  • Second, AI: where the F-1 mentions AI 47 times compared to Nvidia’s 2022 10-K which mentions it 79 times and Qualcomm’s latest earnings transcript which mentions it 28 times.
  • As it stands today, it is difficult to make the case that Arm is an AI company in the same way that Nvidia is an AI company.
  • There is certainly a long-term case to be made as Arm is in an excellent position to capitalise on the opportunity of running AI models on smartphones and other devices.
  • The fact that Nvidia is using Arm IP in some of its AI training products for the cloud is also a plus, but these opportunities are not meaningfully impacting revenues now like they are at Nvidia.
  • Furthermore, the debate over which processor architecture (CPU, GPU, or other) will be best for executing inference on edge devices is far from over.
  • Arm states in its risk warnings in the F-1 (bottom page 35) that if CPUs are not used, then its revenues could suffer if it fails to develop new technologies for its products.
  • This document does not scream AI-company and as such, everything will depend on the meetings that Arm and SoftBank conduct with investors to convince them that an “AI-like” valuation is warranted.
  • Third, Business model & litigation: where substantial question marks remain.
  • These questions spring from the tit-far-tat filings that Arm and Qualcomm have made in their ongoing dispute over the license under which Qualcomm’s Nuvia subsidiary operates.
  • This is the business that Qualcomm acquired in order to return to designing its own processors rather than using Arm’s off-the-shelf designs.
  • Eventually, I suspect that all of Qualcomm’s products will migrate to these new processor designs but the first iterations will be for Windows-on-Arm where Qualcomm intends to match the performance of Apple’s M-series and compete against Intel and AMD.
  • This has no real financial implications in the short-term as there are no products in the market, but seeing as Qualcomm is one of Arm’s largest customers, this has substantial ramifications in the 5-to-10-year time horizon.
  • Hence, while the uncertainty of litigation is never a good idea when conducting an IPO, the immediate-term financial impact looks to be limited.
  • However, Qualcomm has also asserted that Arm intends to completely change its business model from licensing chipset makers to licensing end-device manufacturers in the same way that Qualcomm does.
  • This would completely upend Arm’s business model and greatly increase its revenues and profits, but Arm would have to cancel all of its agreements with chipset makers and move them to device makers instead.
  • Arm has emphatically denied all of these allegations in its response to Qualcomm’s filing, but it does not rule out a business model change in its F-1 (page 39).
  • However, these are risk warnings rather than strategic statements and the example that Arm gives in this warning is a small change in the way that current customers are licensed and is not a change from one customer type to another.
  • Hence, I think it is pretty safe to say that this is not an admission of an impending change to its business model.
  • As a result, I suspect that Arm will stick to its emphatic denial of these allegations during the marketing for this IPO meaning that this uncertainty will remain unresolved.
  • Fourth, China where, like Qualcomm, its exposure is substantial and where revenue growth looks likely to be negatively impacted for at least a few more quarters.
  • Furthermore, the debacle around Arm China still highlights some of the risks to revenues from China as Arm has no control over how Arm China operates.
  • Consequently, I think that Arm’s risk profile in China is as great if not greater than Qualcomm’s especially when it comes to geopolitical sensitivities around the transfer of intellectual property into China.
• The net result is that this F-1 describes the profile of a company that looks more like Qualcomm than Nvidia, which is going to be a problem when it comes to valuation at IPO.
• This is because to achieve a valuation of $64bn, the market needs to put multiples on Arm’s earnings that are more than double those of Nvidia.
• Nvidia is currently trading on a January 2024 PER multiple of 59x but at $64bn the market would be paying 122x March 2023 PER which given FQ1 24 and current market dynamics may be similar for March 2024.
• This is a stretch by any measure that I can conceive of meaning that SoftBank really needs the current AI FOMO frenzy to achieve this valuation.
• Consequently, this is the line that I expect that SoftBank will take when it comes to marketing this investment to investors.
• If this strategy works, then I expect that this IPO will get away quite easily which will be helped by the fact that most of the street is included in the deal and will not be asking too many hard questions.
• Failure will mean a much more modest valuation which I think for the long-term performance of Arm and its reputation in the market, is probably a better outcome.