I’m inclined to start by writing, ‘Here we go again!’, for we have indeed been here before, and none too successfully.
News that the UK Government is planning to “invest in the semiconductor industry to turn the UK into the next Silicon Valley” sends shivers down the spine, not least because most of Silicon Valley is no longer really even silicon any more. That ship has sailed to the Far East, and that level of concentration has been a slow, but inevitable, lesson for the USA in the full implications of going gang-busters on off-shore manufacturing. There is now a major push to reverse that movement in the USA, and the numbers ought to be something of an object lesson for the new UK plans.
The stated aim of the UK Finance Minister, Chancellor of the Exchequer Jeremy Hunt, is both extremely vague and potentially quite precise – and not necessarily in a good way. For example, if he means being the fount of leadership in semiconductor manufacture, it is currently a ‘has been’ objective. Even the USA now only accounts for some 12% of global chip sales, and most of those are produced well away from the area south of San Francisco.
And if the UK really had aspirations of achieving such a goal, £2 billion over 20 years would get the country nowhere, slowly. Semiconductor manufacture is a market the UK will not – cannot afford to – compete in.
Even then, a budget of £1 billlion investment is too little, too late. It will be history repeating itself. In the late 1970s, the Labour Government of the day launched a plan to invest £50 million in the UK semiconductor industry. The results largely sank without trace. Much of the money went to the established electronics companies, such as GEC, Ferranti and Plessey, and they had already largely lost out in the mainstream and only had small, specialist corners of the marketplace. The most successful for a while was Ferranti with its Uncommitted Logic Arrays (ULA) – a family of pre-built standard chips which could be tailored to fit specific specialised applications by a final connection layer that built the required logic. And as I aim to point out, that is still the best direction for the UK to go.
A bit later came another government investment of £50 million, this time going to Inmos, a joint UK/US initiative to produce the Transputer, an innovative parallel-processor design that never created a real market for itself since by then Intel had come to dominate. Most of the £50 million went on buying an existing wafer fab in Colorado Springs rather than attempt to build new in the UK. The technology was eventually sold to French-based chip specialist STMicroelectronics, where elements of the design and the fundamental architecture are still embedded in subsequent specialist devices.
If we want it, this is the cost
By comparison look at US numbers emerging for planning in this area, now it has realized that offshore manufacturing has handed the market first to Japan, and now to Taiwan and China, on a silicon plate.
Having decided to rectify the situation, there are now plans to spend eye-watering sums to redress the balance. Intel has plans in process for new wafer fabrication plants in New Albany, Ohio, with a budget of $20 billion, though it also has a new fab being built in China at a cost of $2.5 billion. That, by the way, is a good marker for the UK Government if it really wants to start getting into the semiconductor business.
So the best option would be for the UK to attract a US, Japanese, Chinese or Taiwanese company to invest in building a plant here - but why should they? It would face the same problem that most manufacturing industries now face in the UK. Being outside of the EU Single Market and facing all the potential tariff and non-tariff barriers that now exist, it would also take several years to build, and no company would build a facility producing brand new chips to new designs and production specs. So it would more likely be a fab to take over production of old, established devices heading onto the downside slope of product maturity while an existing one, staffed with plenty of existing expertise, is re-modelled for the latest technologies.
As an example of the costs the UK would be attempting to compete with, the USA has recently passed a Chip Production Subsidy Bill that sets aside some $76 billion to assist companies in building new fabs, and four of them — GlobalFoundries, Samsung Foundry, TSMC, and Texas Instruments — are joining Intel in starting on new developments.
We can do ‘R’, but will need ‘D’
Chancellor Hunt has talked about a concentration on R&D, but only one of the two obvious areas where R&D work is pressing ahead - AI and Quantum Computing – has a direct connection with semiconductor R&D. The UK claims leadership in both, at least in research, but it is in development that the real benefits will be found, especially when it comes to meeting the dream of building a high tech/high wage economy for the country.
Here AI does have real potential, and ironically it is possible to speculate that the architecture of the Transputer may yet play a direct part in making it work really well. But even here, the difference between research and development can be seen with the success of ChatGPT from OpenAI in the USA. However, this may yet help boost some of the AI research work in the UK, where the technology’s relationship to philosophical issues, such as ethical application, are to the fore. ChatGPT has already demonstrated that going to development and production too quickly can open up more problems than it solves.
Quantum Computing is looking more encouraging for the UK. There is, for example, Quantum Motion, which is aiming to produce qubits in silicon, and at scale. The company has stated funding of some £62 million and has just gained Sony as an additional investor. Though no money has been mentioned, perhaps the more important factor is that Sony is already a known player in semiconductor manufacture and has the ability open up the Japanese market to Quantum Motion’s technology.
The upside of the company’s approach is its aim to realize Quantum Computing in silicon. At the moment quantum technology is largely made of expensive, specially constructed systems. The downside may well be that Sony is not seen as a major player in semiconductor markets, either as a vendor or as a technology leader, such as producing chips using 5 micron and below channel widths. Yet these are likely to be the technologies required to make it a viable product. So without a local manufacturing source, the best that may be likely is the same model as ARM, selling licenses for the design to third parties, with the same results then being quite likely.
A third line of research being mentioned is compound semiconductors, which is a slightly curious topic in that all semiconductors are compound, where a pure silicon substrate has various materials diffused to it to create the logic gates and switches.
Developing new types of compound devices is an obvious target for the UK’s well-respected academia, and could map well on the the work being done at Quantum Motion. But it is unlikely to achieve the Chancellor’s other key goal, turning the UK into a hi-tech, high wage economy, unless there is real development – and subsequent local production – to follow. Yes, research will produce hi-tech jobs, but there won't be very many of them – a few tens of thousands, and a few tens of thousand more support staff, but it won't stretch far across the 60+ million population of the UK.
In addition, while the core research component will be a viable opportunity, it is unlikely that £2 billion over 20 years will amount to much when it comes to development of the technologies into products, and even less so when it comes to scaling to manufacture. In other words, the chances are that the research results will be sold off to the highest bidders at the earliest opportunity. Again, one simply has to think of ARM as an example. Having established itself as a dominant force in the mobile phone marketplace, it was sold off to Japanese conglomerate SoftBank. But it is also a good example of how difficult Hunt’s plan may be to achieve. Despite having a 2021 revenue of $2.7 billion, it currently employs not much more than 5,000 people.
The reason for that is simple: basically, it makes nothing. Instead it has developed a range of chip architecture and instruction sets that can be licenced by chip manufacturers. It is their job to design those into their own chips in ways that meet their own product goals. High tech it certainly is, high wage is also highly likely. But a few thousand employees, at best, is not what Hunt says he wants to deliver.
There is, however, a model on which to base a possible future. If an opportunity exists where a large number of companies can be created that exploit the latest technologies, and even require a wider range of skill sets than just the high level digital/chemistry/physics of semiconductors, there could scope for Hunt’s dream to be at least partially realised.
I suspect such an opportunity is to be found in edge computing. To be effective, this will need to be highly granular, and distributed as widely as possible across user infrastructures. Each step in every production process will need to be observed, measured and managed. More importantly, that data will need processing, and processing locally, with multiple tiers of management across the whole infrastructure of nearly every company, for it will be by remaining on top of all its data sources, in as near real time as possible, that most businesses will remain competitive.
This requirement is opening up new opportunities for a new market sector to open up, one which will combine compute functions, with systems engineering, new applications development and physical engineering to develop and build the huge range of monitoring/control/process management/change management functions that each step in a process requires (be that a machine tool, a sensor, or an assembly line).
Underpinning them all will be the chips and operating software familiar to many, but each application is likely to require, at the minimum, a variant of a type of device. Many will require specific devices designed and engineered to meet their requirements. And they will likely require many who have other skills or understand the needs of the process, rather than the underpinning digital technologies, for them to be effective. That stands more chance of creating the economy Hunt is targeting.