I’m Taylor Sohns, CEO of LifeGoal Wealth Advisors and a CIMA and CFP. The country is seeing a surge in artificial intelligence spending. That surge is now the key fuel for growth. In the first quarter, an estimated 75% of U.S. economic growth was tied back to heavy AI outlays. The question is simple and urgent. Does this massive wave of spending keep markets rising, or does it sow the seeds of a downturn?
AI spending is the only thing keeping our economy afloat. Seventy-five percent of U.S. first-quarter growth came from aggressive investment in AI.
We’ve seen cycles like this. Big ideas often bring big budgets. They also bring booms, busts, and shakeouts. That is why history matters here. It gives a playbook for how investment manias start, accelerate, and, at times, end.
The Surge In AI Spending
Capital expenditures tied to AI jumped after the launch of ChatGPT. The timing is not a coincidence. Companies saw a path to automating tasks, tailoring services, and processing massive data sets. They started ordering chips, building data centers, and hiring teams. The outlays show up in GDP. They also show up in earnings calls and supply chains.
A small group of tech leaders is driving much of the spend. Mark Zuckerberg is pushing Meta into AI infrastructure. Jeff Bezos’s Amazon is building cloud capacity and new models. Microsoft and Google are racing to power new AI tools for customers. This is a classic race for scale. The thesis is that whoever builds first and best wins the customer and the cash flows.
That drive has knock-on effects. Chip makers ramp production. Utilities plan for higher power demand. Real estate firms design buildings that can cool thousands of servers. The market loves the revenue now. The risk is whether the profits later match the price paid today.
View this post on Instagram
What History Tells Us
We have spent even more aggressively in past technology waves. Those periods offer a sobering yardstick.
- Railroads: Spending was about 3.5 times today’s AI pace under Cornelius Vanderbilt and peers.
- Dot-com era: Roughly 3 times current AI levels in the late 1990s and early 2000s.
- Electric motor and cars: Around double today’s pace in the early 20th century.
- Current AI cycle: Lagging those peak figures so far, but rising fast.
In three of those four historical surges, investors overspent. Too many tracks, too many websites, too many factories. Supply outran demand. Capital dried up. Recessions followed. That pattern does not doom the present wave. But it begs for discipline.
Two Clear Paths From Here
There are two main outcomes for this cycle. Both have precedents, and both could unfold over a few years, not a few months.
First, spending continues to climb, matching levels seen in past booms. That path can lift GDP growth and earnings for suppliers and platforms. If real use cases spread and margins hold, markets can continue to advance. Profits would justify the buildout. Winners would enjoy durable cash flows, and even second-tier players could benefit for a time.
Second, adoption lags the hype. The industry produces a glut of capacity, models, or applications that fail to pay for themselves. Pricing erodes. Balance sheets tighten. Layoffs start. Equity markets, which are priced in years of strong cash generation, reset hard. This is the railroad-and-dot-com script. It does not erase the tech. It resets the valuations and the roster of winners.
How To Judge Which Path We’re On
Investors can watch a few simple markers to tell the difference between progress and puffery. None is perfect. Together, they offer a cleaner picture.
- Unit economics: Are AI services sold at or above cost, and do margins improve with scale?
- Real adoption: Do enterprises move pilots into full deployments that cut costs or lift sales?
- Energy math: Can data centers secure power at prices that do not crush returns?
- Supply signals: Do chip inventories rise and lead times fall too fast, hinting at slack demand?
- Pricing trends: Are AI seats, tokens, or inference fees stable, or are discounts spreading?
- Regulation: Do rules raise compliance costs or slow release schedules?
If these markers trend positive, the runway lengthens. If they turn negative, caution grows. Most cycles swing between the two views before the direction becomes obvious.
The Role Of “Drunken Sailor” Spending
Many have called today’s outlays “drunken sailor” spending. That phrase captures speed and size more than precision. Big spending is not the issue on its own. The issue is the quality of the match between the dollars spent and the cash returned. In the railroad industry, many lines never covered their cost of capital. In the dot-com era, many networks became valuable later, but early investors lost money when business models failed.
Today’s leaders may be right to spend fast. First-mover scale matters in platforms. But even giants can miss. If too many firms chase the same workloads or models, returns fall. If power or cooling proves costlier than planned, returns fall. If customers hesitate and shift budgets back to core software, returns fall again.
The lesson is to separate scale from waste. The best owners track payback periods, not press releases. They run pilot projects with clean metrics. Then they scale only what works.
What This Means For The Economy
With such a large share of growth tied to AI outlays, the economy has more exposure to a single theme. That can be fine if the theme converts into lasting productivity. We want to see real gains in output per worker. We want lower unit costs for tasks like coding, support, search, and design. If those gains are clear and broad, the GDP boost lasts. Tax receipts grow. Wages can rise without stoking extra inflation.
If the gains do not show up, the risk rises. Companies may freeze hiring. Projects get delayed. Suppliers cut shifts. Such reactions can flip GDP from a tailwind to a headwind. History has shown that the hangover can be sharp when spending outpaces results.
Investor Playbook: Practical Steps
There is no single trade for a cycle like this. But there are sound habits that can tilt the odds.
First, focus on cash. Favor firms that generate free cash flow today and can self-fund projects. They are less likely to need new equity when markets turn. Second, look for pricing power. If a company must slash prices to win users, the model may not scale. Third, check capital intensity. Some parts of AI are asset-heavy. Others sell tools with lighter outlays. Mix matters.
Diversification also helps. Concentrated bets can work, but carry outsized risk in regime shifts. Balanced exposure across chips, cloud, software, and end users can smooth outcomes. Fixed income with strong credit quality can hedge equity drawdowns if a glut forms. A cash reserve offers optionality when prices reset.
Above all, avoid stories that need perfect conditions. If a thesis only works with endless demand, cheap power, easy money, and light rules, pass. Seek businesses that can adapt under tighter conditions.
Signals From Management Teams
Earnings calls reveal the truth over time. Listen for discipline. Are leaders setting clear milestones for spend and returns? Do they discuss cost per inference and customer payback in months, not years? Are they willing to slow builds when utilization slips? Do they admit where models underperform?
Also, watch how they treat shareholders. Clear capital allocation beats hype. If buybacks halt to fund capex, understand the trade. If stock-based pay soars while free cash flow falls, ask why. These are small clues that add up to a stronger or weaker case.
Lessons From Past Booms
Railroads built the skeleton of modern commerce. The dot-com bust still gave us the web we use every day. The early car makers launched a century of mobility. The tech lived on even when many investors lost money. That split is the point. Transformative ideas can win while early equity bets fail.
What separated winners was timing, cost control, and real customer value. Those markers matter again now. AI will settle into tasks where it is best: large data problems, code, search, drug design, and targeted insights. It may not replace every job, but it can change how many jobs are done. The pace will vary by sector and policy. It will not be smooth.
Where I Stand Right Now
I believe AI spending will stay high for the next few years. The strategic stakes are large for firms with capital and reach. That can keep GDP supported, and markets engaged. But I also expect a shakeout. Some projects will not clear their hurdle rates. Some competitors will merge or exit. Valuations will adjust when that becomes clear.
For portfolios, that calls for balance. Participate in growth, but protect against a hard stop. Track the markers listed earlier. Be ready to trim exposure if pricing weakens or paybacks slip. Likewise, be prepared to add exposure if adoption proves sticky and margins expand.
The Bottom Line
We are living through a major buildout. A few companies are spending at historic rates. Their bets now account for most of the growth we just recorded. History says big buildouts can go too far. It also says they can lay the tracks for years of progress. Both can be true at the same time.
Investors should keep a clear scorecard. Follow cash, pricing, adoption, and energy costs. Expect volatility. Do not confuse headlines with returns. If discipline holds, this cycle can deliver real value. If it slips, prepare for a reset before the next leg higher.
Frequently Asked Questions
Q: How can I tell if AI projects are creating real value?
Look for measurable results. Are customers renewing at higher rates, and do they expand usage? Do margins improve as usage scales? Are projects hitting payback targets within 12 to 24 months?
Q: What could trigger an AI-led market pullback?
Warning signs include falling prices for AI services, rising chip inventories, delays in data center builds, higher power costs, and management walking back AI-related revenue targets.
Q: Where might AI spending show the most staying power?
Areas with clear productivity wins tend to hold up. Examples include software development tools, search and support automation, data analytics, and select healthcare and industrial use cases with direct cost savings.
Image Credit: Shubham Dhage; Pexels
