Semiconductor supply amid the generative AI revolution

Our survey of over 400 middle market business leaders found that 82% of businesses in the real economy are either actively using or experimenting with generative AI. This incredible uptake of generative AI is fundamentally dependent on advanced semiconductor technology. Due to the significant adoption of generative AI tools, there is significant strain on the advanced semiconductor supply chain. Analysis by Bloomberg estimates that year-on-year upside on demand against prediction will be up to 48% in 2027 for Graphics processing units’ (GPU's).

The availability of advanced semiconductors is essential to the UK establishing dominance in the race to be a generative AI superpower. Against this backdrop governments across the globe are competing to increase inward investment that is shifting the footprint of the global semiconductor supply chain. The winning nations will be those that move swiftly and boldly.

The underlying technology and dominance of Taiwan

At the forefront at delivering against the demand for machine learning and AI chips is US based Nvidia. The inventor of the GPU dominates the market, underpinning the majority of AI applications. Its latest H100 chip has been specifically designed for AI applications with the earlier A100 chip powering many of the world’s supercomputers.

In the complex global semi-conductor supply chain, Taiwan is key. Taiwan dominates the fabrication of semiconductors, producing two thirds of the worlds semiconductors and until recently was the sole producer of the most advanced chips. Other nations contribute to this supply chain however fab capacity is limited to less advanced chips and international focus has been in other areas of the supply chain such as design and R&D. The semi-conductors underlying Nvidia’s GPUs are produced by a single site in Taiwan by the Taiwan Semiconductor Manufacturing Company (TSMC).

Governments have recognised the global complexity of the supply chain and the revolutionising role that generative AI will play in society. Demand for chips is expected to grow globally, as nation states are motivated to secure their own supply in the face of growing political tensions between the US and China.

The US and EU

Due to a lagging capability to produce the most advanced semi-conductor technology, the US’s Society of Manufacturing Engineers (SME) reports that only 12% of the world’s semi-conductor chips are made in the United States, down from 37% in 1990. In the face of growing political tensions, the overdependence on imports has created concerns for the supply chain and national security.

In response to these tensions and as part of a package to further boost the US’s position, Biden’s landmark Chips and Science Act was announced in August 2022. This seeks to provide incentives for manufacturing and research to strengthen the US economy, national security, supply chain resilience and technology leadership. The act includes $39bn in subsidies, a 25% tax credit to promote manufacturing onshore and an additional $13bn of investment in chip research.

The US Semiconductor Industry Association (SIA) reports that the act is already seeing success including 15 new fabs and the expansion of 9 across 12 states .The posterchild of this success is TSMC’s new Arizona fab announced in December 2022 which will produce 3nm chips, the most advanced in the US.

Likewise, the European Chip Act, first announced in February 2022, seeks to establish the EU as a leader in the semi-conductor market by doubling its market share to 20% of the global market by 2030, in addition to securing an EU supply chain. The act seeks to mobilise more than €43 billion funding through three main pillars:

1. R&D and innovation;
2. State aid exemptions covering semiconductor manufacturer; and
3. Steps to monitor and intervene the security of the supply chain as necessary.

The UK’s lagging position

The UK lags behind in both availability and manufacturing of advanced chips for generative AI applications. A review into the UK’s computing capabilities published this year highlights that researchers have insufficient public access to top spec Nvidia GPUs to enable meaningful advancements to AI research with less than 1,000 available. This pushes large-scale generative AI projects into the private sector in addition to requiring collaboration with global counter parts.

The UK has approximately 25 fabs including research and academic facilities. Most of the industrial capacity meets the need of less advanced chips, but the UK, like most countries, does not have any large volume advanced semi-conductor fabrication. The capital costs and lead time for such facilities are significant and not feasible for most nations.

The onshoring of the full supply chain should not be the end goal. Rather, through the National Semi-conductor strategy the UK Government has instead sought to double down on the UK’s strength in design, R&D and in compound semi-conductors. With a committed £200 million investment over the next 3 years and possibility of a £1 billion investment over the next decade to:

• improve access to infrastructure;
• power more research and development; and
• facilitate greater international cooperation.

This focus on IP and R&D plays to the UK’s strengths, which has a long history of excellent university-based research and successful spin out companies. UK developed ARM architecture underlays many advanced chips, including the important H100.

Through the National Semiconductor strategy, the Government is recognising the importance of securing the UK supply chain. However, such policy measures are minute compared to the global stage and the UK needs significant additional commitment to keep pace.

Changing landscape

With the increasing demand a proactive awareness of the supply chain will be critical for many in the UK’s middle market. These business leaders should be mindful of the complexity of this fragile supply chain as well as the the wider geopolitical and capacity tensions.

The impact of challenges in the system may range from frustration in procuring replacement IT assets to significant production delays caused by unavailability of key components. Some companies can seek to take advantage of incentives and government policy to enhance inward investment and secure their own supply chain. Others may need to take a proactive approach in having a full understanding of their own supply chain and making contingency plans in the case of disruption.