Despite innumerable challenges the global semiconductor industry continues to perform relatively well, especially those companies that are involved in the generative AI chip market.
However it’s performing though the industry is facing possibly one of the most disruptive periods of the last 25 years, according to Ankur Gupta, Sr Director of Product Management at Model N, a software solutions company. “Of course, there are geopolitical pressures but the speed at which technology is evolving is also a challenge. Gen AI is changing everything with shifting demand patterns and the emergence of new competitors.
“Tariffs and trade restrictions are now adding even more volatility to an already delicate supply chain, so companies are having to build much greater resilience and that requires better auditing to identify and manage any vulnerabilities.”
According to Gupta data and analytics solutions will enable companies to be more agile at a time when they need to navigate fluctuating market conditions.
“Traditionally, the semiconductor industry has relied on long, fixed term contracts in relatively stable markets. Today those markets are changing almost weekly so they need to be more agile and flexible. The supply chain also needs greater transparency, especially as regulatory requirements are tightened.
“Data will be critical as it impacts everything. Companies need an intelligent data-rich strategy that can enable and support better informed decisions. Strong analytics and leveraging AI will ensure faster and smarter decisions.”
Over the course of 2024 the industry saw growth of almost 20% and supply chains performed relatively well but with the demand for AI chips increasing some analysts believe that the industry may be more vulnerable to supply chain shocks going forward – a view held well before the Trump Administration embarked on its policy of tariffs.
According to Deloitte, although the industry is likely to become less concentrated geographically, thanks to various chips acts and the trends of onshoring, reshoring, and nearshoring, the industry remains “highly vulnerable for the next year or two, at least.”
The speed of innovation may be accelerating but it remains extremely expensive and many of the chips being used for training and inference of Gen AI cost tens of thousands of dollars to develop, and specialised chips are expected to gain increased prominence over general-purpose ones, as certain AI workloads require more customised approaches to designing chips.
Imec’s Technology Forum, held in Antwerp last month, saw Luc Van den hove, the organisation’s CEO, talk about the need to develop more reconfigurable chip architectures if the industry wanted to avoid becoming a bottleneck for future generations of artificial intelligence.
According to Van den hove rapid AI algorithm innovation was fast outpacing the current strategy of developing specific chips, which was having an impact on energy, cost and hardware development speed.
"By the time the AI hardware is finally ready, the fast-moving AI software community may have taken a different turn," Van den hove said. He pointed out that while companies were building custom chips to speed up innovation, he warned that approach was risky and for most companies uneconomical.
Van den hove believes future chips will see their capabilities merged into building blocks called supercells, with a network-on-chip steering and reconfiguring these supercells so they can be quickly adapted.
This approach will require true three-dimensional stacking, a manufacturing technique where layers of logic and memory silicon are bonded together, Van den hove said.
But while a reconfigurable approach will give more companies the ability to design their own hardware for specific AI workloads, it will introduce a host of new challenges related to arranging, assembling, validating, and testing and so the industry may need to consider new ways of handling complex design processes.
Already, the chip industry is looking to use digital twins to emulate and visualise complex design processes and designers are having to work more closely with EDA companies to strengthen design, simulation, and verification and validation tools and capabilities.
New tools and methodologies may require the broader chip industry, including EDA and design houses, to consider their long-term direction and goals.
“The use of AI to manufacture AI chips can be the future and will help to reduce operational costs,” said Gupta. “But using AI effectively will take time and there are few first movers willing to engage. They need more proof points before accessing an IT budget that is already under pressure. There’s still a lot of unpredictability around embracing AI.”
In term of longer-term goals, TSMC’s recent decision to open its first European Design Centre (EUDC) in Munich to support European customers in designing high density, high performance, and energy-efficient chips is a good example.
Seen as marking a strategic shift for TSMC, which has traditionally concentrated on chip manufacturing, the decision may have been driven by a shortage of advanced design expertise in Europe and the need to closely support customers in maximising the value of the wafer fab TSMC is currently constructing, with various partners, in Germany.
The fast pace of manufacturing will certainly require new capabilities and capacity, and the industry is thought to need to add a million skilled workers by 2030. From core engineering to chip design and manufacturing, operations, and maintenance, while AI may be able to help alleviate some engineering talent shortages, a skills gap still looms, nonetheless.
“We need a more open and transparent talent pool and while there’s a lot of talk around the supply chain, a key ingredient is going to be the talent and the workforce. It’s critical and needs more attention but the attitude towards free movement of people is moving in the wrong direction,” Gupta suggested.
The industry needs to merge strategic goals to workforce development and government should be looking to help by developing expanded training programmes, as well as vocational and professional education. The industry should also be looking to collaborate with educational institutions to develop suitable training and development programmes.
The big ‘what next’ question, however, is whether tariffs continue or even escalate. Their impact on sourcing talent, investment and manufacturing will be significant and could result in the ripping up of the global semiconductor and electronics supply chain.
