How should mobile phone manufacturers make a decision when self-developed chips become "gold swallowing beasts"?

The sudden shutdown of ZEKU, a chip design company under OPPO, closed the door for mobile phone companies to develop their own chips.

Currently, there are two main paths for mobile phone companies to independently develop chips. One is that chip design work is entirely dependent on oneself; The second is to only define and partially design by oneself, and entrust the rest to partners for collaborative completion.

The former is typically Apple, Huawei HiSilicon, Samsung, and the recently announced termination of Zheku, while the latter is mainly vivo.

Apple's self-developed A, M, and other series of chips have made its terminal products a great success, even in such a cold smart terminal consumer market, Apple has repeatedly become the "contrarian". This has become one of the key incentives for many mobile phone manufacturers to decide on their own chip development capabilities.

But chips are also gold swallowing beasts, as strong as TSMC's major customer Apple and global chip company Top1 Samsung, and to this day, they have not fully completed a fully commercial SoC chip. At least so far, Apple is still outsourcing baseband chips from Qualcomm, which are part of the SoC. Market news shows that Apple is about to make new progress.

After the confirmation of the dissolution of Zheku, some practitioners in the chip industry lamented that this is a question of whether the division of labor in the industry is more or less. Perhaps the mobile phone industry needs to seriously consider what to do with self-developed chips.

Self research, even giants cannot afford it

Many people may not know that Samsung also has its own SoC called Exynos (Orion). But Samsung's true adoption of more SoCs still comes from Qualcomm.

Although Apple has its own self-developed chip A-series, this path is also bumpy. A complete SoC includes modems (modems, also known as baseband chips) and application processors (APs). Apple has been outsourcing baseband chips from Qualcomm throughout its history and integrating them into its own SoC. Essentially, baseband is a work that requires too much long-term sedimentation, but this time can be ten years.

Previously, a senior industry insider expressed to 21st Century Economic Report reporters that they were not optimistic about mobile phone manufacturers developing their own modems. Because this requires certification to the global market and different technology types in sequence, which is a capability that Qualcomm and Unisplendor have accumulated in their historical development, but mobile phone manufacturers need to start from scratch.

In history, when Apple had some disagreements with Qualcomm, it once turned to Intel to purchase baseband chips, but it also caused some models to show obvious signs of poor performance. Later, Apple turned to Qualcomm to purchase baseband chips, and later announced the acquisition of Intel's baseband division.

Market information shows that by the end of 2023 or 2024, Apple may no longer purchase baseband chips from Qualcomm and can achieve self development. Qualcomm is also prepared for this. But it should be noted that this means that Apple's self-developed chips were actually not as complete before.

There is no other reason, it is too difficult to develop a complete SoC chip by oneself.

The Orion constellation under Samsung has also experienced ups and downs. As early as around 2011, Samsung had launched its Orion processor and subsequently collaborated with the launch of its self-developed CPU architecture, the "Mongoose". However, it was ultimately discovered that the energy consumption performance was not as good as the ARM public version, and the "Mongoose" team was disbanded.

In recent years, Samsung has been iterating Orion through various methods and has also collaborated with domestic manufacturer vivo for research and development, but ultimately did not achieve a significant increase in size.

In the analysis of the progress of self-developed chips by domestic mobile phone manufacturers, many people are willing to regard the former OPPO Zheku as the shadow of the next Huawei Hisense, but there are actually significant differences.

Previously, it was reported that Zheku has been working hard on its AP research and development, and is currently the most in-depth and costly domestic mobile phone manufacturer to invest in self-developed chips. As for where the baseband comes from, some analysts believe that they may consider purchasing from MediaTek.

But OPPO and Huawei are vastly different. Huawei is not considered a mobile phone manufacturer, but rather a communication infrastructure provider. Before investing in mobile phones, Huawei had base stations, core networks, optical transmission, and now it also has businesses such as cloud, car, and intellectual property. However, OPPO's business so far is still mobile phones, which means that the two companies' "blood and volume" are completely different in magnitude, and their risk resistance capabilities vary greatly.

Samsung, which is constantly striving for Orion, has much better health than ordinary mobile phone companies. Samsung's businesses include memory, panels, wafer foundry, mobile phones, and more, with more than one business ranking first in the world, naturally having the confidence to constantly try and make mistakes.

Left hand burns money, right hand risks

Mobile phone companies develop their own chips, and the core is of course for personal use. This means that self-developed chips cannot become a universal product for the entire market, highly limiting the company's cost tolerance.

A semiconductor practitioner analyzed the 21st Century Economic Report reporter that mobile phone AP chips produced using the 4nm process cost $40 million per chip, which is extremely expensive and has a high failure rate. It's like a 'golden devouring beast'.

According to the reporter's understanding, OPPO had already incurred a single chip cost of tens of millions of dollars when developing the 6nm Mariana X chip. Fortunately, the technical personnel had a solid foundation and successfully completed the first chip. The second self-developed Mariana Y chip, although successfully taped, has undergone some design adjustments, resulting in a higher cost compared to Mariana X. However, this chip is currently not yet commercially available.

At that time, Jiang Bo, Senior Director of OPPO Chip Products, admitted that if measured solely from a cost perspective, using third-party chip solutions would be more cost-effective. The costs behind this include three aspects: talent acquisition costs, procurement costs for tools such as EDA, and IP cooperation costs.

Due to starting from scratch, Zheku must adopt a high price approach to attract talents, but the complete introduction of external talents can easily lead to internal cultural conflicts.

These are all "money" withdrawals that need to be considered.

The risks to be faced are not only possible and high sunk costs, but also geopolitical factors worth considering.

At present, there is an analysis in the industry that suggests that considering that OPPO's self-developed Mariana X and Mariana Y are both 6nm advanced processes, the AP being developed should also be the most advanced 4nm process, which completely steps on sensitive areas in the geopolitical environment.

As mentioned earlier, there is only OPPO for mobile business, and if the mobile business encounters an accident, there is no other business to help carry it through. Moreover, as a lesson from previous experiences, it is still under pressure today.

From this perspective, the vivo self-developed chip model, which has been relatively lightweight in research and development, may appear relatively stable at present.

Especially, the current mobile phone industry is still particularly cautious about the outlook. Many international chip giants have mentioned in their financial reports that the recovery of China's mobile phone market consumption is weaker than expected, becoming one of the factors affecting performance. Not to mention that the global market is facing more severe economic headwind pressure.

Many chip practitioners understand but regret the termination of Zheku's business, and should respect them if they have tried; Some also believe that it is necessary to respect the industry rules of industrial division of labor and delegate professional tasks to professionals. But no matter what, once Zheku carried his dreams and had to face the harsh reality.