There are two bad ways to look at AI stocks right now.

The lazy take says the entire opportunity is over because NVIDIA and a handful of obvious winners already exploded.

The reckless take says the buildout is so big that any company with “AI” in its investor deck deserves your money.

Both are wrong.

The first phase rewarded people for noticing AI. The next phase will reward people who can separate the companies controlling real bottlenecks from the companies selling an expensive story.

The trade changed. It did not disappear.

THE FIRST WAVE RAN. THE SPENDING DID NOT STOP.

Start with the latest physical evidence, not a stock chart.

In April 2026, the International Energy Agency reported that the capital spending of five large technology companies had surged to more than $400 billion in 2025 and was set to increase by a further 75% in 2026. It also found that electricity demand from data centers rose 17% in 2025, while demand from AI-focused facilities grew even faster. (IEA, Key Questions on Energy and AI; IEA April 2026 release)

That is not a promise that AI stocks go up. It is proof that the companies building AI infrastructure are still writing enormous checks.

NVIDIA reported $81.6 billion in quarterly revenue for the period ended April 26, 2026, up 85% from a year earlier. Data Center revenue reached $75.2 billion, up 92%. Networking revenue climbed 199% to $14.8 billion. (NVIDIA Q1 FY2027 results)

Microsoft spent $31.9 billion on capital expenditures in the quarter ended March 31, 2026. Roughly two-thirds went to shorter-lived assets, mainly GPUs and CPUs. The rest went to long-lived infrastructure that Microsoft expects to support monetization for 15 years and beyond. (Microsoft FY2026 Q3 earnings call)

Alphabet said its 2026 capital spending was expected to reach $175 billion to $185 billion, while Google Cloud backlog ended 2025 at $240 billion. (Alphabet Q4 2025 earnings call)

One company can overspend. One quarter can mislead. But hundreds of billions of dollars, spread across competing platforms and committed to chips, servers, networking, land, power and cooling, tell you something important:

The machine is still being built.

THE BOTTLENECK MOVED

In the first wave, the market focused on who made the fastest AI accelerator. Now the constraint is wider.

A world-class chip is useless if the wafer cannot be fabricated, the memory cannot feed it, the network cannot connect it, the rack cannot cool it or the grid cannot power it.

The IEA estimates that data centers consumed roughly 415 terawatt-hours of electricity in 2024. Its base case reaches approximately 945 TWh by 2030—about 128% higher. The agency also estimates that grid constraints could delay roughly 20% of global data-center capacity planned through 2030. (IEA, Energy demand from AI; IEA grid-delay analysis)

That delay risk matters. The next winners may not be the companies with the loudest AI branding. They may be the companies that remove the constraint everyone else is forced to pay around.

FOLLOW THE MONEY THROUGH FIVE PHYSICAL GATES

  1. COMPUTE — accelerators, systems and the software that keeps them productive.
  2. MANUFACTURING — advanced wafers, lithography and packaging.
  3. MEMORY + INTERCONNECT — high-bandwidth memory, networking, optics and rack-scale fabric.
  4. POWER + COOLING — generation, grid equipment, power delivery and thermal control.
  5. RACKS + CAMPUSES — systems integration, powered land and finished data-center capacity.

The opportunity is no longer “AI” as one giant trade. It is a fight over who controls each gate.

HERE IS THE PART THE HYPE ACCOUNTS LEAVE OUT

A growing industry does not automatically create a good stock entry.

A brilliant company can still be overpriced. Revenue can rise while the multiple collapses. A supplier can post record orders and still get crushed when one customer cuts spending. A backlog can look powerful until financing, permitting or delivery turns it into dead weight.

AI efficiency is improving fast. The same IEA update that documents surging demand also warns that energy use per AI task is falling rapidly. Better chips, better models and better utilization could change where the money lands.

That is why “AI is growing” is not an investment thesis. It is only the beginning of the research.

  • Who owns something customers cannot easily replace?
  • Where is demand already visible in revenue, margins, orders or cash?
  • How durable is the advantage if competition catches up?
  • How much success is already trapped inside the price?
  • What exact fact proves the thesis is broken?

If you cannot answer all five, you do not have a process. You have a ticker and a feeling.

THE FIVE TESTS: NO FANBOY PICKS. NO FREE PASSES.

1. CONTROL

Does the company own a bottleneck customers cannot easily bypass? Not popularity. Not a famous CEO. Control.

2. PROOF

Is demand already visible in operating results? Revenue. Margins. Orders. Cash. Customer commitments. Stories do not count.

3. DURABILITY

Can the advantage survive competition, a weak cycle and a major customer changing direction?

4. PRICE

How much future success are you being forced to pay for today? The best business on the board can still be the worst entry.

5. BREAK

What hard fact makes you stop adding, walk away and rebuild the case from zero? Most free stock content gives you a bullish case. Very little gives you the number that kills it. That number is where discipline begins.

“AM I TOO LATE?” IS THE WRONG QUESTION

You can be early to a bad company. You can be late to a great company and still overpay. You can also arrive after the first headline move and find a second-order supplier whose proof is accelerating faster than its reputation.

Where is the next dollar being forced to go—and what has the stock price already assumed?

That question turns FOMO into a research problem. Research problems can be solved.

THE OPERATING MAP BEHIND THE THESIS

The AI Stocks Power Play ranks seven Core leaders and seven higher-risk Challengers with current proof, valuation, next checks, hard stops and a complete execution system.

See the AI Stocks Power Play

WHAT THE POWER PLAY ACTUALLY GIVES YOU

The AI Stocks Power Play is not fourteen personal buy commands. It is the operating map we wanted before putting real capital near this theme:

  • Seven Core companies with stronger operating proof and more established control.
  • Seven Challengers with higher upside pressure, lower certainty and stricter break lines.
  • Current price, market cap, sales multiple and operating proof on every company card.
  • The next dated metric that has to confirm the case.
  • A specific walk-away condition for every name.
  • A side-by-side board showing where proof, price and risk separate.
  • A Watchboard explaining why Bloom Energy, Arista Networks and AMD did not make the final fourteen—and what would earn a promotion.
  • Position Control, Move Control, an Earnings Protocol, Thesis Control and a one-page Command Sheet.
  • A complete source register so important claims can be checked instead of merely trusted.

THREE QUESTIONS BEFORE YOU LEAVE

DID I ALREADY MISS NVIDIA?

Maybe you missed the easiest price. That does not tell you whether NVIDIA is attractive today, and it says nothing about the companies supplying the rest of the stack. Price and proof have to be judged together.

IS AI A BUBBLE?

Parts of it can be. Real infrastructure demand and dangerous valuation can exist at the same time. The label does not make the decision for you.

WHY NOT JUST BUY AN AI ETF?

An ETF can diversify company-specific risk, but it also owns whatever the index includes at whatever weight it assigns. The Power Play is for readers who want to understand the individual control points, evidence, valuation pressure and failure conditions before deciding how—or whether—to participate.

THE BOTTOM LINE

The easy phase was noticing AI.

The harder phase is following the money after the headlines, finding the bottleneck and refusing to pay any price just to feel involved.

The spending is still accelerating. The power wall is real. The execution risk is real. The valuation risk is real.

That is not a reason to rush. It is a reason to get sharper.

You do not need another hot ticker. You need a map, a price and a line that makes you walk.

THE CORE SEVEN. THE CHALLENGER SEVEN.

Forty-eight pages. Fourteen company cases. Thirty-eight linked sources. One system for knowing the move before the market moves.

Open the AI Stocks Power Play

SOURCES

  1. International Energy Agency, Key Questions on Energy and AI, April 16, 2026.
  2. International Energy Agency, Data centre electricity use surged in 2025, April 16, 2026.
  3. International Energy Agency, Energy demand from AI, April 10, 2025.
  4. International Energy Agency, AI and energy security, April 10, 2025.
  5. NVIDIA, Q1 FY2027 financial results, May 20, 2026.
  6. Microsoft, FY2026 Q3 earnings call, April 29, 2026.
  7. Alphabet, Q4 2025 earnings call, February 4, 2026.

King Research Group publishes general educational research, not personalized investment advice. This article does not consider your objectives, risk tolerance or financial situation. Securities can lose substantial or all value. Past performance does not predict future results. Industry projections are uncertain and may differ materially from actual outcomes. Review current filings, market data and relevant conflicts disclosures before making any decision.