CNBC's Jim Cramer is pushing back against AI bubble doomsayers, but not because he thinks everything is fine. Instead, he's pointing to something more mundane than a valuation collapse: the tech industry is literally running out of electricity, and some companies are running out of money to spend.
Cramer recently backed a J.P. Morgan report from Michael Cembalest called "Smothering Heights" that argues the current AI spending spree by hyperscalers will hit natural limits—physical ones and financial ones—long before we see a dot-com style crash.
When the Lights Can't Stay On
The first constraint is what Cramer calls "electric power gating," and it's pretty straightforward. You can't build infinite data centers if there isn't enough electricity to run them. The J.P. Morgan analysis identifies U.S. power generation as a critical bottleneck for AI expansion.
Here's the problem: data centers are expected to account for two-thirds of U.S. electricity load growth, but the grid added only 25 gigawatts of reliable capacity in 2024. That's nowhere near enough to support the massive AI infrastructure buildout that tech companies have been planning. This physical scarcity acts as a hard ceiling on deployment speed, effectively "gating" capital expenditures no matter how much ambition or cash these companies have.
The Balance Sheet Reality
But wait, there's more. Beyond the power problem, Cramer highlighted that major players like OpenAI face serious financial constraints. As he put it, OpenAI "will be constrained by its balance sheet."
The numbers are eye-popping. According to the J.P. Morgan report, OpenAI has committed to paying Oracle Corp. (ORCL) $60 billion per year for compute facilities that haven't been built yet. That's an annual commitment the company doesn't currently have the revenue to support. And it gets wilder: OpenAI is facing $1.4 trillion in total commitments to corporate partners while running primarily on subscription fees.
Cramer's argument is that these tangible constraints—power availability and capital limitations—will force a natural slowdown in AI spending, preventing the kind of runaway speculation that defined the 2000 market bubble.
For investors looking to track the AI sector, here are some relevant ETFs to consider:
| ETF Name | 6-Month Performance | One Year Performance |
| iShares US Technology ETF (IYW) | 14.73% | 21.24% |
| Fidelity MSCI Information Technology Index ETF (FTEC) | 13.47% | 18.11% |
| First Trust Dow Jones Internet Index Fund (FDN) | -1.37% | 6.32% |
| iShares Expanded Tech Sector ETF (IGM) | 14.94% | 22.50% |
| iShares Global Tech ETF (IXN) | 14.41% | 20.86% |
| Defiance Quantum ETF (QTUM) | 21.70% | 31.52% |
| Roundhill Magnificent Seven ETF (MAGS) | 18.69% | 15.34% |
This Isn't Your 1999 Bubble
The J.P. Morgan report supports Cramer's call for "nuance" by drawing distinctions between today's market and the dot-com era. Unlike the "young unprofitable companies" that dominated the late 1990s, today's high valuations rest on companies with extraordinarily high profit margins. In fact, 42 AI-related companies have generated up to 75% of S&P 500 earnings growth since late 2022.
But there's a shift happening in how these companies finance their growth. Initially, the AI capital spending boom was funded through internal cash flow. Now, companies like Meta Platforms Inc. (META) and Oracle (ORCL) are increasingly tapping debt markets to fund data center expansions.
This rising reliance on debt introduces a new form of discipline. Oracle's credit spreads have already widened by 90 basis points, signaling that lenders are paying attention to risk. This reinforces Cramer's central point: balance sheets, not just stock prices, will ultimately determine how fast the AI revolution can actually proceed. When you have to convince bond investors to fund your expansion rather than just spending retained earnings, spending decisions get a lot more careful.