Capital Markets Update #12

We ran into a surprising chart earlier this week which noted that the S&P 500 ranks 24th on a list of top 30 global equity market performers YTD with a negative 1.0% return (pre-Iran conflict selloff).  We found this surprising.   Consider the massive capital investment and undeniable innovation taking place in our domestic economy at the moment.  Add to this our constructive capital markets environment, unprecedented consumer net worth gains (59% gain since Q4 2019, FRED), supportive real wage growth, a 4.3% unemployment rate, and strong corporate balance sheets.   

You’d think that would set up to a stable, bullish market and, obviously, it hasn’t.  The culprit is an unknown, unquantifiable and undoubtedly disruptive AI threat to the business models underpinning much of the broader S&P 500 market cap.  What likely gets less press is the fear impulse sitting right below this surface concern, summed up by a potential reality where human capital is less valuable in the future than it is today. If this threat comes to pass, will we all effectively become day laborers in service of Gen AI?  It’s almost embarrassing to write about, we’re not a fan, but it’s the narrative we’ll knock around here.

We’re don’t pretend to have all of the answers as to the extent to which AI will disrupt existing business models.  Accordingly, does it make sense to deflate equity multiples across a few sectors?  Sure, we suppose that’s rational.  However, we’d probably say the same thing if the market gained 4% YTD.  We’re not going to add our voice to the chorus of investors, researchers, theorists and part-time day traders taking sides (including sitting on the bench) as it relates to the magnitude of the AI disruption. Is EBITDA at Charles Schwab, JPMorgan, Crowdstrike, ServiceNow, Salesforce, Meta, Nvidia, etc. going to grow faster in the go-forward five years than it has in the past?  The problem with making a determination here is we’re operating with a dearth of data, drawing broad conclusions from relatively unique, narrow and often conflicting sources of information.  That’s risky ground on which to make a decision, as we see it, so we try and stay calm and wait for a more thorough information suite to set our stance. 

One critical question underpinning the next decade will be the supply-demand balance for labor in light of the expected productivity surge associated with broad-based adoption of AI across the enterprise value chain.  Therein lies critical assumption number one; but, assuming widescale AI adoption does take place, we’re not overly concerned for the fate of humanity.  History tells us humans are pretty adaptable, albeit with a slight lag.   

From 2007 – 2020, total enrolled computer science and computer engineering students increased from about 50,000 (2007) to 100,000 (2013) to nearly 210,000 (2021 – CRA Taulbee Survey Data).  So that’s 300% growth in just over a decade.  According to the National Center for Education Statistics, enrollment in Business / Finance degrees nationally has experience less than 10% growth over a similar period (2010 – 2020).  The only other major of statistical significance to see comparable growth to computer science / engineering degree enrollment would be health services, which has experienced 87% growth in enrollment from 2010 – 2020.  This is an example of would-be job seekers adjusting to what was then projected to be the most stable and economic source of labor demand four years out.  Assuming aggregate labor demand does not shrink materially (critical assumption number two), there’s no reason to believe AI’s impact on the economy writ large will not get pushed down through the educational system and, within a period of 5 – 10 years, force a rise in an altogether new field of study or renewed emphasis on an existing field.

Reflecting on aggregate labor demand during historical periods of technological and societal change is actually quite instructive, we believe.  We took a look at an interesting period in history not often referenced in current discourse.  During the first industrial revolution in the United States (1800 – 1860 ish), as modern machinery created efficiency in the conversion of agricultural and other man-made products into consumer goods, the US economy experienced dramatic, fundamental change.  The United States, previously a largely agrarian society, grew up and welcomed an industrial age.  An MIT article (Ricardo, Thompson - 2024) cites that real wages for handloom weavers (spinning cotton into fabric) more than halved from 1806 – 1820.  It basically asserts that while productivity enhanced the economy, it didn’t do so equally for all laborers.  However, a separate analysis conducted by Thomas Weiss (1990) notes the total US labor force nearly doubled from 1.7MM workers in 1800 to call it 3.1MM workers by 1820, and then nearly quadrupled to 11.1MM workers by 1860.  From 1800 – 1820, estimated US GDP increased 27%, on its way to literally doubling from 1820 to 1860.  Forest for the trees, MIT, forest for the trees.  Now, the industrialization of labor didn’t replace the most unique of human functions, which is our ability to generate human capital.  That is, our ability to literally conceive of the industrial revolution in the first place.  Some say AI will do just that, to which we say sure, prove it.  According to many, we were all supposed to be flying around in cars twenty-five years ago.  It’s not a perfect example but our point is don’t forget, these futurists, regardless of professional prominence or societal significance, do not have a perfect track record.  Instead, think about how laborers must have felt in 1800 as the first sawmill, steam engine, or cotton gin was invented; it must have felt like a labor market Armageddon.  They must have said to themselves “what the heck am I going to do now, I don’t know how to work those machines” – and here we are.  We work largely with our brains now, do you actually think we’re just going to stop thinking in a decade because of ChatGPT?  Humans use tools very very effectively.  We’ll take the over on humanity’s ability (in the aggregate, it won’t be entirely perfect) to leverage Gen AI like any other tool.  Yes there will be some reskilling, yes that has happened many times before, no that’s not a bad thing.   

The oft-circulated indeed jobs data shows, despite announced layoffs at major tech companies over the last year, that job postings for Software Engineer roles are actually outperforming the broader labor market by about 15%.  Software Engineer job postings on indeed have recovered to mid-2024 levels while the broader labor market has seen a relatively linear decrease in job postings over the trailing 24 months.  A segment of the market, largely made up of those supportive of the thesis we’re putting forth now, points to this data as empirical evidence that there’s plenty of demand for both AI and human capital in this new world.  Again, we think of this as more noise than signal, just as Block firing half of its labor force can also be looked at as a bloated fin-tech company returning to pre-pandemic labor levels.   We think talk about the fragile state of the US labor market is a discussion worth having, but likely less critical than journalists would have you believe.  For reference, the latest JOLTS data shows we’ve got about 400k fewer job openings today than we had in Q4 2019 and Openings divided by Unemployed Workers is now below the important 1.0x threshold at a reading of 0.87x (Dec 25 data, FRED and BLS).  But, we have to look at this in the context of the current labor market environment.  The current employment environment is literally the third tightest jobs market in the last 50 years (2019 and 2022 got down to 3.6%, 2000 got down to 3.9%).  Before that, you’d have to go back to 1969 to see a tighter labor market (FRED).  Many would disagree, but the reality is we could be sightly over-employed at present, leading to the marginal unit of employment pressure driving unemployment naturally upward towards stasis vs downward.  Many companies objectively and admittedly misread the tea leaves and over-hired in 2021.  So, while worth watching, JOLTS data and corporate labor pruning trends all need to be contextualized with these facts in mind.

Geopolitics aside, we are excited about the current state of economic potential within the US and how new productivity tools will reshape, not degrade, the labor market in years to come.