In this piece:
The emergence of ChatGPT only two years ago rapidly revolutionized human and machine interaction and just as quickly became a hot topic among investors. Since then, technology companies have rushed into developing products and services based on Generative AI models. It’s easy to understand why: Artificial Intelligence (AI) is transforming industries, improving productivity, and streamlining workflow efficiency. But with all new technological innovations comes oversight and unintended use. For investors, it is crucial to consider how—in combination with regulators—we can mitigate risk at each stage of the AI development cycle while promoting innovation to enhance business operations, create opportunities, and solve some of the world’s most significant issues.
Like previous technology innovation cycles, developing the infrastructure necessary to run AI models is a significant investment that technology companies must make before they can reap the rewards. Simultaneously, several critical considerations must be addressed for AI innovation to proceed efficiently, ethically, and sustainably with broadly accessible business and societal benefits. You can use the drop-down menu below to learn about the considerations impacting AI development, adoption, safety, and energy consumption.
AI Development Stack
With newly developed large language models (LLMs), Gen-AI greatly improved machine Natural Language Processing (NLP) capabilities. To ensure that machines generate accurate output, the AI models go through “training” by combing through large datasets, and running complex computations. This process requires faster data processing and super-computing powers to ensure the output is accurate and produced in an efficient and timely manner. Significant investments in computing hardware, networks, public cloud, and energy supply are building new and improving existing infrastructure. With these upgrades, models can be developed and adopted into software applications to integrate with existing systems and operate autonomously as intelligent agents, as visualized in Figure 1.
While AI applications are primarily accessible today through cloud and internet services for scalability, they are following the same trajectory as previous technological innovations, gradually becoming locally accessible through PCs and smartphones.
Figure 1. AI Development Stack
AI’s development is still in its early stages, with most investment funneling into the hardware space. Servers are in high demand, and chipmakers, key component suppliers, AI server providers, and cloud operators are outperforming the market. Continuous investment will allow AI hardware and infrastructure upgrades, which are needed to further evolve mega models and AI applications.
At this stage, software companies have identified customer needs for AI services and are in the customer engagement phase, encouraging adoption through various trials and implementation strategies. Once initial integrations occur, software developers can observe the models in use and spot areas for improvement. As improved AI models and adoption processes incentivize increased deployment, monetization is likely to follow.
Adoption – The Race to Capture AI Value
While 2024 saw a surge in AI adoption compared to previous years, adoption remains relatively modest across industries as organizations attempt to justify and capture AI value. The allure of AI’s promise to transform business processes and create opportunities still remains, with signs of small cost savings and revenue generation. However, the critical challenge organizations face is integrating AI into existing systems. As Figure 2 shows, highly digitized sectors with existing data infrastructure have been the fastest to adopt AI. As with all technological innovations, the risk analysis of adoption creates factors organizations must consider, such as high-level data governance, security, and ensuring ethical, accurate use.
Figure 2. Share of US firms using AI by sector, %
Source: Census Bureau, Goldman Sachs GIR
Energy Consumption
While AI’s value is primarily viewed through development and adoption, an external industry—the energy sector—is essential for AI to run. With the requirement for stronger computing power comes exponential energy demand for data centers. The inference phase, when models apply learnings to new datasets and generate conclusions, accounts for the majority of AI energy demand.
Nuclear power has been revived as the energy supply of choice for powering data centers. As a reliable and low-carbon energy source, nuclear energy allows technology companies to meet power needs without compromising their Net Zero commitments. While it may be the best option, radiation leaks, nuclear storage, and transportation remain environmental hazards.
Safety and Misinformation
As more companies adopt AI, the tension between responsible implementation and profit-seeking is a growing concern. With the vast amounts of data collected and stored for AI models to process, data breaches are increasingly likely across industries. The financial, healthcare, and communication services industries are the most exposed to data security risks, and healthcare is exposed to the costliest cyberattacks due to the abundance of patient information healthcare organizations must store. AI could improve cybersecurity technology and defense and generate sustained demand for cybersecurity firms, but it could also be used to develop more effective malware and cyberattacks.
AI also has the potential to generate inaccurate and misleading information. Flawed or biased datasets will produce more flawed and biased datasets, making data transparency crucial. The ability for users to command output also raises the issue of misinformation and deep-fake images, a major concern during 2024, when half the world’s population saw elections.
While new technology can improve efficiency and productivity, it can also be used maliciously. It is essential to identify use risks early so developers can build safety controls and limit misuse before their models are released. Safety considerations are lagging the pace of development, but recent regulations and investor engagement are encouraging steps toward better guidance for developers in AI risk mitigation and management.
Regulating a Rapidly Developing Technology
The EU has been leading the world in AI regulation efforts, and most markets have followed suit, opting for soft law surrounding AI as regulators seek a balance between fostering innovation and minimizing risk. Adopted in September 2024, the Global Digital Compact established the first global governance framework for AI design, use, and governance for the benefit of all. It embeds human rights and international law, prioritizes online child safety, and aims to make data more open and accessible.
While these regulatory measures minimize the societal risks of AI, regulating a fast-changing technology requires constant monitoring and updating as new capabilities and impacts arise. To ensure risks continue to be mitigated and responsibility is assumed for the development and usage of AI while new regulation is slow to keep pace, investors are looking at strengths in company AI risk management. Formulating and disclosing Ethical AI principles as part of developing and deploying AI demonstrates accountability and responsibility across the AI value chain and ensures self-regulation of data security, output accuracy, and transparency in testing and implementation. For those with limited AI disclosure and governance, shareholders have the power to engage with tech companies and encourage them to establish guiding principles.
For the Future
Investors and companies have a consequential role in ensuring AI development has net positive benefits for the world. When investing in AI and its transformative value, responsible and active investors must be mindful of the long-term risks the technology poses. Engaging early with industries across the AI value chain will help ensure that the world-altering developments still to come will create a positive impact for years to come.
This piece is intended as an editorial review of a survey of risk considerations and does not constitute investment advice, a solicitation, or an offer to buy or sell any security. Past performance does not guarantee future results. All investments involve risk, including the risk of loss of principal.
References to third-party research, publications, and data sources are provided for informational purposes only. Boston Common Asset Management does not warrant the accuracy or completeness of third-party information cited herein. Readers should consult original sources and qualified professional advisors before making any investment decision.
