The semiconductor industry holds significant strategic importance, with the rise of artificial intelligence (AI) further amplifying its significance. AI's progression relies on increasingly advanced computer chips, intensifying the implications. As AI generates new indispensable tools, it is poised to become a fundamental element of the modern economy, businesses, and careers. This article delves into the initial strides being taken towards capitalizing on generative AI.
While the discourse surrounding AI has emerged rapidly in the public sphere, the underlying technologies, such as the internet, smartphones, and powerful computer chips, have taken decades to mature. Initial AI iterations primarily revolved around replacing or expediting tasks like customer service chatbots, automatic translation, and enhancing search engines. However, with the growing capabilities of generative AI, these tools hold the potential to enhance human learning and productivity. They leverage extensive data across various industries while democratizing creativity by granting access to new code for everyone. Anticipated consumer applications encompass computer vision, natural language comprehension, tools for expediting coding, and other instances where automated decision-making by vehicles or robots is crucial.
Despite extensive discourse about the promises and perils of AI, its adoption within enterprises remains largely experimental. Consequently, investors are uncertain about which business models will prosper, and corporations lack clarity on how to monetize this technology. Nevertheless, the pace of adoption is poised to accelerate as heightened interest translates into novel solutions.
In a recent turn of events, Nvidia Corporation, a US-based computer chip manufacturer, astonished the market with sales and earnings projections extending into 2024, far surpassing expectations. Consequently, its market capitalization neared the USD 1 trillion mark. Historically focused on computer gaming, Nvidia's foray into producing intricate chips pivotal for constructing AI models within data centers propelled its market capitalization to rank just behind Amazon and ahead of Meta (formerly Facebook).
Initially, this seems to contrast with the general semiconductor market trends. The surplus of computer memory chips due to high inventories and reduced consumer demand for computers and mobile phones has dampened the semiconductor market in 2023. This oversupply mainly affects less advanced memory chips. However, a distinct narrative emerges when considering Nvidia's specialized domain of graphics processing units (GPUs), which constitute about 80% of the market.
Nvidia's focus on GPUs brings a distinct perspective. These specialized GPU "cores" efficiently handle numerous memory-intensive computational tasks in parallel. Historically, these GPUs were primarily employed for gaming and video editing purposes. Nonetheless, the evolution of generative artificial intelligence heavily relies on training and inference within its machine learning algorithms. These computationally intensive processes heavily lean on GPUs. As a result, the pursuit of AI development will likely continue to propel advancements in semiconductors, spanning calculations, storage, and data transmission capabilities.
After machine learning models have undergone their training phase, a significant portion of the data processing tasks performed by AI applications can be offloaded onto more cost-effective and energy-efficient microprocessors. Data centers, which consume around 1% of the world's power, are known for their substantial energy requirements. Consequently, the prolonged process of AI training is prompting cloud providers to enhance the hardware that operates these data centers.
Nvidia serves as an example of the foundational infrastructure that renders AI applications viable, marking the initial phase of value creation in this technology realm. Envisioning a second phase, we anticipate the emergence of generative AI offered as a service. This would involve bundling it with software products and integrating it into upgraded versions of current applications, compatible with personal computers and smartphones. Providers could then apply a premium to access these advanced offerings.
In the coming times, we can anticipate a rise in regulatory measures. A range of concerns exists, spanning from job displacement, often termed 'creative destruction,' which goes beyond the scope of this article. Nonetheless, artificial intelligence (AI) encounters obstacles related to trust and precision, issues concerning safeguarding intellectual property, and addressing data bias.
Emerging legislation will need to address one of AI's inherent weaknesses, which is its inability to provide an audit trail showcasing the factors that influenced its outcomes. While this might not significantly impact creative domains like film scripting, it holds significant liability implications for areas such as autonomous vehicles and medical applications.
Regulation will also need to grapple with the concept of 'AI alignment,' involving the responsibility of ensuring that generative AI aligns with human values. This aims to prevent models, for instance, from responding to potentially harmful inquiries. Presently, discussions about the potential misuse of these tools, including their potential for weaponization, cyberattacks, and disinformation campaigns, are unavoidable.
A preliminary step was taken during the May Group of Seven (G7) summit, where leaders agreed to establish a regular working group under the banner of the 'Hiroshima AI Process,' named after the summit's location. This step underscores the importance of addressing these challenges.
Investment and Strategic Considerations
The initiative undertaken by the G7 serves as a reminder that the realm of semiconductors extends beyond being merely a global industry; it has evolved into a matter of strategic significance for governments. This notion is most evident in the financial support legislated over the past couple of years to bolster production capacity. These endeavors akin to "nationalization" of semiconductor industries are poised to stimulate growth, foster competition, and propel innovation within the sector. It's noteworthy that establishing a cutting-edge fabrication plant, a process spanning approximately two years, commands an investment as substantial as USD 12 billion.
However, the intricacies don't end there. The fabrication of semiconductor chips entails a multitude of intricate stages, with a limited talent pool of scientists capable of engaging in the development and advancement of these technologies.
The United States' Chips Act, ratified in August 2022, stands as a commitment of USD 280 billion in subsidies across the coming decade. This substantial funding includes a portion of USD 39 billion earmarked to invigorate domestic semiconductor production, accompanied by additional tax credits totaling USD 24 billion. Similarly, in the European Union, the approved expenditure from April 2023 furnishes incentives to facilitate both public and private investments in manufacturing facilities catering to chipmakers and their associated suppliers. The financial outlay, estimated at around USD 47 billion, notably allocates USD 33 billion for the construction of new fabrication plants. Notably, Japan, South Korea, and Taiwan have also implemented measures such as tax incentives and subsidies to fortify their respective semiconductor sectors.
Meanwhile, China is actively engaged in bolstering its semiconductor industry with support rivaling the combined efforts of the US, EU, and Japan. An estimated sum of USD 143 billion in subsidies and credits is being allocated over a span of five years. While the US aims to curb the sale of advanced semiconductor chips to China, this restrictive approach might inadvertently compel the Chinese industry to embark on innovative pathways. Ultimately, this could lead to the creation of solutions that not only enhance productivity but also expedite progress over the long term.
Despite the multifaceted challenges rooted in ethics, geopolitics, and technology, we foresee that generative AI will serve as the breeding ground for the upcoming wave of startups. Much like the advent of the iPhone catalyzed an entire ecosystem centered around mobile applications, and the ascent of cloud computing gave rise to an entirely new realm of software enterprises.
From an investment standpoint, the worldwide semiconductor industry amassed revenues totaling USD 574 billion in 2022. Within this sum, 31% can be attributed to the most cutting-edge computing chips. In the immediate future, the diminished demand for less advanced chips is gradually chipping away at substantial inventories. Our preference leans towards semiconductor manufacturers catering to the burgeoning cloud market, thereby becoming intertwined with the progressions in AI and electrification. This aligns seamlessly with our overarching inclination towards superior technology firms as the economic landscape evolves.