On June 24th, Gerg Holland, the portfolio manager of Invesco's high-end client investments, posted that AI will drive changes in market share and profit capabilities across various industries in the entire economy, and people may not fully recognize the impact of AI on consumers. The current problem is not whether global AI consumption will see growth, but in the shift of market shares, which companies will benefit and which will be eliminated. Investors should not only focus on semiconductor or AI infrastructure, but rather look at the future AI services and software as a bigger and more long-term opportunity.
In the consumer industry, areas that are usually considered favorable, such as e-commerce, digital media, and shared travel, are now enhanced through artificial intelligence. Examples include using generative AI to summarize thousands of customer reviews into a short paragraph, optimizing streaming quality through AI based on internet speed and device capability, and implementing rule-based machine learning models to match drivers and passengers.
Gerg Holland believes that he is no longer only focusing on AI infrastructure, which was the factor that drove the stock market last year, but taking a more long-term approach. Although he holds semiconductor openings, he believes that overall, AI software and services are the more lasting and bigger potential opportunities.
How will AI as an investment theme play out? Gerg Holland believes that if viewed as a longer cycle, it will initiate three waves, and expects that this cycle will cover the next three to five years.
The first wave, data centers and semiconductors, is the current investment opportunity. Server capacity is vastly increasing, and funding comes from large-scale data centers (cloud computing service providers). This is a competition between large companies, limited to those with annual product development expenditures of over $50 billion. Capital expenditures have more than doubled in the past three years. In May alone, capital expenditures for 2024 were raised from $157 billion to $182 billion (an increase of $25 billion) (capital expenditures for 2023 were $128 billion).
Gerg Holland points out that this is an arms race, and it can be foreseen that related expenditures will remain high at least until mid-2025. In addition, computer and mobile phone chips can run better games faster, and car chips can enable more automated safety and self-driving.
At the same time, funds are flowing into GPU chip makers. Faster GPUs/accelerators will be launched this year, and the average selling price will further increase. Network tools, parts, electronics, and HVAC refrigeration solutions also benefit from the upgrade of AI data centers.
The second wave is AI infrastructure. Large language models (LLMs) and cloud service providers will sell AI solutions to each company. These companies are now investing for future returns because they recognize the attractive return on investment (ROI) of AI. Today, most publicly traded companies are evaluating AI "concept verification" projects to determine their ROI. Some will quickly deploy such capabilities, creating revenue for large data centers and LLMs. Most companies will begin to deploy their AI capabilities for themselves 12-36 months later.
The third and largest wave is AI software, services and peripheral equipment. Much like the cycle of mobile phones and the Internet, most wealth creation comes from the shift in market share, time efficiency and/or overflow growth brought about by increased productivity. Many deployments are long-term investments, and beneficiaries will only be seen in 2025 and beyond. AI content recommendations, content creation, and task/productivity assistants will change the way consumers interact and consume. In terms of automation, robots that perform tasks or vehicles that provide transportation services are driven by AI.
In addition, Gerg Holland believes that potential beneficiaries include electronic game developers and media content creators. Game developers can significantly improve coding speed and significantly reduce costs through AI encoding recommendations. In addition, the quality of user-generated content developed with AI tools is stunning and continues to improve. Invesco predicts that this will affect the value of media companies' content libraries.