Why Smartphone Makers Struggle to Build Their Own Processors

Source | Reposted with permission from WeChat public account: 差评 (chaping321)

Author | 差评君

A few days ago, OPPO released its flagship phone Find X5 Pro.

In the review article at that time I told everyone that OPPO placed a self‑developed chip in the phone.

Yes, it is the independently developed imaging chip that OPPO bragged about — Marianas X.

Phone manufacturers making chips is not new; Huawei's HiSilicon Kirin, Samsung's Exynos, etc.

Even Xiaomi once produced a Pengpai S1 — although it ultimately failed.

Those manufacturers built proper phone processors, essentially the central brain of the phone.

Recently, however, we see OPPO's Marianas X, another company's V1, and Xiaomi's revived Pengpai C1.

They are all just imaging chips — single‑function units that sit beside the main processor to assist image performance.

Why are so many companies suddenly flocking to make imaging chips?

Where did the courage to build full processors, like Huawei and Xiaomi once had, go?

In fact, OPPO, other manufacturers, and Xiaomi ultimately want to make phone processors.

Processors are not something you can just think up.

Because they are not only hard to make, but also burn huge amounts of money.

>/ Want to make processors, but can’t.

Modern processors are more like complete System‑on‑Chip (SoC) solutions.

Take Apple’s A15 as an example: besides the CPU, it includes a GPU, DSP, ISP, wireless modem, NPU, video codec, system cache, and many other components.

Each part’s design is far from simple; consider the modem as an example.

Apple has long touted a closed ecosystem and even produced the M1, yet it still cannot create a usable 5G communication modem.

After forcing a two‑year use of Intel’s modem, which performed poorly, Apple finally had to re‑partner with Qualcomm and embed a Qualcomm modem in the iPhone.

Having the ability to design these functional modules is only one side of the coin.

Manufacturers also need the ability to integrate these modules together.

Combining functions sounds like building blocks, but the challenges are huge.

How to connect these parts, balance power consumption and performance?

How to choose appropriate data paths and make hardware layout compatible with software design?

How to control leakage and reduce static power at nanometer scales?

How to use new etching techniques for process optimization and handle competition and noise in high‑speed signals?

This “building‑block” problem is extremely difficult.

Can manufacturers solve these issues? Some can, but only a few do it well.

Apple, Qualcomm, Samsung have each presented their own solutions.

New entrants that cannot keep up with the pace will be harshly criticized, while established players benefit from decades of experience that are hard to surpass.

Huawei began investing in 2009, spent tens of billions over ten years, and finally could compete with the likes of Apple and Qualcomm.

Xiaomi’s Pengpai S1 suffered from high heat and low performance, supporting only 4G.

In later years Xiaomi did not launch a S2, turning instead to imaging chip C1 and charging chip P1.

What exactly happened is unknown, but it hints at the arduous path of chip development.

>/ Designing chips requires money!

Many aspiring chip makers cannot even reach the processor design stage.

After defining technical requirements and assembling a team, the problem returns to startup capital.

Chip development is not something a garage with a few people can accomplish with sand.

Today, entering chip making requires a team of hundreds; otherwise, you cannot even announce it.

Large projects also entail huge salary costs for the team.

Beyond human resources, phone processors must pay ARM for architecture and IP core licensing.

In addition to visible costs, there are hidden expenses that are even more concerning.

Just as software inevitably contains bugs that require repeated debugging, chip design cannot be completed in one go; countless functional tests are needed.

Some functional tests can be performed via software simulation, called simulation .

Other tests require real silicon small‑batch production, known as tape‑out (flow).

Tape‑out is far from free.

In a 14 nm process, a single tape‑out costs tens of millions of yuan.

In the 5 nm process popular in 2021, a tape‑out can reach roughly 300 million RMB.

How many tape‑outs does a chip need? It varies; a single successful tape‑out may suffice, or multiple iterations may be required.

So how much does designing a phone processor actually cost?

For example, HiSilicon’s Kirin invested 480 billion yuan over ten years, with 1 317 billion yuan spent in 2019 alone.

OPPO announced a 500 billion yuan investment for chip R&D, spread over three years.

Complex design and insufficient funding make it hard for manufacturers to create their own processors; many resort to experimenting with imaging chips instead.

>/ A small step in chip making, a big step in R&D

Although current investments are insufficient to recreate a Huawei‑like miracle, they at least show that manufacturers have woken up.

Looking three to four years ahead, manufacturers are eager to add every possible feature—flip cameras, under‑screen cameras, under‑screen fingerprint sensors, etc.

However, after a few years, consumers are not convinced, and sales remain modest.

The best‑selling phones are still those with self‑developed technology, like Huawei.

After Huawei’s sanctions, manufacturers shifted to a strategy of quickly adopting the latest Qualcomm processors or Samsung screens, resulting in only minor differences such as back‑cover color or camera layout, followed by price competition.

After many years of effort, manufacturers realize that to build a moat and become No. 1, they must have self‑developed technology and unique features.

If a 500 billion yuan imaging chip succeeds today, it could pave the way for another Kirin‑like processor in the future.

Such a scenario is not impossible.

Authors: 小陈 (writer), 面线 (editor), 焕妍 (designer)

Image and data sources: Marisilicon X official page; Yuan Chuan Research Institute, Mate 40 retrospective; Wu Yunfei, why we need a hard‑tech investment boom; TSMC 7 nm EUV chip first tape‑out success; various internet sources.