As we all know, the world of semiconductors can be a complex and highly specialized industry that involves intricate cross-technologies and geopolitical considerations. Consequently, it cannot be reduced to simple tags like red and blue OR black and white.
While most people may not be interested in the facts or the truth, they still want to be right.
Zeku’s shut-down touched me deeply, and I have taken the liberty of revising it for everyone’s criticism.
Before Kirin, Huawei’s mobile chips were named by “Mountain Peak.” The application processor was the K3 from Karakunlun, and the baseband processor (or modem) was the lesser-known Balong Snow Mountain. What many people don’t know is that the 3G chip, Meri or Meili (Snow Mountain), failed. Zeku named his chip “Mahriya Trench”. Both meanings suggest the difficulty of this venture, albeit with some directional deviations.
Prologue Link to heading
Sixteen years ago, when Steve Jobs released the first-generation iPhone, he was still uncertain about it. Before the launch, Jobs rehearsed for six days, but the problems persisted: occasionally the iPhone could not make calls or connect to the internet. Furthermore, the baseband chips provided by Infineon Technologies at that time did not support 3G, despite Nokia and Motorola already having 3G phones for 4-5 years.
To catch-up with the summer release schedule of AT&T and secure the contract, Jobs had to release a half-finished iPhone first that was just a music box could make phone calls.
There was no App Store and it couldn’t install software, had only 2G, no GPS, and its first shipment was still six months away.
This hasty release gave Google plenty of time to study and copy the iPhone. The next year, when the first Android phone (G1) was released, it caught up entirely with the iPhone’s progress, providing 3G and an app store, with GPS and navigation, and even replaceable batteries, all with stable signals from Qualcomm’s SoC. Since then, Android’s market share has skyrocketed, leaving the iPhone behind.
The originally thought-to-be-leading-by-at-least-two-years Steve Jobs was furious when he shout : “I will spend my last dying breath if I need to, and I will spend every penny of Apple’s 40 billion in the bank, to right this wrong. I’m going to destroy Android, because it’s a stolen product.”
Looking back, why did Apple choose Infineon as their communication chip supplier, which was not quite leading at that time? Friends who are new to the semiconductor industry may not know how brutal the competition for mobile chips was in the 2G-3G era.
Let us take a closer look.
1.The Battle of Giants Link to heading
During the analog mobile phone (1G) era, Motorola was undoubtedly the leader, with over 70% market share. Back in the day, Motorola’s semiconductor division (later to become Freescale) was a major player, with CPUs even outperforming Intel in some cases.
In an effort to compete with Motorola, European countries partnered together to create the GSM standard, resulting in a flood of mobile phones hitting the market from companies like Nokia, Ericsson, Philips and Alcatel, one big brand a country.
Interestingly, no British mobile phone brands emerged at the time, but ARM, a company once called Acorn, rose to dominance worldwide after.
Apple’s Newton tablet, which was based on ARM, was discontinued by Steve Jobs. Texas Instruments (TI) recognized ARM’s potential and introduced it into Nokia’s business phone, giving ARM a decade time to succeed.
In the US, the semiconductor landscape was more complex with companies like TI, Skyworks, ADI, Agere, Broadcom, Marvell and Qualcomm making mobile phone chips.
However, Motorola stood out as the leading mobile phone manufacturer with its strong chip capabilities. Many chip giants tried to establish customer bases outside of the US, resulting in a fierce battle that ended with many fallings.
This surge in the number of chip manufacturers entering the GSM market highlights the explosive growth of mobile phones during the 2G era, despite the lower technology barrier required to make mobile phone chips.
Unlike today, chip integration was not high, and each design was unique, often composed of multiple chips and components. This made development and debugging extremely time-consuming and difficult, leading to the emergence of independent “Design House” companies that specialized in providing design prototypes or modules directly to mobile phone brands.
Today, 99% of these Design Houses have disappeared and the rest become big mobile phone ODMs.
2. European Players Link to heading
In 1999, Siemens’s semiconductor division spun off to become Infineon. When I work with Infineon, there was a standard in mobile phones and offered the unique benefit of free reimbursement for emplyee’s lost phones.
The quality of Siemens’ mobile phones was truly remarkable - they were practically indestructible.
However, in that era when phones didn’t do much, looks and ringtones took precedence over substance, and slow-moving companies like Siemens began to fall behind.
In 2005, Siemens failed to sell their mobile division to Motorola and ultimately give away it with 350 million euros cash as gift to Taiwanese company BenQ.
Unfortunately, BenQ’s own dreams of creating their own brand were dashed due to the slow of Germans.
In 2005, Infineon Technologies’ struggling wireless division faced difficult times as Siemens, their single large customer, experienced problems with their mobile phone business. However, Infineon’s fortunes were about to change for the better with the introduction of their X-Gold single-chip solution for low-cost phones, which quickly gained industry recognition as a leading solution. Some key players such as Nokia, LG, Samsung, Konka, and ZTE quickly adopted the technology.
Secretly, Steve Jobs, who was developing the iPhone, was also on the lookout for a high-integration, low function phone solution. We’ll come back to that later.
After abandoning Siemens phones, I switched to the Philips 9@9c, which was not only lightweight and attractive, but also had the remarkable feature of being able to last a month without needing to be charged. It was truly a myth in today’s society to not have to bring charging cables when traveling.
Sadly, the Philips brand lasted only a year longer than Siemens before being sold off to CEC.
The chip platform limped along for a few more years before collapsing, but the super-efficient baseband design may have found its way into Huawei through engineers.
The Alcatel phone brand was sold to TCL in 2005. Let’s take a look at the rest European big names:
In 2002, Alcatel mobile phone semiconductor division was sold to STMicro.
In 2006, Philips Semiconductor became NXP.
In 2008, NXP Wireless Division merged with ST and founded ST-NXP Wireless.
In 2009, ST-NXP Wireless merged with Ericsson mobile phone R&D aand founded ST-Ericsson.
In 2013, ST-Ericsson was shut-down.
Meanwhile, Nokia was dominating the market with Texas Instruments as their main partner. However, Nokia eventually became dissatisfied with their bargaining power and started a new multi-supplier strategy in 2007.
Companies like STMicro and Infineon provided low-profit quotes in an attempt to break in. As a result, TI began to detest the business of baseband chips as the rapid updates every year were too fast and the ROI was far inferior to TI’s many industrial chips with decades-long life spans.
In 2008, TI announced its gradual exit from the baseband business and forced Nokia to complete a full platform switch by 2012.
The result was coupled with the sudden invasion of Apple and Android, and Nokia began its long decline in market share from 2008.
Nokia also had a small Modem department which mainly focused on baseband technology research and development rather than chip manufacturing. The end products included USB dongles for Internet use on computers. This thing was much simpler than a mobile phone as it did not require an operating system or screen keyboard. ZTE and Huawei relied on this technology to enter the consumer business and dominated Europe and America with low prices and high quality, laying a solid foundation for future mobile businesses. Nokia’s Modem department was sold to Japan’s Renesas in 2010.
In 2011, Nokia fully adopted Qualcomm’s platform for the Windows Phone series Lumia and marched towards the cliff without looking back.
3. US Players Link to heading
In 1999, Conexant was spun off from Rockwell Automation. At that time, a Conexant wired telephone dial-up modem on a notebook computer was a symbol of high-end technology.
In 2002, Skyworks was spun off from Conexant, focusing on wireless communication, with baseband modem as its main direction. The new Skyworks quickly became a strong platform for TechFaith Wireless and spread to many medium-sized factories in China. However, good times don’t last long, as MediaTek’s turn-key strategy swept China in 2004, Skyworks announced the abandonment of the baseband business. Thereafter, Skyworks focused on RF front-end and became an RF giant along with TriQuint and RF Micro to form Qorvo, relying on Apple, Samsung, and Huawei’s mobile phones.
ADI is a very diligent company that has been quietly cultivating in China. Its mobile phone chips have appeared in many second-tier brands in China but barely managed to survive. After MediaTek achieved great success in the knockoff function phone market, it urgently needed TD technology to enter the mainstream brand 3G market and compete with Spreadtrum, while ADI had TD chips. In 2007, ADI sold its mobile phone department to MediaTek for 350 million dollars, which can be regarded as a win-win deal.
ADI is an atypical American semiconductor company that has not been acquired by a big company and is still thriving. The reason for not being eaten up is simple: founder Ray Stata has been at the helm and did not step down as chairman until he turned 88 last year, yet still sits on the board. This aligns with my other article “the Story of BIOS & PCs.”
Broadcom has always occupied a leading position in Wi-Fi, Bluetooth, and GPS areas, but its baseband manufacturing capabilities are rarely seen and its basebands are often sold as bundled products. But as a communication company with “com” in its name, cutting off this big market of baseband is really unacceptable. Therefore, Broadcom has been persistently supporting the 3G era with its baseband technology. In 2012, Broadcom acquired RISE’s LTE platform in an attempt to make a comeback in the 4G field. However, the development of LTE chips was too expensive and not progressing smoothly, resulting in delays. In 2014, they announced they were giving up on baseband.
Marvell, a Chinese-American found company, has a strong ability to catch trends, and they firmly grasped the market in storage and Wi-Fi. In 2006, Marvell showed their foresight supporting Palm by acquiring Intel’s ARM (XScale) mobile platform just a year before the birth of the iPhone. When TD-SCDMA was not favored by anyone, Marvell launched chips that supported TD, hitting China Mobile. Blackberry was a big customer for Marvell’s mobile platform, but the baseband market became too fierce, causing Marvell’s previous generation smartphone customers to suffer. In 2015, Marvell also had to disband their baseband team. Agere, formerly known as Lucent Technologies, was a traditional player in the voice communication industry, and it was impressive that they managed to survive until the new century.
In 2006, LSI acquired Agere, and in 2007, LSI sold Agere’s mobile baseband to Infineon. It is worth noting that in order to resist antitrust laws, Qualcomm once authorized CDMA IP to LSI, who then sold it to VIA Telecom in 2002. Later, VIA Telecom authorized the IP to MediaTek and Intel, enabling them to avoid paying Qualcomm CDMA royalties and making full network communications possible.
Today’s hot topic Nvidia also went through a rough patch in baseband. Jensen Huang, the founder, has an unparalleled feel for cutting-edge technology. When the smartphone wave arrived, Nvidia naturally did not want to miss out. Tegra chips had captured a lot of business in the early days. However, Huang also underestimated the difficulty of making baseband modems and had to give up in 2015. This was a blow to Huang at the time, as it meant that Nvidia was no longer related to the hottest smartphone market. Perhaps for this reason, Huang decided to acquire ARM several years later. However, avoiding the chaos in the mobile domain may have allowed Nvidia to spread their wings and explore the potential of the GPU in-depth (CUDA/HPC/Crypto/AI…).
4. MediaTek Link to heading
Finally, MediaTek, also known as MTK, acquired BenQ’s wireless phone market failure, Airoha, which had a strong WLAN and RF department, in 2006.
This acquisition gave MediaTek an advantage in 3G while also allowing them to supplement their baseband shortcomings. MTK and Taiwan generally have a unique ability in channel control.
They have trained countless unheard-of smartphone manufacturers and design houses through agents, and also regulate market demand and prices through distributors. With the love of domestic small brands for spot goods and the temptation of large manufacturers to engage in hoarding, MTK’s achievements are difficult for European and American giants to replicate.
Nowadays, there is a rush to develop various chips including silicon carbide, which is similar to the rush to develop design houses and Shan Zhai phones in the past. The threshold is low for small-scale operations, but it is not easy to become bigger and stronger. The desire of this generation to make quick money is far greater than the sense of mission.
During the 4G era, MTK didn’t have an easy time as their product development always lagged behind Qualcomm. At that time, giants like Xiaomi, Huawei Oppo and Vivo hardly used MTK. This shows that basebands are not easy to develop and even experienced companies might trip.
It wasn’t until later when MTK took a gamble with 5G and reversed Qualcomm’s sales with its Dimensity series chips.
5. Finale Link to heading
In the world of modem players, Europe has none, the rest are Samsung in South Korea, Qualcomm in USA, MediaTek in Taiwan, Huawei, Spreadtrum and ZTE in China.
For friends who have made it this far, it shouldn’t be necessary to talk more about Qualcomm and Huawei. Mentioning the excellent comprehensive quality of Hi-Silicon’s Balong modems makes us both proud and regretful.
With the various shuffling of international baseband platforms, more than 90% of Chinese Design Houses, which were once in the limelight, have died out. For instance, CEC Sailong, who collaborated with Philips, lost its competitiveness due to Philips chip problems. While JiaSheng LianQiao, who worked with Infineon, became a well-known runaway company later on.
Design Houses that had early transitions to MediaTek platforms (MTK acquired MStar in 2013) or early transitions to mobile ODM production have emerged from chaotic battles in 2G-3G, such as Wingtech, Loongcheer and Huaqin.
During the time of the Warring States period, several chip companies emerged in China’s TD-SCDMA standard, such as Tianqi, which had investment and technology sharing from Philips and Motorola, Kaiming, which had investment from Nokia and TI, as well as Datang Lianxin. Tianqi was eventually sold to ST-Ericsson, while Kaiming went bankrupt.
The story of TD is estimated to be more than 100 pages long and is essentially a story of burning up billions of dollars, with only Spreadtrum benefiting from basebands in the end.
6. Apple Link to heading
Lastly, let’s return to Apple and Infineon. Infineon is only a cheap backup second or third source of supply with Nokia, Samsung, LG, and other customers. In addition, as the sales of the first three generations of iPhones were not high, Infineon’s wireless department continued to lose money.
When the groundbreaking iPhone 4 was launched, Qualcomm joined as the CDMA2000 baseband provider, even though Infineon was still the main WCDMA supplier. The two companies were compared and Qualcomm dominated in CDMA. Apple had no better choice as Qualcomm held a monopoly in CDMA.
Initially, Steve Jobs chose Infineon for the iPhone because of its high level of single-chip integration, which would keep the board size small. Additionally, Infineon’s low customer base and high dependence on Apple suited the company’s needs as other chip manufacturers showed little interest.
However, even Jobs couldn’t foresee that 3G/4G would develop at such a rapid pace to catch up Wifi, causing Apple to focus on developing its own A-Serie SoC processors and neglecting baseband technology.
Infineon chose to add 3G based on its 2G platform and it turns out to be a bad choice than make a brand new one. Infineon’s slow progress in developing its 3G platform made it difficult for Apple, with signal issues plaguing iPhones running on Infineon baseband. To counter this, Jobs incorporated a stainless steel antenna band around the iPhone 4 to boost signal strength, leading to the infamous “Antenna-gate” controversy.
Despite his disdain for Qualcomm’s expensive patents (5% of the selling price of expensive iPhones), Jobs was forced to switch to Qualcomm from the iPhone 4s as the company’s technology excelled in the upcoming 4G LTE platform.
In 2010, Infineon sold its unprofitable wireless department to Intel for $1.4 billion, which was a great outcome compared to other companies that had to shut down their Baseband departments.
However, the sale meant that Intel lost Apple as its largest customer, leading some tech experts to suggest that Apple should have bought Infineon.
But with Intel as a deep-pocketed parent, Infineon’s Wireless department was able to develop and overcome tens of billiion losses. When Apple reintroduced Intel LTE baseband for the iPhone 7, many criticized its performance compared to Qualcomm, leading Apple to limit Qualcomm’s speed to offset the deficiency of Intel’s chips.
The release of the iPhone Xs marked the end of this struggle as Intel’s XMM7560 baseband modem replaced Qualcomm, fulfilling Jobs’ wish and saving Apple a lot of money in the process.
However, when it came to 5G development, Intel faced another round of delays, causing the iPhone 11 to miss out on 5G, and even the iPhone 12 was at risk of being delayed.
In 2019, Apple had to buy the Infineon baseband team and announced that it would work with Qualcomm again. However, four years had passed, and Apple was still unable to release a proprietary baseband, which was a crucial missing piece for Apple Silicon.
With Apple’s robust finances, where did the problem lie?
Developing analog or mixed-signal chips is not like developing digital chips where you can simply buy IP and throw money and wafer size at it.
It requires a long-term accumulation of know-hows. Coordinating baseband and RF also requires a lot of experience, and 5G RF power amplifiers (PAs) are notoriously difficult to make.
The Intel platform has fewer customers, and therefore fewer chances for trial and error. Furthermore, 5G is an evoving standard and it releases new features almost every year.
The development and testing costs are too high, and if there is only one customer in the end, the cost allocation is clearly not cost-effective.
For example, the most advanced Qualcomm 5G baseband now integrates RF and can reach speeds of up to 10Gbps, not only supporting 2G-4G but also supporting different frequency bands in different countries, including millimeter waves and even satellite communication.
Engineers have to run all over the world to test various base station environments, which is the value of long-term accumulation. This is not like developing software where a bunch of bugs can be quickly converged in the first version. If there are too many bugs in this kind of complex chip, it’s over, as the launch window and super high cost do not allow for repeated tape-out and iteration. And it’s not feasible to rely on software to handle bugs like with those pure digital chips.
Therefore, over the past two to thirty years, the accumulated knowledge or know-hows from 2G to 5G is all valuable, which is very similar to the path of industrial software. Perhaps this is also the reason why there are hardly any new players in the baseband market now.
Looking back, the wave of baseband manufacturers competing for dominance more than ten years ago (chip/design companies (IDH)/ODM, etc.) has cultivated a large number of key talents and laid the foundation for technology and the market, and has also fostered many outstanding mobile phone companies in China. This article only throws out bricks from a narrow perspective and welcomes everyone to review that crazy and burning era together.