Demand for Customized AI Hardware Drives $53.4B Growth in Market
The market for customized AI hardware is set to experience substantial growth, with projected revenues of $53.4 billion in 2023, according to research by Gartner. This represents a 20.9% increase compared to the previous year and highlights the rapid expansion of the industry. Leading hardware companies, including Nvidia and Intel, are spearheading this growth.
While public AI models like ChatGPT have garnered attention from consumers, enterprises are seeking specialized hardware to efficiently meet their specific requirements at a lower cost. As a result, there is a growing demand for high-performance graphics processing units (GPUs) and optimized semiconductor devices.
Alan Priestley, VP analyst at Gartner, explains that the rise of generative AI and the increasing use of AI applications across various sectors necessitate the deployment of AI chips. These chips will replace the current architecture, particularly discrete GPUs, for a wide range of AI workloads. The demand for customized AI chips is expected to continue growing, with projected revenues of $67.1 billion in 2024 and over $119.4 billion by 2027.
Nvidia has emerged as a major player in the market due to its powerful dedicated AI hardware and cloud architectures tailored for AI models. The company recently unveiled its next-generation GH200 Grace Hopper chip, which can be combined to create a supercomputer configuration called the DGX GH200. Nvidia’s latest hardware, such as the L40S GPU, enables accelerated training of trillion-parameter large language models (LLMs) like GPT-3.5.
Intel is also vying for a share of the AI market and has committed to AI dominance by 2025. The company has outlined a roadmap for AI chips, encompassing central processing units (CPUs), GPUs, and dedicated AI architecture. This includes the Gaudi 2 processor, which outperforms the competition in deep learning inference workloads, and next-gen Xeon CPUs like ‘Sapphire Rapids,’ which offer a ten-fold performance boost compared to previous generations.
Google, too, has made strides in AI hardware with its tensor processing unit (TPU) chips. These chips provide comparable performance to Nvidia’s L40S while being more energy-efficient. Google employed TPUs in a supercomputer cluster to train its own LLMs called PaLM and PaLM 2.
Despite the increasing spend on AI, Gartner’s research reveals that AI has not significantly impacted overall IT spending. The 4.3% increase in IT spending is largely driven by software and IT services. John-David Lovelock, VP analyst at Gartner, suggests that AI may be seamlessly integrated into software without a price increase and that tracking AI spending at the end-user level can be challenging due to its utilization of various technological channels.
In conclusion, the demand for customized AI hardware is driving significant growth in the market, with revenues expected to reach $53.4 billion in 2023. Companies like Nvidia, Intel, and Google are leading the charge in developing specialized AI chips to meet the diverse needs of enterprises. As the deployment of AI systems continues to expand across different industries, the demand for customized hardware is likely to grow further, shaping the future of AI technology in the years to come.
