Smoke, Mirrors and Silicon
Why Huawei’s 7nm chipset is not a sign of Chinese strength…
The media frenzy surrounding the launch of Huawei’s Mate 60 Pro phone has been nothing short of breathless. The hype, fuelled by claims of “new” ground-breaking 7nm chips, reports this as a dynamic shift in the strategic technological landscape.
Many seasoned observers, even Bloomberg, appear to be swayed by the hype, quickly pronouncing this as the beginning of the end for US sanctions on Chinese technology. Others have interpreted this development as a quantum shift in China’s technological capabilities, inferring hidden depths to Chinese microprocessor manufacturing capabilities, suggesting even that China is now equipped to bypass western sanctions, and further are equipped to potentially overshadow future western advancements in the field.
These analyses are incorrect, and as usual the reality is far more nuanced…
So what exactly is this Chip?
As a new device consumer, there is certainly a lot to be said for the phone — with a triple camera setup offering high resolution imaging and video in wide, periscope and ultrawide. A 5000mAh Li-Po battery capable of fast charging, a 120Hz LTPO OLED high res display amongst many more crowd pleasing enhancements.
However, what got the world interested though, was the fact that the chipset, a novel Kirin 9000S SoC (System-on-a-Chip) consisting of an Octa-core CPU running between a single 2.62 GHz Cortex-A720 to a quad 1.53GHz Cortex-A510, backed by a Maleoon 910 GPU, was manufactured to a 7nm transistor (and other components) size.
This fact is being held as a an abrupt technological advancement. Which is surprising given that there were reports of the Chinese semiconductor giant SMIC (Semiconductor Manufacturing International Corporation — the firm manufacturing the majority of Huawei’s chips) had managed to create 7-nm chips as long ago as Q2 2022.
Even at that time, it was not a particularly innovative process. Samsung and TSMC were creating scaled logic and memory bitcell integrated 7nm chips as far back as early 2016. Rather it represented a thoughtfully executed utilization of accumulated resources and capabilities. Future similar advances are far from assured.
That is not to say that the new chipset is not interesting or lacks merit. On the contrary, the Kirin 9000S SoC showcases several intriguing features — for example a dedicated chip for satellite connectivity. Satellite connectivity in mobile devices, while not a novelty, is mostly made possible through clever workarounds; for instance, iPhone 14 devices send text messages to emergency services via satellite by repurposing the phone’s antennas. However, the Mate 60 Pro’s approach could potentially signify a leap in sophistication.
However, some other aspects perhaps reveal more about the chip’s inefficiencies. For instance, the 7000mm² vapor chamber is considerably larger than those featured in most rival devices, and significantly exceeds the size of its predecessor. This increase suggests a need to offset greater heat output, potentially indicating less-than-optimal energy efficiency. An oversized vapor chamber might be an effective solution for managing thermal issues, but it also underscores the room for improvement in the design and operation of the system-on-chip (SoC).
In conclusion a closer look at Huawei’s announcement reveals a less than impressive reality. The technology isn’t ground-breaking and reflects developments achieved by others years ago.
Okay, but manufacturing 7nm chips is not exactly “easy”
True, the production of 7nm chips is not a simple task. The highly intricate and delicate process involves depositing multiple layers of various materials onto silicon wafers, which are then patterned using advanced photolithography techniques. The smaller the node size, such as 7nm, the more complex the process becomes, necessitating precision equipment, stringent quality control, and an environment free from even the smallest particles of dust. Smaller nanometer values indicate smaller and more densely packed components, which can lead to improved performance, power efficiency, and smaller chip sizes.
The process of semiconductor and other integrated circuit fabrication at its most basic entails three steps:
Wafer Substrate Preparation: prepping silicon wafers, which are typically thin, flat, and round discs made from a single crystal of silicon which are thoroughly cleaned to remove any contaminants and particles that might affect the quality of the final semiconductor devices.
Layer Deposition: Multiple layers of various materials, including silicon dioxide (SiO2), silicon nitride (Si3N4), and conductive materials like polysilicon, are then deposited onto the silicon wafer using techniques like chemical vapor deposition (CVD) or physical vapor deposition (PVD). With each layer serving a specific purpose in the IC’s design, such as insulation or conducting electricity.
Photolithography: Finally after each layer is deposited, a process called photolithography is used to define patterns on the wafer. This process requires the application of a layer of photosensitive material called a photoresist onto the wafer which is then exposed to UV light through a mask (a patterned template). This photoresist is “developed” to selectively remove areas of the photoresist, leaving behind a pattern that corresponds to the desired circuitry. The final circuit can be tweaked by etching or otherwise modifying the underlying layer according to the pattern on the photoresist.
The only part of this process wherein the CHIPS act limited China’s access to vital machinery was in the final process, the Photolithography as all other aspects of this fabrication are well known and widely reproducible.
So why is Photolithography the focus?
To understand why photolithography is the critical focus, it is useful to get a brief overview of lithography techniques, particularly deep ultraviolet (DUV) and extreme ultraviolet (EUV) lithography.
DUV (Deep Ultraviolet) and EUV (Extreme Ultraviolet) are two distinct forms of photolithography technologies utilized in the manufacturing of semiconductors to design patterns on silicon wafers during the fabrication of integrated circuits (ICs). They are differentiated primarily by the wavelength of light applied and the precision they offer.
DUV (Deep Ultraviolet):
- Wavelength: DUV lithography employs light with a longer wavelength in the deep ultraviolet spectrum, typically around 193 nanometers (nm). The specific wavelength is often associated with an argon fluoride (ArF) excimer laser.
- Resolution: DUV lithography is used for older technology nodes (e.g., 90nm, 65nm, 45nm, and larger). Its larger wavelength restricts its ability to create extremely small features on silicon wafers.
- Multiple Patterning: To achieve smaller feature sizes with DUV lithography, a technique called multiple patterning can be used. This involves exposing the wafer multiple times with different masks and carefully aligning them to create the desired patterns.
- Complexity and Cost: Multiple patterning increases process complexity and production costs. It also extends the time required for semiconductor manufacturing.
EUV (Extreme Ultraviolet):
- Wavelength: EUV lithography uses extremely short wavelengths in the extreme ultraviolet spectrum, typically around 13.5 nm. This wavelength is generated using a plasma source and is significantly shorter than the wavelength used in DUV lithography.
- Resolution: EUV lithography allows for much finer resolution, making it possible to create smaller and more intricate features on silicon wafers. It is crucial for advanced technology nodes (e.g., 7nm, 5nm, 3nm) where smaller transistors and denser circuitry are required.
- Simpler Patterning: EUV lithography simplifies the patterning process because it can create finer features with a single exposure, reducing the need for multiple patterning steps.
- Challenges: While EUV lithography offers significant advantages, it also presents technical challenges. EUV light is absorbed by most materials, so specialized optics and masks are needed. Controlling the source and managing the reflective optics are also challenging.
In summary, the key difference between DUV and EUV lithography is the wavelength of light used. EUV lithography uses much shorter wavelengths, enabling it to create smaller and more intricate features on semiconductor wafers. This makes it a critical technology for advancing semiconductor manufacturing processes and achieving smaller technology nodes with higher performance and energy efficiency. However, EUV technology is also more complex and requires sophisticated equipment, which can result in higher manufacturing costs.
So what does that mean for the 7nm chip in Huawei’s new phone?
The relevance lies in the fact that since mid-2019, the sale of EUV machines to Chinese entities has been banned, which at the time marked a significant turn in the global tech trade dynamics. This ban created significant obstacles for China’s ambitions in semiconductor manufacturing, as EUV lithography was and remains the cutting-edge technology that enable production of smaller, more efficient chips.
However, up until the 1st September 2023, just a week ago, DUV machines were still largely available for purchase by Chinese entities from the world premier Photolithography machine manufacturer, the Dutch company ASML.
Why?
When the Netherlands implemented an export control measure at the end of June this year which specifically targeted ASML Holding NV’s chip-making machines, preventing their shipment to China they set the effective date as September 1. Whilst the rule did not explicitly mention China, it stipulated that ASML, the sole manufacturer of highly sophisticated chip-making machines worldwide, would be required to obtain a license before exporting their advanced deep ultraviolet lithography (DUV) systems to China. A license which would likely be refused. The decision made by the Dutch government was widely perceived as a move to align with the United States and Japan.
However, the four month gap between June and September, provided ample time for Chinese chip foundries to import ASML’s latest cutting-edge immersion deep ultraviolet (DUV) lithography system, the TwinScan NXT:2000i — a system reportedly capable of producing chips using the 5-nanometer process.
Furthermore, a simple examination of the figures reveals that Chinese semiconductor manufacturers have clearly been accumulating machinery in preparation for the limitations enacted by the United States and its allies.
According to a report published by Chinese semiconductor industry consultancy JW Insights, citing data from China customs, Chinese imports of lithography machines made in the Netherlands, primarily from ASML, witnessed a remarkable year-on-year growth of 64.8%, amounting to US$2.58 billion from January to July. Notably, ASML had projected that its sales to China for the year would remain stable at approximately 2.2 billion euros (equivalent to US$2.36 billion), accounting for 14% of its total annual revenue. In July alone, China’s imports of lithography machines from the Netherlands reached a value of US$626 million, nearly eight times higher compared to the same month last year.
Given these facts, it really should have come as no surprise that China possesses the capacity to manufacture 7-nanometer chips. Their strategic accrual of equipment from ASML clearly indicate a readiness for advanced chip production. Moreover, as evidenced by the large purchases of the TwinScan NXT:2000i’s — Chinas capability to produce chips with a 5-nanometer probably is already exigent.
Far from being a clandestine stride, this is a meticulously planned response to the anticipated limitations imposed by international sanctions, and it underlines the massive national effort China’s leadership is expending on enhancing China’s semiconductor sector.
The plan
China’s “Made in China 2025” strategic plan, which aims to transform the nation into a global high-tech manufacturing economy, has seen the central government invest heavily in boosting domestic Integrated Circuit (IC) manufacturing capabilities. An initiative not merely an attempting to decrease reliance on foreign technology but a strategic move in expanding their semiconductor industry’s footprint.
Vast amounts of capital have been funnelled into research and development centres, fabrication plants (foundries), and semiconductor companies across the country. According to a report by the Centre for Strategic and International Studies, China’s IC industry grew by 20.9% in 2019, almost twice the global average.
A combination of government funding, tax breaks, and stimulating domestic demand was aimed at creating the worlds most favourable environment for domestic IC manufacturers, as well as trying to reduce reliance on Western technologies, prioritizing the growth of its domestic microprocessor industry.
Whilst there are multiple competing drivers for this policy, cost is a key one, with China spending a staggering US$ 432 billion on imported microprocessors in 2021, which surpassed grain and crude oil imports and in the context of deteriorating US-China relations — it was clear to China’s leadership that safeguards were needed.
The policy has had some success; China’s microprocessor industry has made significant strides in manufacturing high quantities of microprocessors, especially at the 24nm production nodes and beyond. However, producing cutting-edge microprocessors at 5nm and soon 3nm remains a considerable challenge. This advanced ability remains dominated by two corporate giants: Taiwan Semiconductor Manufacturing Corporation (TSMC) and Samsung from South Korea.
SMIC may well have managed to produce 7nm chips — but it has done so by doing it the hard way, using Deep Ultraviolet Lithography (DUV) tools which expose the silicon to light three or even four times in contrast to the once needed for Extreme Ultraviolet Lithography. Whether mass-production at this level can ever be commercially competitive remains uncertain. Ongoing subsidisation is likely to be required.
This is going to be a problem, it has become evident that China’s flagship initiative for promoting indigenous microprocessor manufacture, the China Integrated Circuit Industry Investment Fund (aka the Big Fund), is in trouble. Set up in 2014 and backed by the Ministry of Finance, the Big Fund has received over US$ 40 billion of capitalisation. However, in late 2022 a review conducted by Vice-premier Liu He confirmed that there was little to show for this investment. Those heading the fund are now under investigation for corruption. This failure has been replicated in provinces and municipalities many of whom have capitalised start-ups often headed by individuals with no background in the industry, resulting in a series of failures.
There are some signs of progress, according to the Semiconductor Industry Association (SIA), China’s share of the global chip sales market was only 7.6 percent in 2020. However, this number is rapidly increasing, thanks to the country’s expanding domestic market (semiconductor device sales in China have surged from $13 billion in 2015 to $39.8 billion in 2020 alone). China’s homegrown chip industry is growing. In 2011, there were just under 1,300 registered chip companies in China, but by 2020, this number had skyrocketed to 22,800 then 70,000 in 2021 — it has however since begun to fall.
The issue is, whilst China may well have been able to brute force its domestic manufactures into creation of commercially viable 7nm chips, and maybe even a few 5nm chips, the future of technology lies in advanced applications like artificial intelligence (AI) and quantum computing, both of which demand extremely high-performance microprocessors. These applications require chips with smaller nanometer (nm) sizes to perform complex computations faster and more efficiently. For instance, chips with 5nm and below are the bare minimum for the enhanced processing speeds and energy efficiency, required for for heavy-load technologies like AI model training.
China’s capabilities will fall short here. The market will not stand still and the capability gap, not to mention the global supply chain for these advanced microprocessors, will inevitably move further and further from Chinese fabs.
Looking ahead
Huawei’s announcement this week may represent a propaganda win for Chinese tech and by extension the ruling CPC, however it’s important to consider the potential downsides of such a strategic build-up of exclusively foreign made equipment.
While it’s clear that the procurement of these specialized lithography machines has given China a temporary ability, this is unlikely to be sustainable in the long term. Manufacturing chips at any grade, let alone at the 7-nanometer level is a complex and demanding process, and the equipment required is not only sophisticated but also highly sensitive.
Inevitably, these machines will experience wear and tear, and their hyper-specialized nature means that repairing them, or even procuring replacement parts, will present a significant challenge. This is to be compounded by the fact that ASML, as a Dutch company, is bound by the same international sanctions that prompted China’s strategic stockpiling in the first place. In the event of a breakdown, the inability of ASML to assist with repairs or provide parts will significantly disrupt China’s chip production capabilities.
This exposes a critical vulnerability in China’s semiconductor sector: a high dependence on a single source for crucial machinery. The strategy of stockpiling might mitigate the impact of trade sanctions in the short-term, but it also underscores the inherent risks of relying too heavily on imported technology.
Looking forward, it is reasonable to anticipate more announcements from China about supposed ‘ground-breaking’ advancements in 7nm, and possibly even 5nm, technology. However, these announcements need to be seen in context. The reality is that these developments, while noteworthy, do not necessarily signify a breakthrough in technology, but rather a demonstration of China’s stockpiled manufacturing capabilities.
The use of 7nm and 5nm technology could be seen as a strategic move to maintain industry momentum during a period of regulatory and supply chain pressure. However, a true test of China’s semiconductor industry’s resilience and innovation capacity would be in an ability to develop and manufacture advanced chips beyond this established technology, particularly considering the rapid pace at which global chip technology is advancing.
In the face of such challenges, it remains to be seen how China will manage these risks and ensure the longevity and stability of its semiconductor industry.
