5G Risk? Huawei Is Small Potatoes Compared To TSMC (Who?)
“5G will bring about an unprecedented era for semiconductors.” – The CCP
“Modern wars are fought with semiconductors.” – a U.S. Senator
When the (former) chief spy-master of the United Kingdom, John Sawers, goes public with a Financial Times op-ed entitled “The UK Should Bar Huawei From Its Network” – well, it may at last be worth taking seriously the prospect that Huawei could be excluded from at least one crucial European market.
Huawei is of course the world’s largest manufacturer of telecommunications equipment, and a leader in the development of 5G technology for the next generation of wireless networks. Huawei has supplied about a third of the current 4G wireless infrastructure in Britain. Ripping out their installed UK base (which would be necessary, evidently, if Britain disqualifies them for 5G) would cost, at least £2bn. Perhaps more. Huawei has also said it would spend another £1bn to build an advanced research center near Cambridge – presumably the loss of that investment would be part of Britain’s opportunity cost.
But what one might assume, based on the Sawers’ background as the former head of the James Bond world of MI6 – that the argument for the ban would be based on strict security concerns, state espionage, cyber threats, and the like – appears, oddly, not to be his main concern here. The national security risk is certainly the argument against Huawei that others have made, including the American government and, lately, the former CEO of Google. But it was apparently not enough to push the UK all the way to a full ban. As recently as January, the UK was prepared to resist American pressure and retain a significant 5G role for Huawei. So, what has changed, exactly?
There are three additional reasons that might cause Britain sufficient concern to consider a full ban.
The developing backlash against China, the charges of Covid-19 disinformation, its “wolf warrior” diplomacy against its critics (which Sawers references at some length); in this scenario, Britain is really at odds with the Chinese government. Huawei would be “collateral damage.” China’s diplomatic aggressiveness certainly doesn’t help Huawei any. China’s ambassador in London cautioned that it was “not in the UK’s interest” to make this move.
- “The China business community are all watching how you handle Huawei. If you get rid of Huawei it sends out a very bad message to other Chinese businesses…. If you want to make China a hostile country, you will have to bear the consequences… If you dance to the tune of other countries, how can you call yourself Great Britain?”
Tough talk, and obviously counterproductive. But the Brits are used to absorbing this sort of abuse. They won’t shoot Huawei just because of the clumsiness of some of Xi Jinping’s emissaries.
In a previous column, I excerpted the Report of the UK’s reviewers who have tracked Huawei’s technical performance, and challenges. Without repeating that material here, the summary conclusion is that there may be significant shortcomings, both in terms of equipment “vulnerabilities” and (more seriously) the integrity of Huawei’s engineering processes.
These technical issues are worrying, but with new technology there are always technical issues. If Huawei responds constructively, it need not be a fatal affair.
Supply Chain Vulnerabilities
The U.S. has been ratcheting down on Huawei’s critical suppliers, to deny Huawei access to U.S. technology that it has needed for 5G. The initial restrictions blocked U.S. companies like Qualcomm and Google from selling directly to Huawei. But in May, the supplier-ban was extended to include foreign companies that use American technology or equipment.
It is a real possibility that Huawei might not be able to obtain the components it needs to build its 5G equipment. This is Sawers’ “business case” for a full ban. Here’s his assessment:
- “[Prior to this] Huawei’s use of reliable suppliers, such as the Taiwan Semiconductor Manufacturing Company, for crucial components also meant we could assure its supply chain…. But the latest US sanctions mean that reliable non-Chinese suppliers to Huawei can no longer work with the company. UK intelligence services can therefore no longer provide the needed assurances that Chinese-made equipment is still safe to use in the UK’s telecoms network. There are now sound technical reasons for the UK to change January’s decision.”
But why is this new factor decisive, even more than national security concerns? Why – after weathering so many storms – does Huawei itself admit that this new move may put its very survival at stake?
The History and Structure of the Semiconductor Industry (The Simple Version)
Modern wars are fought with semiconductors. And this latest American move is a “surgical” strike, with a very specific direct target, identified in Sawers’ comments – a non-American supplier to Huawei, Taiwan Semiconductor Manufacturing Company, generally known as TSMC. But who exactly is TSMC? And why is this potentially a checkmate move in the American game against Huawei?
First of all, let it be noted that TSMC – which trades on the New York Stock Exchange – is now the largest semiconductor company in the world by market value.
But to understand the power the company holds over Huawei, we need to understand TSMC’s business model. And to understand that, we need a bit of history.
The Evolution of IC Technology
Jack Kilby invented the integrated circuit at Texas Instruments in 1959. He later received the Nobel Prize for this accomplishment. It was the Physics prize – which was the appropriate field – because creating the first IC was basically a challenge of materials science. Kilby’s “chip” was a lump of germanium with some wires and solder. It had just two transistors, a proof-of-concept, but hardly a commercial product. Learning how to mass-manufacture this thing was the name of the game from that point on.
The rate of improvement in IC manufacturing is described by Gordon Moore’s famous “Law” which holds that the number of transistors on a chip doubles every two years or so. This trend has turned Kilby’s crude device into the incredible processing engines that power the digital world today. 60 years after Kilby, a single chip may incorporate more than 50 billion transistors, at a density of 3 nanometers per node.
Modern semiconductor production relies on precision manufacturing at its most extreme, down nearly to the quantum limits. It’s all about hardware. Peruse here a standard summary description of IC fabrication:
- “Semiconductor device fabrication is the process used to manufacture semiconductor devices, typically the metal-oxide semiconductor (MOS) devices used in the integrated circuit (IC) chips… It is a multiple-step sequence of photolithographic and chemical processing steps (such as surface passivation, thermal oxidation, planar diffusion and junction isolation) during which electronic circuits are gradually created on a wafer made of pure semiconducting material.”
The Economics of IC “Fab”
Modern IC fabrication technology is spectacularly expensive. Moore’s less well known “Second Law” states that the cost of a semiconductor plant doubles every four years. TSMC just announced plans for a new plant in Arizona, at a cost of $12 billion. They are spending another $20 Bn on a new Asian plant. An IC fab is 5 to 10 times more expensive than a deepwater oil drilling platform. It is several times more expensive than a nuclear power plant. These IC facilities are probably the most expensive private-sector technology investments of all time.
Moreover, the investments do not retain their competitive edge for very long. Whereas a nuclear plant will operate for decades, an IC fab facility will be obsolete – noncompetitive – within 5 years or so.
- “To remain competitive in the semiconductor industry, firms must continually invest a significant share of revenues in both R&D and new plant and equipment. The pace of technological changes in the industry requires that companies develop more complex designs and process technology and introduce production machinery capable of manufacturing components with smaller feature sizes. The ability to design and produce state-of-the-art semiconductor components can only be maintained through a continual commitment to keeping pace with industry-wide investment rates of roughly 30 percent of sales.”
Such savage economics have split the industry into two camps:
- Foundries, who make the heavy capital investments as manufacturers, and
- “Fabless” semiconductor companies that design chips, but do not manufacture them, and avoid the high cost of fabrication facilities
Today, most of the American semiconductor industry is fabless. Companies like Qualcomm and Nvidia don’t “make chips.” They perform high-value design, and contract with foundries – like TMSC, the largest foundry in the world – to produce the actual chips, sidestepping Moore’s Second Law. As an asset-lite business model, the fabless approach is highly successful. Investors are enamored.
But there is a catch. There aren’t very many companies left who can play the foundry game (and why should they, when financial markets seem to frown). It is also a business where scale is critical. The result is that as TSMC has become the market leader, they have also become a critical supplier for almost all the fabless semiconductor companies in the world.
The Checkmate Move?
Huawei is one of those fabless IC companies. They rely on TSMC to fabricate both the chips of their own design and the chips designed by American fabless IC firms which they have used. In fact, until recently, Huawei accounted for 20% of TSMC’s revenue, a situation of mutual dependence, though skewed in TSMC’s favor.
TSMC is not a U.S. company, and was not affected by previous bans on supplying Huawei. The earlier rounds affected American fabless suppliers (like Qualcomm). Huawei was still dependent (as of 2019) on chips sourced from these companies. Its response to the bans was to accelerate its efforts to design its own chips. But the latest U.S. government measures enjoin any company, anywhere, from using American technology to supply Huawei – and that catches TSMC in the anti-Huawei net, because TSMC uses semiconductor-manufacturing equipment that does come from U.S. companies (such as Applied Materials, the market leader in that category).
Huawei can probably, in time, design chips to replace the ones designed by American companies like Qualcomm. But Huawei is not a foundry, and has no plausible pathway to become one any time soon. It’s Moore’s Second Law again, and what we might cite as the corollary: the later you enter the business, the harder it is to become a viable fab. In other words, the IC fabrication business has a long learning curve, and is not mastered quickly even with lots of cash to spend.
(It is interesting that, following the U.S. ban, a much smaller Chinese foundry, Semiconductor Manufacturing International Corp. – SMIC – announced a large and apparently oversubscribed Initial Public Offering in Hong Kong. SMIC was cofounded by Huawei, and has struggled to gain its footing. It was delisted from the New York Stock Exchange in 2019. It had been sued for intellectual property theft by TSMC, and made a large payment to settle with them. A sadly familiar pattern.)
In any case, the general view (at the moment) is that Huawei has no good alternative to TSMC. Which is why Mr. Sawers has now concluded that they are out of the 5G game, at least for the UK.
The Real 5G Risk?
But the TSMC situation highlights a larger risk, which is under-appreciated: the reliance of the entire industry on TSMC. As Earl Lum (CEO of EJL Wireless Research) puts it: “I hope there is no earthquake in Taiwan. If TSMC were taken offline, it would cripple the entire industry.” TSMC so dominates the IC manufacturing business that it is just as critical for many other 5G supply chains besides Huawei’s. Apple, Qualcomm, and Nvidia all rely significantly on TSMC to manufacture their chips. TSMC might be considered, in engineering terms, as the “single point of failure” in the digital economy.
Is this a risk for 5G? In principle, it is. One of the consequences of the American enamorment with asset-lite business models (across many industry sectors) is that the U.S. technology base in semiconductors is now skewed away from the hardware end of the business. In 2015, just 13% of the world’s foundry capacity was located in North America. (We have 69% of the fabless IC market.)
Congress has roused itself to the threat, proposing last month to channel more than $20 billion of government funding specifically to support investment semiconductor fab.
- “The bills call for the State and Commerce departments to create a federal program that would match state incentives offered to companies that build foundries.”
It won’t work. $20Bn might buy one state-of-the-art plant. Which might be competitive for 5 years. But the critical private-sector investment that Congress would hope to stimulate in support of this program won’t mobilize. The financial markets hate capital expenditures. Intel – the last major American player in the IC fabrication business – has suffered from this investor bias, as its P/E shows. The company has struggled with costly manufacturing challenges that its fabless competitors aren’t exposed to. Its manufacturing investments are not really seen as assets by many investors.
- “It held on too long to a vertically integrated business model and has been outflanked by a new industry structure that has unleashed fleet-footed, fabless competitors.”
Asset-heavy business models are so Henry-Ford. Apple has coined the new mantra: Designed in California, Manufactured Wherever. (I just bought a T-shirt that touts itself as “Designed in Brooklyn, Made in Peru.”) We’ve committed ourselves as a country to the high-value-per-dollar-invested end of the business – which is all about brand, design, software, content – not expensive, rapidly depreciating hardware. Asset-lite is where the digital value-added — most of it — is to be found. The IC fabrication business is an exception, perhaps, but TSMC now owns it, and for the economic reasons cited above, a panicky and passing government handout won’t reverse things. The U.S. is still a player in fab, but not the dominant player. We’ve placed our strategic bet in fabless. Which means we’d better hope for the continued good health and fortune of our Taiwanese friends.
If Huawei were to disappear tomorrow, it would disrupt but not derail 5G. If TSMC were to disappear, or be neutralized, it would cripple 5G and many other facets of the digital economy.
Could this happen? Well, it’s not earthquakes that could neutralize TSMC. In Taiwan, they’re watching the news, like all of us. They know it’s pretty much only the 7th fleet that stands between them and the CCP goons now taking over Hong Kong.