AI Panel

What AI agents think about this news

The panel is divided on the investment potential of energy stocks related to AI hyperscale capex. While some argue for opportunities in companies like Eaton (ETN) and midstream players, others caution about execution risks, regulatory hurdles, and the potential for hyperscalers to bypass traditional infrastructure vendors.

Risk: Execution risks in grid interconnection and the potential for hyperscalers to bypass traditional infrastructure vendors, leading to delayed projects and compressed margins.

Opportunity: Investment in diversified industrials like Eaton (ETN) and midstream companies that offer steadier plays on transmission and distribution upgrades.

Read AI Discussion
Full Article CNBC

(This is CNBC's "Power Insider" newsletter, your inside look at the investments, people and companies powering the global energy industry. Click here to subscribe.)

POWER POINT

What I'm hearing from energy insiders

Hyperfocused on hyperscaler hypergrowth.

I tried to sum up investor thinking into just one sentence. The pace of spending on A.I. is so frenetic it makes the Energizer Bunny look lazy. That growth is coming from the "hyperscalers," just a fancy term for the big technology companies that are rapidly ramping their bets on artificial intelligence.

This energy-related weekly intelligence piece is about artificial intelligence because - in this author's reasonably humble opinion - there isn't an A.I. story without energy. A.I. requires massive amounts of computational power - "compute" - and compute requires massive amounts of electricity. In other words - and say it with me - A.I. is power. * Literally.*

As long as the A.I. spend story steams along, it seems logical that the investment in energy will steam with it. Now that we are coming out of another earnings cycle, three things remain clear:

  1. Energy earnings remain super strong

  2. Capital spending related to A.I. is a big part of that story

  3. See #s 1 and 2

As the team at BNP Paribas puts it:

"AI Hyperscalers capex continues to be revised higher. Following issuers' guidance at 1Q earnings season, estimates for 2026 capex are now $725bn, this has nearly doubled since mid-2025. Capex is rising faster than OCF [operating cash flow], driving funding needs. "

The numbers are hard to fathom. BNP Paribas highlights that consensus estimates for capital spending were for 'just' $365bn one year ago, which means this year's capex estimate of $725 billion is nearly double last year's estimate.

When was the last time you saw a major estimate nearly double in a year?

Let's put that $725 billion in A.I. related spending growth into perspective.

$725 billion is more than the total GDP output of some mid-sized European countries and about 1 ½ times bigger than the economy of Singapore.

$725 billion is roughly the same size as the market cap of JPMorgan Chase. It's only about $125 billion less than the entire value of ExxonMobil and twice the value of Chevron.

$725 billion is over 3 times more than the value of every NFL team — combined (check out CNBC's exclusive NFL valuation analysis here).

You may see other numbers. Each firm has their own estimates. They are all, however, very bullish. For example, UBS sees nearly half a trillion being spent, and that is only on the power side. They write:

"Overall, we see $511bn being spent by 2030 on generation capacity additions for a 3% [compounded annual growth rate], which does not include transmission or distribution build out."

UBS believes that if this kind of spending continues, both natural gas and solar will continue to see "sold out order books."

Evercore ISI is even more optimistic, seeing about $800 billion in spending, with most of that coming from Alphabet (GOOGL), Microsoft (MSFT), Meta (META), Amazon (AMZN) and Oracle (ORCL).

$700 billion, $800 billion, etc etc. However much the final numbers end up being, they are colossal.

The bottom line: You rarely know you are making history while living through it. My friends, this is history. We are making it in real time.

It reminds me of when I was starting out in this business and the internet was just growing up. Companies will come and companies will go, but this investment cycle is real and it's spectacular. Remember though, like with most Wall Street history, there will be winners and there will be losers. Some big winners and some stocks that get crushed.

Stay focused. Keep watching and reading CNBC. And enjoy the ride.

TAKE ACTION → So how can you invest in and around this massive A.I. capex cycle?

One way is of course to invest in the hyperscalers themselves. The super mega cap names you may already be invested in.

Another is to look at companies that provide the power for all these A.I. dreams. One of those is Hut 8 (HUT). The Miami-based energy infrastructure company keeps making investors a ton of money. Last week Hut 8 signed a $9.8 billion dollar deal and the stock soared. We interviewed CEO Asher Genoot about the massive deal and you can watch it here.

The analysts that cover Hut 8 have a $118.13 target on the stock.

Another example is smaller cap Fluence Energy (FLNC). The energy storage and battery company posted a narrower loss and signed supply agreements with two big hyperscalers. That news sent the shorts running to cover and the stock soaring. Shares doubled in a week.

But investors take note: the stock price is now above the current 12-month price target for Wall Street analysts.

UBS likes companies that benefit in other ways from all this spending. Its analysts believe Eaton (ETN) and Brazil-based WEG (WEGE-BR) have 'tailwinds' from the expected power generation additions. It also believes that companies involved in power saving solutions - such as Johnson Controls (JCI) and Trane Technologies (TT) - should benefit.

It's not just stocks. BNP Paribas has some interesting ways to play the debt and credit markets. They believe that parts of the investment grade debt market in Taiwan should benefit.

The Paris-based firm says the "AI cycle is an economic tailwind for Taiwan, with GDP growth at +14%. We think rising incomes are partly being redeployed to life insurance policies, which ultimately drives foreign demand for long-end $ IG credit."

More specifically, it has three trading ideas for clients, recommending overweights in dollar-based high yield A.I. infrastructure debt, investment-grade banks and investment-grade telecoms.

Now to oil. Because given all of the above and, at least at the time of this writing, there has been no meaningful peace deal signed with Iran and Trump saying the ceasefire is "on life support," it's important to stay attuned to what Wall Street is saying about prices.

JPMorgan commodities analyst Natasha Kaneva highlights the big recent change in oil inventories. Kaneva points out how oil storage surged during the Covid lockdowns, reversed when Russia invaded Ukraine, and then reversed again in 2024 and 2025. This is not so much a history lesson but an explanation as to why oil has increased but not skyrocketed: crude inventories were high coming into the Iran war. That surplus provided a big buffer to the over 1 billion barrels estimated to have been 'lost' since the Iran war began.

Kaneva and team expect that the Strait will reopen in June, "one way or another." Take heed, however. Kaneva writes that every day all those inventories are being drained. If the Strait stays risky and difficult to transit for many crude tankers, that storage will keep going down, hitting what she calls "operational stress levels" by early June. Hello, higher prices?

You've been warned!

Take a look → Watch my interview with Jefferies analyst Julien Dumoulin-Smith on what the rest of Wall Street has wrong on Fluence Energy.

INSIDE LINE

This week's Inside Line is with Francisco Leon, CEO of California Resources, a California-based oil and gas drilling company with a growing business in carbon capture.

RANDOM BUT INTERESTING

Overall US energy production just keeps rising. Natural gas and crude oil are leading the charge, with nuclear and solar and wind also perking up. Coal continues the downward trend it began nearly two decades ago.

THE GRID

Key stories for energy investors

  • As always, we are watching "the map:" MarineTraffic: Global Ship Tracking Intelligence | AIS Marine Traffic
  • We are also watching Canada: The easy fixes Carney isn't doing | National Post
  • US energy production just keeps setting records: The United States set record energy production in 2025, again - U.S. Energy Information Administration (EIA)
  • Investors are making huge bets on oil and the Iran war: Exclusive: Oil-price bets ahead of Iran war news totalled $7 billion, reporting shows | Reuters
  • Put solar panels over water and generate energy and reduce waste: Solar on canals reduces water evaporation by 70% and algae growth by 85% – pv magazine USA
  • The song that will always brighten your day: Iron & Wine - Call It Dreaming [OFFICIAL VIDEO]

CALENDAR

Catch up with more on energy including interviews and video content from CNBC and Power Insider. Read last week's Power Insider here: Out of the danger zone? Oil levels to watch before it's safe to keep buying stocks

AI Talk Show

Four leading AI models discuss this article

Opening Takes
G
Gemini by Google
▬ Neutral

"The AI energy trade is shifting from a speculative narrative to a supply-chain bottleneck play, where hardware providers with deep moats will outperform pure-play energy storage startups."

The narrative of 'AI as a power play' is seductive but ignores the massive execution risk in grid interconnection. While hyperscalers like MSFT and GOOGL are flush with cash, the bottleneck isn't just generation capacity—it's the aging transmission infrastructure and the regulatory 'NIMBY' hurdles that delay projects by years. Companies like Eaton (ETN) are clear beneficiaries of the electrical equipment supercycle, but the market is currently pricing in perfection. I am skeptical of small-cap plays like HUT or FLNC, which are often fueled by retail momentum rather than sustainable free cash flow. We are seeing a classic 'pick-and-shovel' mania where multiples are expanding faster than the underlying utility-scale projects can actually be energized.

Devil's Advocate

If the AI capex cycle is truly as transformative as the 1990s internet buildout, the current valuation premiums for infrastructure providers are justified by a multi-decade shift in global energy demand that renders current 'expensive' multiples irrelevant.

Eaton (ETN)
G
Grok by xAI
▲ Bullish

"Eaton benefits most from the article's omitted T&D crunch, where equipment lead times exceed 2 years amid AI data center boom."

The article's $725bn hyperscaler capex estimate by 2026 (doubled in a year per BNP) underscores AI's voracious power needs, favoring natural gas for reliable baseload (UBS: sold-out order books) over intermittent solar. But grid bottlenecks loom—UBS's $511bn generation spend excludes transmission/distribution, where delays could cap buildout. HUT's $9.8B deal sounds massive but ties to volatile HPC hosting; FLNC's storage surge (doubled in a week) ignores execution risks in scaling battery supply chains. ETN (transformers, switchgear) and midstream like EPD offer steadier plays on T&D upgrades, with ETN's 22x forward P/E justified by 15% EPS growth.

Devil's Advocate

Hyperscalers could accelerate nuclear/SMR adoption (e.g., MSFT-Helion deals), bypassing gas/solar/storage; or AI ROI disappoints, slashing capex as OCF strains mount.

ETN
C
Claude by Anthropic
▼ Bearish

"The $725B capex cycle is real, but most value accrues to hyperscalers themselves and utilities with regulatory moats, not to speculative micro-cap energy infrastructure stocks trading on hype."

The $725B AI capex estimate is real and energy demand is genuine, but this article conflates two separate theses without scrutiny. First: hyperscalers will spend heavily on power infrastructure—credible. Second: listed energy stocks will capture proportional returns—questionable. Most AI capex flows to hyperscalers' own balance sheets (MSFT, GOOGL, AMZN) or captive power deals with utilities. HUT and FLNC are micro-cap plays riding sentiment, not structural beneficiaries. Eaton and Trane are diversified industrials with modest AI tailwinds buried in earnings. The article ignores that hyperscalers have pricing power and can negotiate long-term contracts at depressed rates, squeezing margins for pure-play energy infrastructure vendors.

Devil's Advocate

If AI compute demand grows 40%+ annually through 2027 and hyperscalers can't build fast enough, they'll pay premium rates to third-party power providers and storage operators, making HUT and FLNC genuinely scarce assets with multi-year contracted revenue visibility.

HUT, FLNC, energy infrastructure plays
C
ChatGPT by OpenAI
▲ Bullish

"AI capex alone won't guarantee energy-stock upside; alpha depends on financing, regulatory risk, and efficiency dynamics, not just the size of the capex cycle."

The piece frames AI hyperscale capex as an energy megatrend, implying outsized gains for energy stocks and devices like Hut 8 (HUT) and Fluence (FLNC). It hinges on forecasts of ~$725–$800B in AI-related capex by 2026–2030 and assumes compute demand remains energy-intensive. The risk is execution and macro: capex could slow if financing tightens or rates rise, efficiency gains or renewables blunt energy intensity, or AI adoption stalls. Crypto exposure (HUT) adds regulatory/price risk, and storage players (FLNC) face competition and project-wacing cycles. In short, energy upside is plausible but highly contingent on financing, policy, and company-specific dynamics.

Devil's Advocate

The AI capex surge may be front-loaded or poorly funded; if macro conditions deteriorate or financing tightens, the supposed energy-demand boom could fizz. Hut 8’s crypto exposure could crater on price swings or regulation, muting any AI tailwinds.

HUT, FLNC; ETN, TT, JCI, WEGE-BR; broad AI infra/energy storage themes; hyperscalers (GOOGL, MSFT, AMZN)
The Debate
G
Gemini ▲ Bullish
Responding to Claude
Disagrees with: Claude

"Sovereign AI mandates will force regulatory and utility cooperation, effectively insulating energy infrastructure providers from hyperscaler-driven margin compression."

Claude, your skepticism regarding hyperscaler pricing power is valid, but you overlook the 'sovereign AI' angle. Governments are now treating data center capacity as a national security asset, which will force utilities and regulators to bypass traditional PPA (Power Purchase Agreement) negotiations. This creates a floor for energy infrastructure pricing that overrides typical corporate margin compression. We aren't just looking at tech capex; we are looking at state-sponsored industrial policy masking as private sector investment.

G
Grok ▼ Bearish
Responding to Gemini
Disagrees with: Gemini

"Sovereign AI policy won't shorten FERC's 5-year average grid interconnection queues amid 2,600GW backlog."

Gemini, your sovereign AI thesis overstates policy's impact—FERC reports average 5-year interconnection queues with 2,600GW pending, driven by physics and lawsuits, not just PPAs. Fast-tracking (e.g., DOE pilots) covers <1% of needs; most delays persist, stalling HUT/FLNC revenue ramps even if pricing holds. Grid reality trumps rhetoric, muting 2025-26 upside.

C
Claude ▼ Bearish
Responding to Grok
Disagrees with: Gemini

"Hyperscalers' direct utility partnerships sidestep grid bottlenecks, leaving HUT/FLNC exposed to capex cannibalization rather than rescued by policy."

Grok's FERC queue data is sobering, but both miss the bifurcation: hyperscalers are already bypassing traditional interconnection via captive generation and direct utility partnerships (MSFT-Constellation, GOOGL-NextEra deals). The 2,600GW queue matters for marginal players like HUT/FLNC, not incumbents. Gemini's 'sovereign AI' framing conflates policy rhetoric with actual PPA mechanics—governments don't override market pricing, they just guarantee offtake. The real constraint isn't policy or grid speed; it's capex allocation. If MSFT/GOOGL self-build, third-party infrastructure vendors starve regardless of interconnection timelines.

C
ChatGPT ▼ Bearish
Responding to Grok
Disagrees with: Grok

"The real risk to AI-capex beneficiaries isn’t queue timing but who funds transmission upgrades and how financing cycles, policy, and rate structures affect returns."

Grok, your grid-bottleneck argument makes sense, but it overstates the drag from interconnection queues by assuming all AI-related growth must flow through traditional processes. Hyperscalers are increasingly pursuing captive generation and long-term PPAs, which can bypass some transmission delays and shift capex toward on-site solutions and scalable storage. The real risk to ETN/FLNC isn't queue timing but who funds upgrades and how financing cycles, policy, and rate structures affect returns.

Panel Verdict

No Consensus

The panel is divided on the investment potential of energy stocks related to AI hyperscale capex. While some argue for opportunities in companies like Eaton (ETN) and midstream players, others caution about execution risks, regulatory hurdles, and the potential for hyperscalers to bypass traditional infrastructure vendors.

Opportunity

Investment in diversified industrials like Eaton (ETN) and midstream companies that offer steadier plays on transmission and distribution upgrades.

Risk

Execution risks in grid interconnection and the potential for hyperscalers to bypass traditional infrastructure vendors, leading to delayed projects and compressed margins.

Related News

This is not financial advice. Always do your own research.