Irony, Optimism, and AI

“Is AI in a bubble?” That’s one of the questions we’re asked the most these days. No doubt, the AI fervor has propelled the Magnificent Seven to dizzying heights. It’s not just these tech giants, though. Companies across sectors have been quick to highlight AI in their strategic discussions. AI is now discussed at least once on every S&P 500 earnings call. Some companies have even overstated their AI capabilities, leading the Securities Exchange Commission (SEC) to caution against AI washing and urge transparency about actual AI utilization.

Figure 1: Artificial Intelligence Mentions Per S&P Earnings Call


_{Source: Bloomberg, as of 12/31/2023.}

We’re optimistic about the potential for AI-driven growth and productivity enhancements in the next business cycle and beyond. But it isn’t unusual for investors in public markets to get ahead of themselves with new technologies. There’s often a lag between market speculation and the tangible realization of the technological advancements. The biggest benefits to growth from the 90’s tech boom arrived well after their stock prices fell.  

The bubble question has merit, but in my view the better question about AI right now is, “What does it mean for the real economy over the long term?” For example, and likely to the relief of many, technological advancements do not necessarily come at the cost of employment. Rather, new technologies often lead to the creation of jobs that previously didn’t exist. Approximately 85% of employment growth in the last 80 years can be attributed to new technologies.^{[ 1 ]} While the automobile industry rendered horse and buggy drivers obsolete and sent many a horse to the glue factory, it simultaneously generated new opportunities in manufacturing, sales, and repair services. Today, nearly 60% of workers are engaged in jobs that did not exist in 1940. ^{[ 2 ]}

As disruptive as AI seems, the traditional dynamics underpinning productivity growth are already present in the economy, and they’re unrelated to AI. We know from the 1990s, the last significant period of productivity growth, that tight labor markets and fixed investment work hand in hand. In our inaugural Connection, we wrote about how we expect the US to enter a virtuous cycle where tight labor markets spur investment, which in turn increases productivity while expanding the economy and keeping inflation in check. These seeds were planted pre-COVID when the US started running out of cheap labor. In 2018, the unemployment rate fell below 4% from the post-Global Financial Crisis (GFC) high of nearly 10% while wage pressure rose to a 10-year high.^{[ 3 ]}

Beginning in 2018, without a surplus of inexpensive labor, companies started investing for growth. Private fixed investment as a share of GDP recovered from post-GFC lows, and that initial wave is helping to drive productivity growth today. We see AI’s contribution to productivity further out on the horizon and being an additive factor to the traditional drivers present today.


Figure 2: US Productivity
(rolling 10-year average of annual percentage change)


_{Source: US Bureau of Economic Analysis and US Bureau of Labor Statistics, as of 12/31/2023. Tight labor markets defined as period when the rolling 10-year average of the prime age employment to population ratio is 78 or above.}


Productivity can be hard to measure, especially when it’s associated with technology. It was the famous economist Robert Solow who said in 1987 that “You can see the computer age everywhere but in the productivity.” And when you consider that Candy Crush has been downloaded more times than Microsoft Office, we should view the data with some skepticism.^{[ 4 ]}

But at the most basic level, productivity is a measure of efficiency of the inputs (capital, labor and technology) used to produce outputs (goods and services). After a wave of investment in the 90s, productivity boomed and pushed economic growth higher. Contrast that with the post-GFC environment when the economy was saddled with excess capacity and cheap labor and investment fell.


Figure 3: 10-Year Average Labor Productivity and Nominal GDP
(% year-over-year, seasonally adjusted)


_{Source: Bloomberg, US Bureau of Economic Analysis, as of 12/31/2023.Labor productivity represents non-farm business sector labor productivity. Post-tech bubble is from 3/31/2002 to 9/30/2007. Post-GFC is from 9/30/2009 to 9/30/2019.}


The next innovation wave, the widespread adoption of AI, is still out on the horizon. To get there requires large amounts of capital and expertise. Take the demand for data centers as an example. Currently, 98% of data centers are under lease, and to build one takes 18–24 months on average. Utility connections take additional time, with an average wait time for permitting of more than 3 years. It’s also expensive; on average it can cost upwards of $500M to construct a single data center^[5]

Demand for AI-related expertise is also booming. Engineers skilled in large language models command a 44% premium in median annual compensation compared to non-AI skilled software engineers.^{[ 6 ]} As Global Head of Data Science Matt Katz discusses in the Guest Column, a recent survey of approximately 200 Blackstone portfolio company executives found that greater than 50% said their primary challenges were sourcing skilled talent and understanding AI’s potential benefits for their business.