Artificial intelligence (AI) is top of mind for all businesses looking to grow. Economic researchers point to AI as a key factor to boost the economy, and believe that AI could result in billions, if not trillions, of dollars in technology-related spending. However, this potential growth will not be possible in isolation; it requires massive amounts of supporting infrastructure.
Below are some key infrastructure components that will be critical to the continued growth of AI and cloud computing. For these purposes, we are defining infrastructure broadly, including all of the equipment, structures, and facilities needed for the operation of everything surrounding and supporting AI and the cloud.
- Data centers: As more businesses look to implement AI and migrate to the cloud, the demand for data centers will continue to grow. For some businesses, it will make sense to have their own data center(s), but most businesses will rely on third-party data centers on a colocation basis. This shift is consistent with the overall migration to outsourcing of non-core and specialized functions and reducing the need to own and maintain related physical infrastructure and processes. With the current soaring demand for AI chips, it may be also impracticable, if not impossible, for a random business to procure and build out enough data center capacity. One of the open questions with respect to potential capacity issues is whether we will see regulators subjecting data computing capacity and their providers to the non-discriminatory access regulations same as the other essential utilities.
- Power: There is also a growing demand for the supply of electricity to power computing facilities and data centers, largely driven by the use of AI technologies. According to McKinsey, the energy supply is becoming a significant issue for the industry; even with sufficient power generation, connection of a new data center may take years due to the need to develop the grid. The industry is energy-hungry, with a medium-sized data center consuming as much power as several thousand households. In some instances, municipalities in different parts of the globe imposed a moratorium on new data center development to protect local grids from stress. In response, tech businesses seek to implement sustainable solutions to power their facilities through renewable sources and alternative solutions, including nuclear power. For more information, refer to our previous Insight.
- Connectivity: A reliable internet connection is another important feature for AI and cloud businesses, and network infrastructure may differ depending on the type of technology. Migration to the cloud coupled with limited capacity in certain markets will further drive international competition between data center service providers, which may put the capacity of submarine cables and other international infrastructure to the test. While a data center is stationary and its network needs are quite predictable, mobility and business-to-consumer sector users rely on wireless technologies to access AI and cloud technologies. Therefore, the availability of 5G networks and mobile solutions will play a substantial role in the continued growth of these sectors—which is likely to support the existing trend in the telecom sector of the use of towers on a colocation basis and the more prominent role of professional tower infrastructure companies, as well as further tower market consolidation. Morgan Lewis has experience advising both telecom providers and tower companies on portfolio acquisitions, master lease agreements, and portfolio management agreements. In addition, our lawyers regularly advise on deals involving other types of telecom assets, including metro fiber, fiber to the home, wireless, and submarine cables.
- Data: The input data used for training AI affects the output results. As such, AI businesses need to focus on improving the quantity and quality of input data. The use of publicly available online sources and data scraping by AI developers to train models is increasingly facing claims of copyright infringement, especially when the fair use defense is not applicable. To protect their copyright, some copyright holders include restrictions on the use of their works for AI training and embed technical solutions to prevent data scrapping. On the other hand, markets are emerging for platforms that connect copyright holders and AI developers, with the intent to offer high-quality content to AI platforms and secure royalties for copyright holders. A similar shift could occur as seen with the rise of streaming platforms, where industry maturation led to users being willing to pay reasonable fees for high-quality content. Data has been turning into a commodity over the last decade, and we may see further market developments in this direction.
- Talent: As one cloud provider stated, tackling the talent gap is its greatest challenge. It cannot be overstated: the demands of the coming AI era require new ways of training, but also the continuous training of employees, as new tech is developed and existing tech becomes outdated. Although some jobs may be lost to AI, it is also likely that the industry will create many more opportunities for those willing to acquire the necessary skills. Further, the model of education may change dramatically; while the concept of life-long education becomes more popular, educational institutions are likely to shift to short-term programs where students can master specific skills relatively quickly to keep up with the pace of the changing technology.
- Security: The rules applicable to dealing with sensitive and private data will continue to affect the industry. In addition to already existing regulations concerning the handling of personal data, we may see calls for stricter adherence to industry standards as it relates to detecting and preventing risks from the misuse of AI, including implementing zero-trust security models, requirements to conduct stress tests, anomaly identification, promptly responding to identified risks and reporting incidents, as well as implementing other risk mitigation policies. While initially the industry has heavily relied on self-regulation and best practices, we are likely to see more active intervention by governments and independent validation requirements with respect to security of the technology and data in AI-driven businesses. Some existing examples of regulations focusing on risks include the EU AI Act, NIST AI Risk Management Framework, Singapore Model AI Framework, UK Regulatory Framework for Artificial Intelligence, and the Council of Europe Framework Convention on Artificial Intelligence, Human Rights, Democracy, and the Rule of Law.
- Regulations: Some part of governments’ focus on regulating AI is driven by the privacy issues discussed above, but there is also a geopolitical context, as countries appreciate the economic advantages of AI for growing a national economy. For example, the United States is considering a new policy to cap chip sales to certain countries in the interest of national security. Other than export controls, legislative developments may further focus on AI technologies as being strategic assets or critical infrastructure, subject to limitations on foreign investments and antitrust authorities’ scrutiny.
- Finance: Obviously, developing the infrastructure requires investment. The availability of financial resources to support AI and cloud development will become an important competitive advantage. The industry may see the rise of designated lenders, funds, and consortiums of institutional investors focused specifically on AI infrastructure development.
If you are interested in hearing more about the current trends in infrastructure and their effects for the industry, we will be covering this and other AI-related topics at our NY Summit on October 31.