It’s AI All the Way Down

During a recent client lunch, the conversation turned to AI’s impact on education. The client, an educator, shared that college students are increasingly using AI to write papers. While students cheating on essays isn’t new, what struck me was the professor’s response. Rather than encouraging original work, her former colleagues have turned to AI for grading students’ AI-generated assignments. This creates a peculiar cycle where AI-generated writing is evaluated by AI, effectively removing human thought and learning from both sides of the academic process.

AI is quietly propelling the stock market to new heights, even as investor attention focuses on tariffs, economic data integrity, and political upheaval. Through the end of July, just three AI-driven stocks accounted for 55% of the S&P 500’s year-to-date performance: Nvidia (25%), Microsoft (20%), and Meta (10%).¹ Staying with the Western theme, the once-celebrated “Magnificent Seven” tech stocks have essentially become the “Three Amigos,” with AI fever concentrating market gains on an increasingly narrow group of companies. (Meta is clearly Martin Short)

We’ve been dealing with concentrated market leadership for years. It’s challenging to manage a diversified portfolio when one theme, and seven (let alone three) companies, drive performance. Portfolios are forced to accept elevated concentration risk to outperform, which contradicts the core tenet of modern finance: the so-called free lunch of portfolio diversification.

History shows that narrow market leadership may lead to increased volatility, particularly when accompanied by extreme valuations, euphoric sentiment, and crowded positioning. Many of these market signals are present today. While mega-cap technology stocks have driven recent performance, we maintain a risk-balanced portfolio allocation that prioritizes resilience over momentum chasing.

Are Storm Clouds Building?

It’s difficult for investors to consider AI’s investment implications without recalling the late-90s tech bubble. Recent enthusiastic AI research isn’t helping dispel those parallels. Investment bank Morgan Stanley recently published a report suggesting that global firms will invest $3 trillion in AI data centers between now and 2028, with half of this forecasted investment coming from the US.²

Setting aside the plausibility of such massive capital expenditure, AI data centers are notoriously power-hungry. According to the same Morgan Stanley report, the estimated power demand from their forecasted AI infrastructure would require 45 gigawatts (equivalent to 23 Hoover Dams worth of new power generation) over the next three years. This AI spending, based on a yet-to-be-built power supply, is expected to generate returns of over $1 trillion annually by 2028, up from $45 billion in 2024, representing a staggering 2,122% increase.²

It’s important to remember that Morgan Stanley is an investment bank positioned to earn substantial fees from underwriting AI-related debt. Their financial incentive lies in building investor enthusiasm around AI’s future profitability. Wall Street banks have a time-honored tradition of painting the rosiest possible picture to secure underwriting and advisory fees. Investors must determine where reality ends, and empty profit promises begin. This behavior from investment banks always exists but gains the most traction during periods of market excess.

We believe AI will indeed change the world. The technology is improving rapidly and is already impressive in many regards. Yet even good investments can deliver poor returns if investors overpay. History shows that investors consistently overestimate new technology’s short-term impact while underestimating its long-term potential. The recent release of Chat GPT’s new model, which experts found underwhelming, underscores this point.

With policy uncertainty clouding the outlook for much of the S&P 500, investors have understandably gravitated toward AI-related sectors, broadly categorized as TMT (Technology, Media, and Telecom, which we’re using for its historical late ‘90s popular usage). These sectors typically lack the direct tariff exposure of consumer discretionary stocks or the policy uncertainty plaguing healthcare. Their current weighting of 44.2% in the S&P 500 sits just below the all-time high of 44.7% reached in 1999. The long-term average of 26%³ highlights the dangerous concentration of investment risk within these AI-related names.

Market Reality Sets In

Investors are notoriously fickle. When paying record prices for earnings, companies must eventually deliver results. Two members of the high-flying Magnificent Seven learned this lesson recently. Tesla has fallen 30% from its all-time high following disappointing earnings and guidance⁴, while Apple has dropped 11% for similar reasons.⁵ The fact that market leadership has narrowed to just three stocks demonstrates that while investors remain willing to pay premiums for AI names, they still expect performance.

No industry remains fully insulated from economic reality. While AI infrastructure spending has continued despite policy uncertainty, fear of a genuine economic slowdown or increasing long-term interest rates could change that dynamic. The labor market provides the clearest window into economic health.

Labor Market Signals: Frozen, Not Broken

Estimating monthly job creation has always challenged economists. Figures can be revised for up to 18 months after the initial estimate. The Bureau of Labor Statistics’ monthly Employment Situation Summary represents their best estimate based on survey data from employers. It’s invariably wrong to some degree each month, with figures updated as additional survey data arrives. This is referred to as the revision process, but it’s simply seasonal adjustment refinement and the addition of late survey responses from businesses as well as state and local governments about their hiring activity.

The July jobs report, released August 2nd, hit markets with unusual force. Analysts expected 105,000 new jobs, but the report delivered just 73,000 while revising the previous two months downward by a total of 258,000 jobs (the largest revision since May 2020).⁶ Such disappointing headline numbers and large revisions typically signal business cycle turns.

However, we’re not convinced this report indicates a recession for several reasons.

First, the 73,000 headline number likely approximates what’s needed to maintain steady unemployment, given that immigration to the US has essentially stopped. Previously, economists estimated the US needed approximately 150,000 new jobs per month to keep unemployment stable. The apparent halt in immigration during the year’s first six months represents a remarkably swift downshift in labor market supply.

Second, the average workweek held steady month-over-month and has increased year-to-date.⁷ Average weekly hours serve as a crucial cross-check on employment data.⁸ Historically, employers reduce hours before layoffs. Similarly, average wages grew faster than expected and continue outpacing inflation. A labor market truly approaching recession would show both metrics declining.

Supporting evidence came from the August 5th ISM survey, whose employment component registered the weakest reading since 2023’s recessionary scare.⁹ Again, weak but not broken. Further, initial jobless claims have steadily fallen since June, suggesting employers neither want to fire existing workers nor hire new ones due to policy uncertainty and skilled labor shortages.

Rather than indicating recession, these data points collectively suggest a frozen labor market. Weak labor market growth implies slow overall economic expansion, consistent with analyst expectations of 1.0-1.5% annualized GDP growth. This represents weakness, but not recession. It’s also an indication that immigration policies are having a negative economic impact on growth.

Bringing It All Together

Under normal circumstances, a frozen labor market would prompt the Federal Reserve to cut short-term interest rates without hesitation to stimulate the economy and ease employment market bottlenecks. Full employment represents half of their mandate. The other, more challenging part of the Fed’s mandate is stable inflation, which appears to be moving in the wrong direction on most measures.

Core PCE (Personal Consumption Expenditures), the Fed’s preferred comprehensive inflation measure, has risen to 2.8% year-over-year, above their 2% target.¹⁰ Looking at shorter timeframes presents a more troubling picture. Inflation rates excluding volatile energy have been accelerating throughout the year, even before tariff announcements. Annualizing the most recent data pushes core PCE above 3%. Core Consumer Price Inflation, CPI, has also been accelerating for the past three months.¹¹ Inflation is headed in the wrong direction. Rate cuts when underlying inflation is seemingly accelerating generally prove counterproductive in the long run as it generally results in higher interest rates for longer term lending such as mortgages, as well as intermediate and longer-term bonds.

Equity investors are counting on multiple rate cuts later this year. Such expectations are required to justify the lofty valuations. Tariffs, while not expected to be nearly as impactful as COVID, don’t help the inflation side of the Fed’s mandate. The full price impact of tariffs is still working its way through global supply chains.¹² Yet, despite these challenges the market currently expects a rate cut from the Federal Reserve’s September meeting with 96% certainty.¹³

The Fed must choose which of their goals to prioritize. Even before the Trump era, the Fed generally gave labor market health preference over stable inflation when the two conflicted. If this precedent holds and the Fed begins a cutting cycle, the most likely outcome is not recession, but mild stagflation. Stagflation combines higher inflation rates with slow economic growth.

Stagflation represents one of the most challenging economic regimes for investors and businesses to navigate. With stagflation, prices increase but consumers lack sufficient income growth to maintain purchasing power. It’s like being stuck in economic quicksand. The harder policymakers work to fix one problem, the worse the others become.

The market areas that tend to perform best during stagflation are low-valuation, old-economy sectors like healthcare, energy, and consumer staples. These are exactly the sectors that have trailed the broader market for years. Stagflation won’t stop AI investments, but it will create headwinds for investor sentiment and portfolio positioning. Valuations in the technology sector could decline in that environment as it generally results in higher long-term interest rates, which make future earnings from growth companies less valuable in today’s dollars. Essentially, stagflation tends to rotate market leadership from high-growth, high-valuation sectors toward more traditional, defensive industries.

What’s the Stagflation Playbook?

Stagflation, even in mild form, represents a rare market phenomenon. It doesn’t occur frequently because the economic conditions required to create it are inherently contradictory. Large Federal deficits, poor labor market demographics, high debt levels, and tariff implementation may be the eclectic mix needed to trigger stagflation.

Unusual scenarios can produce unusual outcomes. In a stagflationary environment, both equities and bonds may struggle to deliver consistent absolute returns. Historical asset class relationships, where fixed income serves as a hedge against equity risk, may no longer hold true.

This breakdown creates compelling opportunities for alternative investment strategies that capture value beyond traditional stock and bond exposure. Broadly speaking, alternative lending, insurance, and equity market premium strategies can provide unique exposures that may actually benefit from stagflationary pressures while delivering non-correlated returns. While their use is not intended to be a savior, an allocation of 10% to an overall portfolio could prove valuable if stagflation becomes a market reality.

Concluding Thoughts

While artificial intelligence will undoubtedly transform the global economy over time, the market’s current pricing suggests that transformation has already occurred. With TMT sectors commanding 44% of the S&P 500 at historically extreme valuations, investors are paying up for substantial hopes and dreams.

The parallels to previous technology bubbles are striking but not necessarily predictive. Unlike the dot-com era’s pure speculation, today’s AI leaders generate substantial revenues and profits. However, the concentration of market gains in just three stocks, combined with Wall Street’s enthusiastic trillion-dollar projections, echoes the same optimistic fervor that preceded previous corrections.

Economic headwinds are building. A frozen labor market, accelerating inflation, and the likelihood of stagflation create a challenging backdrop for growth stocks trading at premium valuations. Meanwhile, potential catalysts for change, mainly one misstep from the Three Amigos, lurk beneath the surface of seemingly unstoppable AI momentum.

The investment challenge isn’t determining whether AI will change the world (we’re already witnessing that change). The question is whether current prices already reflect decades of future growth, leaving little room for error. History suggests that even transformational technologies can prove disappointing investments when purchased at excessive valuations.

Prudent investors should acknowledge AI’s revolutionary potential while remaining mindful of market concentration risks and valuation extremes. In an environment where traditional stock and bond relationships may break down, diversification into alternative strategies becomes increasingly valuable. The winners in any technological revolution are rarely the highest-priced stocks at the peak of enthusiasm, but rather the companies and investors who maintain perspective when others lose theirs.

As storm clouds gather on multiple fronts, the AI trade may face its first real test. How it responds will determine whether we’re witnessing the birth of a new economic era or simply the latest chapter in the timeless story of market excess.

1. MOV Equity Index Movers, 1/1/2025 – 7/31/2025, Accessed in August 2025.

2. Morgan Stanley, Powering AI: Capital, Power Bottlenecks and Mapping Adoption, July 24, 2025.

3. Technology, Media, and Telecom (TMT) Market Cap, as of 7/31/25, Accessed in August 2025.

4. Tesla (TSLA) US Equity, 8/12/2024-8/12/2025, Accessed in August 2025.

5. Apple (AAPL) US Equity, 8/12/2024-8/12/2025, Accessed in August 2025.

6. Bureau of Labor Statistics, The Employment Situation – July 2025, August 1, 2025.

7. US Non-Farm Payrolls, as of 8/1/2025, Accessed in August 2025; Average Hourly Earnings (AHE) Month-Over-Month Index, 12/31/2024-7/31/2025, Accessed in August 2025.

8. US Average Weekly Hours Total (AWH), 7/31/2019-7/31/2025, Accessed in August 2025.

9. Index US ISM Services Employment (NAPMNEMP), 7/31/2019-7/31/2025, Accessed in August 2025.

10. US Personal Consumption Expenditures (PCE), Price Index, as of 7/31/2025, Accessed in August 2025.

11. US CPI – Core, as of 8/12/2025, Accessed in August 2025.

12. US Effective Tariff Rates, Accessed in August 2025.

13. World Interest Rate Probability, Accessed in August 2025.

Data as of August 2025.

Intended for educational purposes only. Opinions expressed are not intended as investment advice or to predict future performance. Past performance does not guarantee future results. Neither the information presented, nor any opinion expressed constitutes a solicitation for the purchase or sale of any security. Consult your financial professional before making any investment decisions. Opinions expressed are subject to change without notice.