The surge in artificial intelligence could cause a global chip shortage by 2026

There may be another global chip shortage, and a new report predicts a surge in demand for artificial intelligence products and services that suppliers may struggle to meet.

According to consulting firm Bain and Company, AI workload could grow 25% to 35% annually by 2027. However, an increase in demand of just 20% has a high probability of upsetting the balance and plunging the world into another chip shortage.

“The explosion of artificial intelligence at the confluence of large end markets could easily exceed this threshold, creating vulnerable bottlenecks across the supply chain,” wrote the authors of the Global Technology Report 2024.

Our hunger for artificial intelligence will also require the construction of larger data centers with a capacity of more than gigawatts. Existing data centers typically range in size from 50 to 200 megawatts.

Combining demand for AI infrastructure and AI-enabled products, the AI ​​software and hardware market is expected to grow 40% to 55% annually over the next three years.

If large data centers cost $1 billion to $4 billion today, they could reach $10 billion to $25 billion in five years, the report said. This puts the total AI market forecast for 2027 at between $780 billion and $990 billion (£584 billion and £741 billion).

SEE: Gartner predicts that global AI chip revenue will increase by 33% in 2024

The spider’s web of supply and the pressure it’s under

To sustain this growing demand, the AI ​​component supply chain must be able to scale up at the same pace. However, in reality, the chain resembles a more complex spider’s web with chip raw materials at its center.

On the one hand, there are the factories and infrastructure required to scale up chip production, and on the other, there are the data centers needed to run AI products. According to Bain, lead times for each range from three and a half years to more than five years, which poses a significant obstacle to keeping up with demand.

The report shows that state-of-the-art factories producing the most advanced chips are the most vulnerable link. They will need to increase production by 25-35% between 2023 and 2026 to keep pace with projected growth in desktop and smartphone sales of 31% and 15%, respectively.

To keep pace, up to five more state-of-the-art factories would need to be built, at an estimated cost of $40 billion to $75 billion.

There is also a supply chain dedicated to turning chips into smartphones and PCs with AI functionality on devices such as Apple Intelligence devices, which are growing in popularity as consumers’ desire for data security increases.

SEE: Gartner: Computers equipped with artificial intelligence will dominate laptop options for companies

Indeed, the silicon area in the average notebook core processor and smartphone processor has already increased by 5% and 16%, respectively, to accommodate the neural processing engines built into the device. Bain predicts that these products could increase demand for upstream components by 30% or more by 2026.

Packaging is the next part of the Internet, and if demand for GPUs doubles by 2026, suppliers will need to triple their production capacity. Additionally, different power and cooling requirements connect each part of the process to utilities, which will also need to scale based on demand.

The latest global chip shortage

Since the beginning of the current generative AI boom, chipmakers have thrived. Leading graphics processor vendor NVIDIA announced record revenues of $30 billion (£24.7 billion) in the second quarter of 2024, with a stock market value of more than $3 trillion (£2.2 trillion). Switch maker Broadcom and memory chip maker SK Hynix have had similar success.

SEE: Nearly 1 in 10 companies will spend more than $25 million on artificial intelligence initiatives in 2024, according to Searce report

These record profits were achieved only by a few major companies that control large parts of the supply chain. NVIDIA, an American company, designs most of the GPUs used to train AI models. However, they are manufactured by the Taiwanese company TSMC. TSMC and Samsung Electronics are also the only two companies that can produce cutting-edge chips on a large scale.

However, everything in the industry was not always so simple. A global chip shortage emerged in early 2020 as a result of the Covid-19 pandemic. Supply problems among this relatively small number of companies have persisted for more than three years, affecting industries such as consumer electronics and artificial intelligence.

Even before the pandemic, the semiconductor supply chain was on shaky ground due to a number of events, including trade wars between the United States and China, and Japan and Korea, that affected the prices and distribution of goods. Additionally, natural disasters such as drought in Taiwan and three factory fires in Japan between 2019 and 2021 have contributed to raw material shortages.

“Extreme weather, natural disasters, geopolitical conflicts, pandemics and other major disruptions over the past decade have made all too clear how supply shocks can severely limit an industry’s ability to meet demand,” the Bain and Company report said .

The desire for AI sovereignty could worsen chip shortages

It’s not just a lack of manufacturing capacity that could lead to a second global chip shortage.

“Geopolitical tensions, trade restrictions and the separation of international technology companies’ supply chains from China continue to pose significant risks to semiconductor supplies. Factory construction delays, material shortages and other unpredictable factors can also cause tipping points,” the report said.

For example, the United States has applied chip export controls to semiconductor sales to China, as well as the Netherlands and Japan. The UK also blocked most licensing applications for companies seeking to export semiconductor technology to China in 2023.

China’s Ministry of Commerce also announced it would implement export controls on gallium and germanium-related goods “to protect national security and interests.” These rare metals are essential in chip production, with China producing 98% and 54% of the world’s supply of gallium and germanium, respectively.

Governments around the world are also spending billions of dollars to expand their own semiconductor production capacity, with the main reason being to reduce their dependence on other countries. However, data security also plays a role; by keeping the supply chain within their borders, authorities can better protect against espionage and cyberattacks.

In 2022, the United States passed the CHIPS Act to provide essential investments in semiconductor research and manufacturing incentives, and to strengthen the U.S. economy, national security, and supply chains. The White House also launched a draft Artificial Intelligence Bill of Rights to help regulate artificial intelligence at the national level and invested in a proof of concept for a shared national AI research infrastructure.

Intel, TSMC, Texas Instruments and Samsung – the world’s largest memory chip maker – have announced plans to build factories in the US

In August 2023, it was announced that the UK government would commit £100 million ($126 million) to support the development of AI hardware and address possible computer chip shortages. Just this month, Amazon Web Services announced plans to invest £8 billion in data centers in the country over the next five years.

WATCH: UK government announces £32 million for AI projects after withdrawing funding for supercomputers

The European Union offered 43 billion euros ($46 billion) in subsidies to boost its semiconductor sector under the European Chip Act, which was passed in July 2023. The bloc also has a lofty goal of producing 20% ​​of the world’s semiconductors by 2030,

But Anne Hoecker, head of Bain’s global technology practice, said the pursuit of data sovereignty would be “time-consuming and incredibly expensive.”

In a press release, she stated: “While in some respects these projects are less complex than building semiconductor factories, they require more than just securing local subsidies. Hyperscale companies and other large technology companies can continue to invest in localized AI operations that will provide significant competitive advantages.”

Bain’s report adds that small language models with algorithms using RAG, or search-assisted generation, and vector embeddings can benefit from data sovereignty because they handle many of the compute, networking, and storage tasks close to where the AI ​​data are stored.

Guidance for AI supply chain executives on dealing with chip shortages

Bain’s report makes several recommendations for semiconductor companies on how to survive the next global chip shortage:

• Gain a deep understanding of and track the entire AI supply chain, including data center components, PCs and smartphones, and peripherals such as routers and networking equipment.
• Sign long-term purchase agreements to protect access to chips in the event of potential disruptions.
• Design products to use industry-standard semiconductors rather than application-specific chips to maximize cross-vendor compatibility and sourcing flexibility.
• Strengthen your supply chain in the face of geopolitical uncertainty, such as tariffs or regulations, by diversifying suppliers and sourcing components from multiple regions.

The authors of the report wrote: “Executives may continue to feel fatigued by the pandemic-induced semiconductor supply disruptions, but there is no time to rest as the next big supply shock approaches. This time, however, the signals are clear and the industry has a chance to prepare.

“The way forward requires vigilance, strategic foresight and rapid action to strengthen supply chains. With proactive measures, business leaders can ensure their resilience and success in a world increasingly leveraging artificial intelligence.”