Traditional central processing units (CPUs) were designed for efficient execution of general-purpose computing tasks in a single cycle. On the other hand, graphic processing units (GPUs) are meticulously tailored to process multiple, albeit less complex, logic-based computations in parallel. This attribute makes GPUs exceedingly efficient for data-intensive tasks, making them a favored choice for AI training and inference tasks.
In 2022, global spending on GPUs for AI acceleration-related applications neared $16 billion. Nvidia dominates this market, commanding nearly 80% of the GPU-based AI acceleration share. Nvidia’s investments in software frameworks, particularly the CUDA architecture, empower developers and engineers to harness computational efficiency from GPUs. This strategic advantage positions Nvidia ahead of competitors lacking comparable software support.
In addition to GPUs, the AI chip market also includes field-programmable gate arrays (FPGAs) and application-specific integrated circuits (ASICs) from companies like Infineon and Texas Instruments. FPGAs and ASICs contributed approximately $19.6 billion to the market in 2022, demonstrating a remarkable year-over-year growth of about 50%. Collectively, their spending is projected to grow at an annual rate of almost 30%, reaching a total of over $165 billion by 2030.
Escalating Chip Demand Spurs Heightened Investments
Driven by a scarcity of AI processing hardware, developers and engineers are racing to introduce AI-centric products to the market. This rush has led to a six-month backlog in GPU orders and substantial price increases for Nvidia’s A1000 and H1000 GPU line-up, with these chips fetching hefty premiums in secondary markets.
However, this high demand and associated price hikes are expected to benefit the entire semiconductor value chain. This includes foundries, chip designers, and semiconductor equipment suppliers. As AI becomes increasingly prevalent, multi-trillion-dollar sectors such as advertising, e-commerce, digital media, entertainment, online services, communications, and productivity are likely to amplify their investments in turnkey AI hardware setups.
Specialized AI Hardware: A Prerequisite for AI Advancement
The surge in AI’s popularity is poised to drive a wave of upgrades within data centers, consequently triggering a new investment cycle in semiconductor technology, with GPUs occupying a central role. The swift proliferation of large language models is set to generate exponential demand for AI processing power, thereby fostering a surge in specialized chip expenditure. This trend could open a market worth over a hundred billion dollars for the semiconductor industry in the near future. Notably, major cloud hyper-scalers are anticipated to continue investing in research and development, creating their own chips to reduce reliance on established chip suppliers and cut costs. While demand for AI chips may display some fluctuations in the short term, Global X is optimistic about the semiconductor value chain’s ability to seize this opportunity and create a viable investment avenue as AI penetrates novel markets.
