AI Is Real. The Hype Is, Too
How Leaders Navigate Cycles Without Losing the Plot
By Bryan J. Kaus
I’ve spent the last few weeks on the road more than usual—client work, new partnerships, conversations with executives trying to figure out what to actually do with AI beyond the earnings-call theater. That vantage point keeps surfacing one question:
Is AI a bubble… and if so, what do we actually do about it?
My answer: AI is a very real technology wave wrapped inside a very familiar capital cycle.
If you lead a business, a team, or a portfolio, your job isn’t to choose between “all-in evangelist” and “this is all smoke.” Your job is to see the whole system: the technology, the capital, the physics, and the people.
AI is just the current vehicle. The principles apply everywhere.
1. The Temperature Check
You don’t have to squint to see we’re in a full-blown AI moment.
AI is now mentioned on over 40% of S&P 500 earnings calls - nearly triple the ten-year average. One chip designer has become the most valuable company on earth, almost entirely on AI data-center demand. Big Tech is committing hundreds of billions annually to AI infrastructure. Some forecasts push total capex into the trillions over the next five to ten years.
That’s not “emerging tech.” That’s mandate.
And with it comes bubble-like characteristics: concentrated bets, forward-pulled spending, herd storytelling, and a narrative so loud it drowns out dissent.
But that’s only half the picture.
2. Three Truths You Have to Hold at the Same Time
Truth 1: The technology is real
We’re well past the stage where AI is just a clever demo.
We have credible evidence that AI tools are making coders, consultants, call-center agents, and analysts faster and more accurate. They reduce drudge work, accelerate first drafts, and surface patterns humans would take days to find.
I see this personally. Used well, these tools can change the slope of the curve for mid-level knowledge work.
So the right stance isn’t “this is fake.” The right stance is: Where is it real, and what are the constraints?
Truth 2: Capital and narrative are running ahead of fundamentals
When you see AI on nearly half of earnings calls, valuations that assume decades of flawless execution, and multi-trillion-dollar spend plans anchored to use-cases we haven’t fully defined, you’re no longer just in “innovation.”
You’re in a speculative capital cycle.
That doesn’t mean it all collapses. It does mean you should expect:
Some projects and data centers to become stranded assets as tech and demand evolve
A long tail of AI “platforms” and “co-pilots” that never escape the slide deck
Pressure to show short-term AI “wins” that don’t necessarily line up with long-term value creation
Truth 3: Physics and people will bite before the PowerPoint does
The part of the AI conversation that’s still under-discussed is physical and human constraint.
On the physical side:
Data centers already consume a material share of total electricity, and that load is projected to roughly double by the end of the decade. AI-specific facilities are even more power-hungry, with forecasts of their consumption multiplying severalfold.
Much of our grid infrastructure in the U.S. and Europe is old, hard to upgrade, and politically contentious to expand.
On the human side:
Major companies are announcing five-figure job cuts, directly referencing AI and automation as part of their push to “operate leaner.” At the same time, they’re competing for a smaller pool of highly skilled people to actually design, run, and govern these systems.
At some point you have to ask: If we repeatedly rip out 10–20% of our workforce in the name of efficiency, are we genuinely optimizing the system, or are we admitting we mismanaged human capital for years and are now using technology as cover?
3. Substitution vs. Complement: Pulling Up the Ladder or Building a Better One?
A lot of AI implementation talk right now is basically: “Why hire a junior when a model can do their job?”
On a narrow spreadsheet, that logic works. If you view the firm as a living system, it’s dangerous.
We’ve seen healthier versions of this movie before.
Take automotive manufacturing: The Ford assembly line of a century ago relied on thousands of people doing repetitive, often dangerous tasks. Modern plants use automation and robotics to remove the worst physical risk, improve consistency, and reduce cycle times. Headcount per unit is way down. But humans haven’t disappeared; they’ve moved into higher-order roles - engineering, quality, maintenance, process design.
That’s complementary automation: machines remove drudgery and danger; people move up the value chain.
What many organizations are flirting with now is substitution automation:
Eliminate entry-level roles and mid-tier analytical work
Keep a thin strategic layer on top
Assume models will handle everything in between
That might make a couple of quarters look fantastic. It also pulls up the ladder that develops future leaders, institutional memory, and practical judgment.
I’m not arguing against aggressive efficiency. I’m arguing for systems thinking: Use AI to augment people and rebuild workflows, not just to remove headcount and call it strategy.
4. Is This a Bubble? That’s Not the Most Useful Question.
The more productive question is: Where are we in the cycle, and what does that imply for capital allocation and organizational design?
We’ve watched this pattern play out in other sectors.
Fitness hardware and subscription services that saw demand spike during the pandemic, extrapolated that into forever, ramped production and cost structure and then got crushed when demand reverted toward normal.
Shale oil and gas producers that over-drilled in boom years, ignored balance sheet discipline, and then spent years paying down debt after prices corrected.
Dot-com era companies with real ideas but no path to cash flow, swept up in a wave that eventually punished both the flaky and the fundamentally sound.
AI is similar:
The underlying tech is real
The use-cases are multiplying
The capital cycle around it is fully capable of overshooting
On top of that, funding structure matters. If AI and data-center buildouts are consuming most of your operating cash flow, you’re effectively betting your balance sheet on one thesis: that your AI investments will earn more than their cost of capital over time.
That’s where the “bubble” part tends to show up not in whether the technology works, but in whether the capital formation around it has overshot what the fundamentals can sustain over 5–10 years.
5. The Constraint That Will Define the Next Decade: Power and Steel
Because I sit at the intersection of energy and technology, the piece that jumps out at me isn’t the model architecture. It’s the power bill and the concrete.
A few realities:
Data centers are on track to become one of the fastest-growing loads on the grid - this on top of everything we’re already electrifying in transport and industry.
Utilities are planning record capital expenditures, but transmission projects and generation additions remain slow, litigious, and politically contested.
Bringing on new gas turbines, nuclear capacity, renewables, and storage all require huge amounts of capital, long permitting cycles, and policy stability that often doesn’t exist.
Some countries operate as “engineering states”: once they decide to build, they build: dams, high-voltage lines, new power plants. Others behave more like “lawyerly states”: every project is negotiated, challenged, and delayed.
AI doesn’t care which environment you operate in. It just needs electrons.
For AI, that means:
Energy and resilience infrastructure become quiet choke points
Long-term power purchase agreements from hyperscalers at above-market prices will keep many projects viable—but may also pressure retail consumers if regulators and utilities don’t manage the balance
Regions that can build reliable, affordable power and modern grid infrastructure will have a structural advantage over those that cannot
If you’re running an industrial, a data-center strategy, or any asset-heavy business, you cannot treat AI as just a “software story.” It is a steel, copper, concrete, and transmission line story.
6. What Leaders Should Actually Do
Let’s pull this out of the clouds and into a practical checklist.
If you’re an executive running a business:
Before you slap “AI-enabled” on your strategy deck or product sheet, ask:
What problem are we actually solving?
If you can’t explain the value in plain operational terms - cycle time, error rates, safety, customer experience - you’re playing to market buzzwords, not building market substance.
Are we complementing our people or substituting them?
Where can AI remove drudgery, rework, and low-value tasks?
Where are we at risk of hollowing out the talent pipeline that actually runs this place five or ten years from now?
What are the physical constraints?
Power: Do we truly understand our load, contracts, and exposure to price spikes?
Infrastructure: Are there bottlenecks in chips, talent, or grid capacity that could derail our plans?
How concentrated is our risk?
Are we effectively tying our fortunes to the success of two or three vendors, platforms, or regulatory assumptions?
What story are we telling our people?
If the only message is, “AI will make us more efficient, so we’re cutting 15%,” don’t be surprised when the survivors disengage and your best people quietly exit.
The organizations that win this wave won’t be the ones with the flashiest AI slide. They’ll be the ones that treat AI as an operating system upgrade for the whole business, not a one-time excuse to slash headcount or spot fix an issue.
If you’re an investor (formal or informal):
None of this is investment advice, but the lens matters:
Look past the narrative to boring fundamentals: free cash flow, unit economics, pricing power, and input constraints (power, water, chips, regulatory friction)
Assume the capital cycle will overshoot. It always does
Distinguish between AI as a feature (sprinkled onto every app), AI as infrastructure (chips, data centers, grid upgrades), and AI as capability (firms that can actually deploy it into their operations and customer offerings)
The durable returns are more likely in the third category firms that treat AI as a tool inside a broader system, not as a brand.
7. Why This Matters Beyond AI
The reason this conversation matters isn’t just AI.
Every sector has its hype cycles. Every CEO feels pressure to be the cheerleader even when the math says “be cautious.” Every leadership team has to manage the tension between short-term expectations and long-term system health - financial, physical, and human.
AI just compresses the timeline and raises the stakes.
Whether you’re running a pipeline, a refinery, a data-center campus, a mid-cap industrial, or a small business, the questions are the same:
Where are we in the cycle?
What are the real constraints?
Are we building resilience, or just hoping the music doesn’t stop on our watch?
The Point Taken
AI isn’t optional. Hype isn’t optional either. Discipline is.
Leadership, in this moment, is the discipline to be excited about what’s possible while staying relentlessly boring about the numbers, the physics, and the people who actually make it all work.
The organizations that compound value through this cycle won’t be the ones chasing every headline or hiding from every risk. They’ll be the ones who see the technology clearly, respect the constraints honestly, and build systems that work when the narrative changes.
That’s how you win the decade.