Huawei has a chip problem, but it's far from being a write-off

For lovers of melodrama, the annual updates by Huawei's carousel of rotating bosses are often a treat. Hit by US sanctions under former US President Donald Trump, the Chinese company once depicted itself as a fighter plane pockmarked by gunfire, struggling to stay airborne. This year's update was disappointingly low-key. "We've been through a lot over the last few years," said Ken Hu without fanfare, after rotating into the hot seat. But press reports have compensated in wildly diverging ways.

"America's assassination attempt on Huawei is backfiring," blared The Economist in a detailed briefing on June 13. The message was that a sanctions program started by Trump and accelerated under President Joe Biden has largely failed. Huawei has not only avoided a crash landing but also gained altitude and become more self-reliant.

Two weeks later, Quartz weighed in with its own, much shorter assessment. "US sanctions against China's AI chip efforts seem to be working," ran the headline, with a standfirst saying: "Huawei is reportedly having a hard time increasing production of its Ascend 910B AI chip." American policymakers have denied China the cutting-edge tools it needs to produce the most advanced chips. The older equipment available to Huawei just isn't fit for purpose.

The reality is far more nuanced than either of these stories suggests. But to judge the success or failure of US policy, one must first ask what the sanctions were supposed to achieve. They were introduced long before companies and governments became fixated on generative AI, when the in-vogue technology – believe it or not – was 5G.

Given the share prices of 5G stakeholders at the time, markets never expected the next-generation mobile technology to be a money spinner. But governments were fooled into thinking it would connect everything from insulin drips to ballistic missiles and power economic growth. Amid signs of Chinese government assertiveness, letting a Chinese company anywhere near western 5G networks would surely be insane.

Pincer movement

The campaign against Huawei was therefore two-pronged. The first meant exerting pressure on US allies and friendly countries in Europe to flush Huawei out of their networks. Many had grown reliant on Huawei in the 4G era, seeing it as a low-cost but technologically sophisticated alternative to the likes of Alcatel-Lucent, Ericsson and Nokia Siemens Networks. For technological and economic reasons, telcos were likely to stick with the same vendors when upgrading to 5G. And by the time it came along, Huawei was widely perceived to have the technological edge over Western rivals.

But this part of the campaign has had mixed fortunes. Germany, Europe's biggest economy, has spent years prevaricating while its telcos have built nationwide 5G networks with Huawei kit. Even the UK under Conservative Prime Minister Boris Johnson, a supposed Trump ally, wanted a compromise whereby Huawei would be ejected from the core, the cockpit of the system, but allowed to remain sprawled throughout the vast seating area of the radio access network (RAN).

The other prong of the US campaign, then, was designed to gum up Huawei's supply chains and stop it from producing the trustworthy, competitive equipment that UK and other telcos might want to use in their networks. And the US had an ace up its sleeve: its domination of chip design and chipmaking tools, including the machines used by Asian foundries to produce the most advanced components. The hammer blow came when sanctions denied Huawei access to TSMC, a Taiwanese foundry using Dutch and US machines to crank out the smallest transistors in the world.

This tightening of rules ultimately persuaded UK authorities to ban Huawei from selling 5G products, while giving telcos until the end of 2027 to sanitize their networks. Its bigger impact, though, was on Huawei's smartphone business, never identified by US hawks as a potential conduit for Chinese military "backdoors" and subterfuge, as networks had been. Unable to procure the tiny chips needed for its consumer devices, and barred from Google's suite of software apps, Huawei suffered a smartphone collapse. Honor, a key brand, was sold to state-backed acquirers.

When DUV may do

But this business group, formerly responsible for more than half of Huawei's revenues, staged a dazzling recovery last year – to the alarm of US hawks. New and popular gadgets appear to include 7-nanometer chips, thought to be off limits to Huawei. To produce them, it was believed, a chipmaker would need extreme ultraviolet lithography (EUV) machines. ASML, a Dutch company, has a monopoly on their production, and its government has denied it an export license to serve China.

Experts believe SMIC, a Chinese foundry, has instead resorted to older deep ultraviolet lithography (DUV) machines to churn out 7-nanometer chips for Huawei. While not as good as EUV, DUV can employ a technique called multiple patterning to do it. The problem with multiple patterning is that yields – a percentage measure of the functional chips derived from a wafer – tend to be relatively poor. This could partly explain why SMIC's profitability suffered so badly during its recent first quarter. After a 31% year-over-year rise in cost of sales, to $1.5 billion, SMIC's gross margin shrank from 21% to less than 14%.

Huawei, moreover, now appears to have acknowledged its chip problems for the first time. Richard Windsor, the founder of analyst firm Radio Free Mobile, last month drew attention to a keynote address by Zhang Ping'an, a Huawei executive, at the 2024 China Mobile Computing Power Network Event. In it, Ping'an appears to have recognized that US sanctions have put 3-nanometer and even 5-nanometer technology beyond Huawei's reach. Solving the 7-nanometer problem needs to be the company focus, he apparently said.

Windsor has interpreted that as an admission the current yields are not commercially viable in the long run. "Yields need to be 90% or better in most processes to earn a positive return on the capital invested to build and operate the fab," he said in a blog. "By all accounts, SMIC and Huawei's yields are way below this figure which is why the current situation for making 7-nanometer chips in China is unsustainable."

Signs of strength

But there is no sign that Huawei, unlike SMIC, has borne any costs so far. Its cost of sales last year rose just 5%, to about 325 billion Chinese yuan (US$44.7 billion), while revenues were up nearly a tenth, to about RMB704.2 billion ($96.9 billion). At the smartphone-making consumer business, sales grew 17%, to roughly RMB251.5 billion ($34.6 billion).

Huawei's rebound in the smartphone league tables shows consumers care little if there is a 7-nanometer or 5-nanometer chip in the phone (although battery life and performance might be concerns). If Huawei can fix the problem of yields – and Windsor does not put it past the Chinese – it may not have to worry.

Huawei also has less need for tiny chips at its networks business, the part that bothers policymakers. This is partly because there are far fewer 5G basestations in the world than there are 5G smartphones. Basestations are also much bigger than smartphones and therefore not as space constrained. The semiconductors they incorporate have always tended to be a generation or two behind the chips inside smartphones.

Regardless, chip sanctions have not stopped German and various other European telcos from investing in Huawei kit. The 5G networks now deployed in China, where Huawei is the dominant vendor, are regarded as some of the best in the world. In its latest mobility report, published last month, Ericsson notes that midband 5G equipment – the sort needed for higher levels of performance – now covers about 95% of China. In Europe, the figure is just 30%.

If the US campaign has achieved anything on the networks side, it is a bifurcation of the global market along geopolitical fault lines. China and its friends buy Chinese gear while the rest of the world buys from Nordic or other Asian vendors. But nobody is buying much. Enthusiasm for 5G has waned, and spending on the RAN is expected by Omdia, a Light Reading sister company, to fall by 7% to 9% this year after dropping 11% in 2023.

In this bear market, Huawei is doing considerably better than either Ericsson or Nokia, its main rivals. One reason is that China remains by far the world's biggest 5G market and a dependable source of revenues for Huawei. Sales to Chinese customers were up 17% last year. And while this growth was undoubtedly fueled by consumer purchases of Huawei's latest smartphones, China also accounted for more than two thirds of total company sales, up from 52% back in 2018. By contrast, Ericsson's sales to North America, its most profitable market, tumbled 38%.

The fear of Ericsson boss Börje Ekholm is that the West will fall behind China in this bifurcated market. "If the tech world is fragmented east and west then it is going to mean competition between two ecosystems," he told Light Reading during an interview in August 2021. "A Chinese ecosystem will be formidable competition for the west. It concerns me that end users – customers and enterprises – will feel it in their mobile experience."

Branching out

But Huawei's successes last year owe something to a reinvention forced on it by US sanctions. Unlike Ericsson or Nokia, it has expanded into domestic markets outside its traditional kit-making sector, buoyed by Chinese protectionism and antipathy toward the standard American alternatives. Sales at its cloud computing business, for example, rose 22% last year, to about RMB55.3 billion ($7.6 billion). As small as that makes it next to the industry giants, Huawei is slowly gaining ground.

"Huawei has been doing quite well in its local market and has been growing much more rapidly than the two market leaders, Alibaba and Tencent," said John Dinsdale, the chief analyst and managing director of Synergy Research, in a previous email to Light Reading. "Its market share in China is now into double figures (just!). It does, however, remain a long way behind the leaders."

No doubt, the recent AI expansion could prove difficult if sanctions thwart Huawei's attempts to produce more advanced chips. Yet if the AI story turns out to be solely about chips, rather than the services they are intended to support, Nvidia is heading for an almighty crash. Amazon, Google, Microsoft and various other US tech giants all have AI pitches that are not just to do with silicon. There may be a bigger future opportunity for Huawei in software – in large language models, applications or even artificial general intelligence.

The US sanctions program clearly has numerous flaws. Its first is the assumption that a lead in silicon design somehow translates into a pervasive tech hegemony. From the perspective of US hawks, the second must be the poor enforcement of new rules, with companies like Intel and Qualcomm awarded exemptions that allowed them to keep serving Huawei. On the opposite side, others argue that cutting US companies off from the vast Chinese market is counterproductive. Weakened by lower revenues, they will have less to invest in US research and development. 

Worst of all, though, is the old-fashioned view, almost redolent of Western imperialism, that sanctions will permanently hobble China – that homegrown EUV and other such chip wizardry is somehow beyond the capabilities of China's scientists and always will be. There was similar talk more than 20 years ago when Huawei was routinely dismissed as a copycat, a rip-off merchant and plunderer of US intellectual property. By 2019, technology executives within European telcos reckoned it was the 5G company to beat.

Not everyone outside China subscribes to such views. Ericsson's Ekholm seems to be among them, fretting in 2021 about the West's ability to "keep up with the vast R&D spend in Asia – particularly China – that's already happening." The year before, analysts at New Street Research wrote that China's vast resources of human capital would ultimately give it a decisive long-term advantage.

Earl Lum, a semiconductor expert at EJL Wireless Research, has described it in succinct terms. "There are so many people in China to hire," he previously told Light Reading. "It doesn't matter that everyone you are hiring isn't an Einstein. One of them will be."