Analysis regarding the rumors about Marvell Technology kicking the tires of Intel’s Altera and Open RAN groups

Disclaimer: EJL Wireless Research LLC does not own any shares of any publicly traded company mentioned in this article/post and IS NOT advising the purchase/sale of any financial instruments related to any company.

Background

After Intel’s Q2 financial results and plans for a major restructuring including a 15k+ RIF were announced, the rumor mill started around which product groups were underperforming and not aligned with the company’s focus on the Foundry business, the new advanced process nodes, as well as the PC/Server AI market with the launch and ramp of Lunar Lake, Panther Lake, Clearwater Forest and Gaudi 3 over the next 12 months.

There was some mention of Granite Rapids on the conference call but clearly, AI is the focus going forward for the company and NOT telecom.

The Altera product group has been struggling since Q2 2023 with Q2 2024 revenues less than 50% of prior year revenues at USD $361 million, coupled with negative operating income and operating margins. One positive note is that Q2 2024 revenues were sequentially higher than Q1 2024 revenues so it may be that the bottom has already been reached. Intel has already structured the Altera business for an IPO spinoff but with no exact date and timing for this event.

Open RAN Group

So what would be the reasons for Marvell Technology even wanting to entertain the Open RAN group assets (with its x86 architecture for look aside front haul acceleration) if Intel is looking to jettison it? Firstly, if Intel is looking to exit the Open RAN market given the dismal financial performance of this group, would be mean the end of Open RAN like it was for WiMAX? No but it would set the ecosystem back by at least one year if not longer which would force Vodafone and its Europe Big 4 Open RAN partners (Deutsche Telekom, Orange, Telefonica) to rethink its commitment behind the network architecture and the resources needed to continue to back this strategy.

Marvell Technology’s Open RAN solution is based on the Octeon 10 ARM architecture but it was unable to break the Intel juggernaut and subsequently pulled back its focus and conceding to Intel. Buying the Intel Open RAN solution which is x86-based would complicate the manufacturing of the product and require Marvell Technology to enter a supply agreement with Intel for the chips. We do not believe there would be much financial upside for doing this, even though Marvell Technology would immediately gain market share in Open RAN. It would be much better for Marvell Technology to re-pitch its Octeon 10 inline solution to Open RAN customers than to acquire the headache of the Intel group. Additionally, how much would Intel want for this group and how much would Marvell Technology being willing to pay?

Altera Business Group

The Altera FPGA business is a much different product portfolio and business than the Open RAN product group. The end market applications for FPGAs include wireless infrastructure, defense, data center, as well as other sectors. Altera’s main wireless base station customer has been Ericsson for the baseband/DU fronthaul functions. However, the evolution of the Ericsson baseband and DU has integrated the fronthaul FPGA functions into an SoC that is essentially 100% supplied by Intel. While spinning off the FPGA business does not necessarily impact Ericsson directly, control by Marvell would allow for increased revenues to Ericsson for a portion of the baseband SoC and the potential for future conversion to ASIC business from Ericsson. It would also allow for R&D insights from next generation wireless systems for 5G NR Advanced and 6G as initial prototype designs will likely be using advanced FPGAs until product volumes and standards stabilize and would potentially provide a path towards suppling ASICs to this customers.

Acquiring Altera would allow Marvell to create a design pipeline from FPGAs to ASICs with its customers and could allow for integration of FPGAs into SoCs for greater revenue generation.

The main issue is that the Agilex FPGAs are manufactured using Intel’s 10nm SuperFin, Intel 7, and other process nodes. This would require some long term supply agreement for Marvell via the potential acquisition and how would this impact the future R&D roadmap for the Alter product portfolio is also uncertain.

In summary, it doesn’t make financial sense for the Open RAN group but does make financial and strategic sense for the Altera FPGA business. Marvell Technology only had USD $809 million in cash at the end of FY Q2 2024 so any transaction would require financing via a 100% stock transaction or debt and stock for Altera.