A semiconductors supply chain disruption is looming with global chip shortages forecasted by 2026 on the back of strong demand for the chips to power AI, according to a report.
The last global shortage for semiconductors was during the Covid 19 period due to supply chain disruptions as economies closed their borders while demand increased due to remote working.
AI enabled devices put pressure on semiconductor
According to a report by Bain & Co, the growing use of AI across sectors is likely to trigger a scarcity of graphics processing units (GPUs) supply for data centers as demand for computing resources and large language models (LLMs) persist.
The continued adoption of AI-enabled devices is anticipated to drive significant buying of new personal computers (PCs) and smartphones, undeniably leading to constraints in semiconductor supplies.
As it stands, the semiconductor supply chain is intricate and requires a growth north of 20% to offset the likelihood of chip shortage, according to Bain & Co.
The AI wave in the world of technology is anticipated to outweigh supply creating serious hurdles in the supply chain. This comes as big tech firms have been snapping up GPUs, especially from Nvidia.
“Surging demand for graphics processing units (GPUs) has caused shortages in specific elements of the semiconductor value chain,” Anne Hoecker, head of the technology practice in the Americas at Bain, told CNBC by email.
“If we combine the growth in demand for GPUs alongside a wave of AI-enabled devices, which could accelerate PC product refresh cycles, there could be more widespread constraints on semiconductor supply.”
Hoecker.
According to the Bain & Co report, matching the semiconductor supply and demand has been tough, and could be worsened by the tech industry’s rapid evolution, the huge capital requirements, and considerable lead times.
It is against this background that the research firm has also warned chip suppliers to be proactive and beat this impending crisis.
Investments into AI sector adding to the increase in demand
Main cloud service providers are anticipated to grow their annual spending by 36 percent this year buoyed by largely investments in AI and enhanced computing.
The demand in GPU is expected to continue on an upward trajectory as venture capitalists continue to invest more capital into AI start-ups at a time when LLMs are also enhancing their competencies to process multiple data types that is text, images, and audio concurrently.
Based on Bain & Co’s projections demand for current-generation GPUs is set to double by 2026 and suppliers of these key components would be required to upturn their output north of 30%.
Advanced packaging and memory demand will be significant uptakers of the semiconductors.
For instance, Chip-On-Wafer-On-Substrate (CoWoS), the makers of packaging components would require tripling its production capacity by 2026.
In that regard, to pave for AI growth, supply chain essentials must come together to secure access to advanced packaging and sufficient electricity.
The supply chain of semiconductors is spread across companies, for instance, while Nvidia might design its GPUs these are made by Taiwan Semiconductor Manufacturing Co (TSMC).
To perform its operations effectively, TSMC relies on chipmaking tools from countries around the world, such as the Netherlands. Furthermore, the most cutting-edge chips can only be made at a large scale by TSMC and Samsung Electronics, according to CNBC.
However, the Bain & Co report also bring another dimension to the looming shortages, that might worsen the situation. The report cites the geo-political tensions as another factor that might trigger shortages.
“Geopolitical tensions, trade restrictions, and multinational tech companies’ decoupling of their supply chains from China continue to pose serious risks to semiconductor supply,” said Bain
Delays in factory construction, materials shortages, and other unpredictable factors could also create pain points according to Bain.
Source: https://www.cryptopolitan.com/semiconductors-shortages-loom-on-ai-demand/