However, clear evidence that AI’s capabilities can boost productivity and increase profits are still scare. Microsoft’s Copilot and Apple Intelligence are two early attempts to bring everyday productivity boosts to businesses and consumers, and thus their effectiveness will be heavily scrutinized.

In our view, both products should be viewed similarly to the original IBM personal computer (PC), released in 1981. The first PC was a tool with a mere 16 kilobytes of memory (only 0.002% of the memory in an 8 gigabytes iPhone), floppy disks for storage and a handful of software applications. But demand for this novel and potentially transformative technology was strong. Third-party software grew rapidly, steadily increasing the PC’s usefulness to the point that few of us can now get through our workday without our computer and smartphone. A similar example is the Internet itself. When the first web browser was built there was very little content to browse on the web, there were no search engines and one still had to type “www” before a web page’s name.

Like the PC and the Internet, AI will likely change not only how we work, but what we do. The International Monetary Fund (IMF) has estimated that 60% of all jobs in developed countries will be disrupted by AI². While the IMF intended its estimate to warrant greater caution and governmental oversight, the transformational impact of technology on jobs has been going on for decades. According to a study by a group of Massachusetts Institute of Technology (MIT) economists, more than 60% of today’s U.S. job occupations didn’t exist in 1940³. And technology has made other job occupations obsolete. Software applications like Microsoft Word and Google Docs made the occupation of typist virtually unnecessary and greatly impacted the demand for secretarial services.

The lag between innovation and productivity is shortening

Both the PC and the Internet took a long time to become indispensable mainstream technologies that boosted both business and personal productivity. A recent study found that with each subsequent innovation, productivity gains were realized more rapidly. While the choice of innovations in the study is somewhat anecdotal (steam engines in 1769, electricity in 1880, PCs/Internet in 1981 compared with AI in 2023), each saw a halving of the time it took from conception to a demonstratable boost in productivity. For example, steam engines took 61 years to reach productivity, electricity took 32 years, PCs and the internet took 15 years. As such, it was estimated that it could be a mere seven years from now that AI has a similar impact on productivity that the PC and the Internet did⁴.

While seven years may be optimistic, productivity gains inevitably follow transformative innovation—but the length of the lag between these events is hard to predict. For example, it took years after the dot-com boom (and bust) before smartphones were affordable and telecom companies built out the third generation (3G) infrastructure they needed to shine. But, as the hardware and software were developed, everyday uses (such as email, mapping, social media consumer sales) became widespread and ultimately ubiquitous. That said, this proliferation and adoption take time. If you can recall, the first iPhones (launched in January 2007) didn’t have access to the App Store which launched in July 2008. 

In our view, AI models still need to improve, evolve and specialize, with people needing to learn how to utilize the technology. Software applications still have to be envisioned and developed to maximize AI’s strengths. But even with the tools available today, businesses are increasingly eager to engage. In a recent poll, 94% of corporate chief information officers expect to use Microsoft generative AI products over the next 12 months, with Copilot leading the list at 68%5. Bottom line: we are still in the early days of AI, but we have little doubt it is on track to be a transformative technology, similar to the PC, Internet or smartphone.

More investment will come

Enough money has been spent on AI that some have suggested that the stock prices in the technology industry are fated to repeat the tech bubble and burst. While the spending is staggering, the fact remains that valuations of the largest companies making these investments are nowhere near the extremes seen in the early 1990s. The mega-cap companies currently fueling this investment wave are highly profitable, with low debt burdens. The 20 companies spending the most on AI investment today have earnings before interest, taxes, depreciation and amortization (EBITDA) margins near 26% and a net debt-to-EBITDA ratio below one. But the top 20 spenders in 1999 (just before the bubble burst) had EBITDA margins near 16% and a net debt-to-EBITDA ratio of nearly 2.56. In short, the AI investment boom has so far been funded by cash, not debt. Their investment may thus mean less net profit for these companies in the short term, but does not signal greater financial fragility.

Despite the recent dramatic surge in many of these company’s stock prices, their rise also remains below the extremes of the late 1990s. As can be seen in the chart below, the NASDAQ 100 Index® rose approximately 200% in the last five years as AI investments gathered steam. But before the 2000 correction, the NASDAQ 100 Index had risen more than 1,000%.

AI’s momentum will grow

We do not doubt—despite the recent correction in AI-related technology stocks—that the investment in AI will remain robust. A central part of this belief is that we see no greater opportunity for business investment that could match the breadth of AI’s potential to boost future net revenues. Robotics could be a close second, but the largest technology companies have already made their infrastructure commitment to AI, and the application development phase to maximize its uses has already begun. We believe further adoption of AI and innovations that will improve its effectiveness are inevitable.

Furthermore, with large technology firms already assuming the bulk of AI’s infrastructure investment costs, the cost to use AI today for end consumers is minimal. Consider that AI’s computation cost, as a result of significant hardware and software improvements, fell roughly 10x from 2017 to 2020⁷.  While this pace may not be sustainable, many consumers and businesses may be able to use an increasingly more productive AI tool for free or at a minimal cost for the foreseeable future. We believe this will help create a virtuous circle of increased demand through more useful AI services. 

While it is difficult to quantify the likely rate of AI adoption and its effect on productivity, it is clear to us that the technology has sufficient momentum to fuel both an investment boom and eventual productivity gains.

Custom software applications will fuel productivity

To take advantage of the infrastructure that energizes AI and to add quantifiable value, we believe companies need tailored software applications. Consider the value created in the Software as a Service (SaaS) expansion as a relevant precedent. Applications provided by Salesforce, Inc. and other leading software providers quickly emerged as productivity-enhancing tools for a wide range of companies.

Anecdotally, companies our analysts and portfolio managers at Thrivent directly cover have dozens, if not hundreds, of productivity-enhancing concepts that are starting to enter the development phase. While some of these concepts may prove more beneficial than others, we have little doubt about these companies’ commitment to harvesting the lowest-hanging, productivity-enhancing fruit using AI. 

Investing in AI innovation

There is justified debate about the sectors, industries and companies that are best positioned to see the most productivity gains from AI. There are simply too many unknowns and idiosyncrasies to generate consensus on AI’s economic impact across the Fortune 500. For example, it will take time to determine which jobs can be automated cost effectively. And there may be positive surprises for companies as further innovations in either AI’s underlying models or its use cases emerge that particularly benefit one industry, process or labor force.

However, we believe investment will continue in AI infrastructure. This includes semiconductor manufacturers and other ancillary manufacturers that are needed to build the necessary data centers, such as electrical equipment manufacturers as well as engineering and construction companies. The so-called “hyperscalers,” which are companies that can build the dynamically scalable computing infrastructure that AI processing requires, are also likely to remain aggressive in their growth plans. This includes companies like Amazon.com, Microsoft and Alphabet Inc.—larger companies that have the capital to rapidly scale the computing and storage services users will need.

Additionally, we believe the implementation of AI on smartphones and PCs is likely to catalyze a significant upgrade cycle. For example, Apple Inc. announced that its upcoming AI features will only be available on its newer iPhone models and many buyers may choose to upgrade to the latest model to ensure they can fully benefit from its AI features. According to a survey, 25% of Galaxy S24 users are making the purchase for its AI features⁸. As more apps are developed that take advantage of these functions, we can expect greater demand for the phones that provide them.

Finally, we believe investors will focus their attention on the companies that can create the software applications that maximize the power of AI to address the needs of enterprise companies. While we continue to expect more companies will look to build applications in house, change takes time and established software companies like Microsoft Corporation have already demonstrated their ability to add AI features to existing applications. However, we expect the ease of building custom applications with AI will improve and gain more traction, potentially creating a challenging environment for enterprise software developers.

Turning to specific sectors and industries, we believe AI technology has and will continue to have a wide range of use cases. In the near term, it may be more helpful to focus on companies with particularly high exposure to the kind of services that AI may be able to provide. For example, AI may be able to dramatically reduce labor costs in areas such as customer service, data entry or programming.

Ultimately, we expect AI will expand its impact to an increasingly wide range of industries, including manufacturing, health care, financial services and energy. While all companies share a thirst for efficiency, improved processes and reduced labor expenses, one of AI’s strengths is its applicability for an enviably wide range of tasks. For example, in health care, AI has already demonstrated its strength in interpreting medical imaging, but the technology may also be as useful in processing insurance claims or maintaining patient records. Similarly, the energy sector could benefit from AI’s assistance in geomorphological modeling as well as energy usage patterns and trends.

Given the breadth of roles AI can provide, we believe robust bottom-up research and active management can and will play a key role in identifying the individual companies more likely to economically benefit from their approach to integrating AI.

Potential headwinds

AI depends on enormous amounts of data, and there are already high-profile disagreements about who owns the content AI is digesting and what it is worth. For example, The New York Times sued OpenAI, and the AI company has subsequently agreed to licensing agreements for content from other major media outlets. So, while progress is being made, and we remain optimistic that it is in everyone’s interest to resolve the barriers, legal or regulatory hurdles may yet arise, slowing progress.

There are also more practical risks, including a sufficient supply of semiconductors and the potentially rising costs for the energy resources data centers require. And geopolitical uncertainties heighten both of these risks. For example, a large amount of AI’s infrastructure is currently based in Taiwan. While we do not expect tensions with China will disrupt access to these critical services, some risk remains that tensions grow, potentially disrupting access.

Additionally, there is a risk that it takes longer than we or markets expect for companies to figure out the best way to use AI. No  killer application has yet been identified to influence the industry, and no clear transformative examples of firms generating higher productivity and increased earnings have been documented. While there is broad agreement that AI will be useful, it could take time before replicable examples allow broad-based efficiency gains.

Finally, there remains the perennial concern that, when asked, AI may simply refuse to open the pod bay doors, in reference to 2001: A Space Odyssey.

Opportunities abound

The infrastructure investment that has already been made that powers AI, rapid improvements in AI’s performance, the lack of a compelling alternative to achieve the expected scale of efficiency and the wide range of possible uses suggest to us that the AI investment cycle is just beginning.

Consumers, small and large companies, governments and software developers are likely to identify new ways of using technology’s latest innovation. While some will benefit from being first-to-market, we expect the rising wave of AI to lift all boats, providing a strong incentive to stay invested in U.S. stocks. 

Media contact: Callie Briese, 612-844-7340; callie.briese@thrivent.com

The Fortune 500 is an annual list of the 500 largest U.S. companies by revenue.

The NASDAQ 100 Index includes 100 of the largest domestic and international non-financial securities listed on the NASDAQ Stock Market based on market capitalization. Any indexes shown are unmanaged and do not reflect the typical costs of investing.

EBITDA, or earnings before interest, taxes, depreciation, and amortization, is an alternate measure of profitability to net income. By excluding depreciation and amortization as well as taxes and debt payment costs, EBITDA attempts to represent the cash profit generated by a company's operations.

The net debt-to-EBITDA ratio is a debt ratio that shows how many years it would take for a company to pay back its debt if net debt and EBITDA are held constant.

¹ Nestor Maslej, Loredana Fattorini, Raymond Perrault, Vanessa Parli, Anka Reuel, Erik Brynjolfsson, John Etchemendy, Katrina Ligett, Terah Lyons, James Manyika, Juan Carlos Niebles, Yoav Shoham, Russell Wald, and Jack Clark, “The AI Index 2024 Annual Report,” AI Index Steering Committee, Institute for Human-Centered AI, Stanford University, Stanford, CA, April 2024.

² “Jobs may disappear: Nearly 40% of global employment could be disrupted by AI, IMF says” CNN Business. January 15, 2024. https://edition.cnn.com/2024/01/15/tech/imf-global-employment-risk-ai-intl-hnk/index.html. August 8, 2024.

³ David Autor, Caroline Chin, Anna Salomons, Bryan Seegmiller. “New Frontiers: The Origins and Content of New Work, 1940–2018.” Quarterly Journal of Economics. March 2024.

⁴ “How AI can boost productivity and jump start growth” J.P. Morgan Private Bank. 2024 https://privatebank.jpmorgan.com/nam/en/insights/markets-and-investing/ideas-and-insights/how-ai-can-boost-productivity-and-jump-start-growth. August 6, 2024.

⁵ Morgan Stanley 2Q24 CIO Survey Takeaways Maintaining GenAI Leadership Position, July, 11 2024.

⁶ “How AI can boost productivity and jump start growth” J.P. Morgan Private Bank. 2024.

⁷ New York Life, Artificial Intelligence: from imagination to investment, 2024.

⁸ Sumit Adhikari. “Galaxy AI plays key role in Samsung’s Galaxy S24 sales surge.” Android Headlines. April 11, 2024.