Delivering more efficient Fabs

6 mins read

The manufacturing of semiconductors is complex, and the sector is a technology-intensive one that involves huge capital investments, high risk, and long payback periods.

Investment is in the billions of dollars, and it needs to be both significant and sustained. The efficient manufacture of semiconductors requires huge scale if it is to be cost effective. As a case in point, Intel’s new fabrication facility in Magdeburg, is expected to cost in the region of €12bn while Infineon’s latest power-semiconductor facility is set to cost in excess of €5bn.

But it’s not just the cost of building these facilities that is huge, developing an ecosystem for chip manufacturing in greenfield locations requires a suitable infrastructure that’s capable of delivering uninterrupted power, huge quantities of clean water as well as the chemicals and gases necessary for chip fabrication.

Traditionally, the production of chips has been concentrated in just a few countries where the required ecosystem has been developed over time – but that is now changing. The CHIPS and Science Act brought in by the Biden administration has encouraged billions more in domestic investment, and has helped to boost American semiconductor research, development, and production and has catalysed regional economic growth and development. Similar developments are also taking place in the EU, India and Japan.

The investment in semiconductors is seen as a strategic necessity as they are at the heart of modern technology. Semiconductor sales reached $618 billion in 2022, up over 30 percent in just two years, and are on course to smash through $1trn by 2030.

However, despite investment and strong sales growth the sector is confronted by a number of serious challenges. Supply chain disruption, caused by a combination of factors including the COVID -19 pandemic and geopolitical tensions between the US and China, has exacerbated a severe chip shortage globally and forced the industry to look at diversifying supply chains.

As a result, countries are strengthening domestic capabilities while encouraging the likes of Intel and TSMC to invest in new facilities outside of their domestic markets. These companies in turn are collaborating with competitors to deliver new capacity and capabilities.

For example, Intel Foundry Services (IFS) and Tower Semiconductor have recently announced an agreement whereby Intel will provide foundry services and 300mm manufacturing capacity to help Tower serve its customers globally.

In the EU STMicroelectronics’ 300mm capacity for silicon-on-insulator CMOS in Crolles, France forms part of a joint venture with GlobalFoundries, while another 300mm fab for mixed-signal and power devices based in Agrate, Italy is being operated with Tower Semiconductor.

Another issue for the semiconductor industry is the pursuit of miniaturisation driven by Moore’s Law, which requires constant innovation and the need for improved levels of quality control, yield improvement, and time-to-market. That doesn’t come cheap!

But while countries might be committing billions to investing in and developing a semiconductor industry that then brings into focus one of the most serious challenges for the industry, and that’s the lack of trained staff and a widening gap between demand for and the availability of qualified individuals.

The competition for talent is intense and the demand for highly skilled workers with expertise in areas such as materials science, device physics, process engineering and software development is making life hard for the industry.

As a result, it needs to focus on investing in education and training programmes that can cultivate a new generation of experts in this field, and many are calling for increased levels of investment in offering internships, apprenticeships and mentorship programmes that can give talent the start they need.

That very lack of skills, training, and education is not an academic debate, it’s already had an impact. TSMC announced that it was delaying the start of production at its Arizona fab from 2024 to 2025, due to a shortage of skilled technicians.

Skills shortage

Research from the US Semiconductor Industry Association (SIA), in partnership with Oxford Economics, found that the United States faces a significant shortage of technicians, computer scientists, and engineers and predicted a projected shortfall of 67,000 of these workers in the semiconductor industry by 2030.

The report, titled “Chipping Away: Assessing and Addressing the Labor Market Gap Facing the US Semiconductor Industry,” made several policy recommendations to help close the talent gap as well complement the workforce development initiatives already being carried out by companies.

According to Matt Johnson, president and CEO of Silicon Labs and SIA board chair, “Effective government-industry collaboration can overcome the talent shortage facing our industry, build the strongest American tech workforce possible, and unleash the full potential of semiconductor innovation.”

The study presented three core recommendations to strengthen the US technical workforce. The first being support for regional partnerships and programmes aimed at growing the pipeline for skilled technicians for semiconductor manufacturing and other advanced manufacturing sectors; the second, was to develop the domestic STEM pipeline for engineers and computer scientists; and finally, it called for the retention of more international advanced degree students within the domestic economy.

While the CHIPS Act has certainly set the stage for long-run investment and increased global competitiveness it will require tens of thousands of new workers to increase productive capacity in the US.

Of the total estimated semiconductor technical workforce gap of 67,000 by 2030, the study estimates approximately 39% of the gap (26,400 jobs) will be in technician occupations, 41% (27,300 jobs) in engineering occupations, and 20% (13,400 jobs) in computer science.

A technical solution?

To combat these challenges, there is growing interest in artificial intelligence, machine learning and automation to help streamline the complex design and manufacturing processes to reduce costs.

Smarter manufacturing will help to improve the efficiency, quality, and sustainability of the production process optimising the performance and yield of the chips by adjusting process parameters and equipment settings in real time.

The sheer complexity of chip production is huge, as it’s a process that can entail hundreds of different steps, generating more manufacturing options than there are ‘atoms in the universe’.

“This hinders agility, stalls progress, and prevents fabs from fulfilling their potential,” according to Jamie Potter, CEO of Flexciton, a UK company that has developed a radically new solution for innovative chip makers to address the issue of complexity and to help drive new developments in manufacturing.

“We have developed an intelligent advantage platform that can enhance decision making by doing the heavy lifting for you. It analyses and automatically selects the best manufacturing option from trillions of possibilities – making it easier and faster to meet your fab’s full potential,” Potter explains. “Our aim is to help fabs drive progress efficiently, profitably, and sustainably.”

The manufacture of semiconductors is probably the most complex process to undertake and Flexciton looks to bring advanced technology to play to enable things to be done better.

“We’ve been working with the industry for some time,” said Potter, “and started with Seagate, who helped us bring out technology to market.

“Essentially, what we are doing is building a type of ‘brain’ for the factory, one that removes human decision making from the process. A process that is far too complicated to rely on people to solve. Consequently, there’s a massive gap in efficiency and more intelligent technology can be used to significantly improve both productivity and efficiency,” said Potter.

Most fabs will have collected vast amounts of data, data which can then be used by Flexciton to address production issues and to programme their platform to achieve specific objectives.

“Where FabTime, a web-based dashboard that processes operational transactions from the factory execution system (MES), is available we can install our platform and be operational in a matter of hours. Otherwise, the collection and processing of data can take up to 3 months.”

Potter makes the point highlighted in the report published by the Semiconductor Industry Association that with so many new facilities being built the pressure on finding the necessary skilled workers will be immense.

“Just where are these people expected to come from? And this skills shortage seems to be getting worse.  A few years ago, the semiconductor industry was very attractive and could get the talent it needed. Today engineers join Google, Facebook or a host of new AI start-ups. So, the challenge is that there is a massive performance gap at the factory level between what factories are and what they could be.”

When it comes to manufacturing a semiconductor the process involves hundreds, often thousands, of different machines with wafers having to go through the same process repeatedly whether that’s wafer fabrication, photoresist coating, lithography, etching, ion implementation or assembly and packaging.

“As a result, you’ll see thousands of wafers at any one time moving around a fab passing through the same machines. The process isn’t efficient, and you’ll see queues for some tools, while other machines will lie idle. Based on data provided by the fab our tool looks to boost productivity and we assess the whole facility so that we can optimism the flow of wafers by automatically selecting the best manufacturing option.”

According to Potter orders for the platform have surged in the past year.

“It’s still early days and it’s a hard industry to engage with – visionaries and early adopters are crucial – and its an industry which tends to exhibit a herd mentality, but we are seeing a strong uptick in customers talking to us. The industry needs to be more open, as there are plenty of companies like us that could have a big impact but are struggling to be heard.”

Based in Austin, Texas, SandBox Semiconductor is another company whose technology is being used to boost semiconductor manufacturing productivity. It has developed an AI-enabled software platform for semiconductor process engineers that can reduce the long development cycles necessary to create new chips while also providing better insight and enabling new processes not possible before.

By using a computational modelling approach it’s been able to accelerate how process engineers create recipes during the ramp up and volume manufacturing stages of semiconductor manufacturing.

The SandBox Studio AI platform combines physics-based modelling with AI to allow engineers to more quickly - and with less dependency on expensive physical experiments - streamline the thousands of steps involved in semiconductor manufacturing- particularly etch and deposition feature level optimisation. 

The company’s technology is already being used by foundries, memory suppliers and equipment companies and features that include data visualisation and hybrid metrology, are being added to the platform.

While heavy subsidies are going into advanced chipmaking as New Electronics reported earlier this year, there is a realisation that no single region will ever be self-sufficient in semiconductors but what companies like Flexitron and SandBox can provide is that where those investments in manufacturing are made, they can be significantly more productive and efficient.