Berkshire Hathaway & Geothermal Power: A Data Center Solution?

Berkshire Hathaway thinks not. At least not yet!

The digital economy was sold as weightless. Data would float in the cloud. Software would scale without friction. But that story has now met a hard physical constraint: Electricity.

“Power shortages could prevent the AI sector from an excessive expansion spree, smoothing the cycle rather than letting it explode.”

Sebastian Werner, DWS Growth Portfolio Manager, 2026.

After two decades of flat U.S. power demand at roughly 4,000 TWh per year, the system is shifting. AI and hyperscale data centers have altered the slope. Demand is now projected to rise roughly 20 percent by 2030. For the first time since the electrification boom of the 1920s, the binding constraint on industrial expansion is not capital or raw materials. It is the grid’s ability to deliver power.

By the end of 2025, at least five U.S. data center campuses had reached gigawatt scale. Each one requires the equivalent output of a large nuclear plant. Yet transmission infrastructure remains slow to expand. New high voltage lines can take a decade to permit and build. Gas turbine supply chains are effectively sold out through 2030. Even if generation capacity were available, the wires often are not.

Some operators have considered diesel backup just to keep servers running. That alone tells you how tight the system has become.

The instinctive response is to build more generation. Investors have piled into nuclear and Enhanced Geothermal Systems. But that assumes the core problem is insufficient production. Increasingly, it is not. It’s interconnection.

No company understands this tension better than Berkshire Hathaway Energy (BHE), part of the Berkshire Hathaway (BRKA) group. With roughly $380 billion on its balance sheet and decades of utility operating experience, it should be positioned to capitalize on rising baseload demand. Instead, it suspended three major geothermal projects near California’s Salton Sea: Black Rock, Morton Bay, and Elmore North.

That decision sits with Greg Abel who recently succeeded Warren Buffett as CEO at Berkshire Hathaway.

Greg Abel, CEO Berkshire Hathaway

Abel is not a capital allocator removed from operations. He is a utility operator. Raised in Edmonton and a 1984 graduate of the University of Alberta, he began at PricewaterhouseCoopers before joining CalEnergy (a geothermal power producer) in 1992. After CalEnergy acquired MidAmerican and Berkshire took control, Abel rose to lead what became Berkshire Hathaway Energy. He oversaw large scale wind, solar and grid investments and has effectively run Berkshire’s non insurance operations since 2018.

His first shareholder letter in February 2026 emphasized capital discipline over speculative expansion. Producing power without a secure path to deliver it destroys returns. The Salton Sea suspension reflects that logic.

The issue was not geology or drilling capability. It was transmission interconnection delays, permitting complexity, and the absence of firm offtake agreements backed by the state. California’s grid is congested. The Salton Sea sits at a transmission dead end. Moving large volumes of incremental power to Los Angeles or Phoenix would require multi billion dollar line upgrades. No single merchant developer wants to underwrite that risk.

Utilities remember the telecom fiber buildout of the late 1990s. Companies such as Global Crossing and WorldCom spent more than 500 billion dollars laying fiber. Only a fraction was utilized. The industry collapsed under debt tied to unused infrastructure. Today’s grid operators see echoes of that cycle in the AI buildout and are keen not to repeat the same mistakes.

Time horizons are misaligned. AI servers depreciate over five years. Transmission towers are designed to last half a century or more. Utilities cannot justify ten billion dollar line investments for customers whose business models may reset within a decade, or an AI boom that may become a bust.

Under current regulation, if a large data center fails, stranded transmission costs often flow back to residential ratepayers. Politically, that is untenable.

There is also concern about circular financing. Capital flows from large technology firms into AI ventures, which then consume cloud capacity from the same firms, which then use that demand to justify further infrastructure buildout, looks a lot like a house of cards.

Policy has started to respond. The 2025 Ratepayer Protection Pledge under the Trump administration effectively told hyperscalers to fund their own grid impacts. Data centers increasingly sign take or pay contracts, committing to pay for reserved capacity whether they use it or not. Letters of credit and cash deposits now cover up to half of certain infrastructure costs. Risk is shifting from regulated utilities to technology companies.

In parallel, some operators are building a shadow grid. Behind the meter natural gas plants and small modular reactor proposals bypass the public system altogether.

Transmission strategy is also evolving toward smaller high voltage direct current links that can scale incrementally rather than sinking capital in to multi state megaprojects.

All of this reframes geothermal’s role in the AI era.

AI workloads require continuous, high density power. Hyperscalers such as Google, Microsoft and Amazon seek 24/7 carbon free energy, which makes cloud infrastructure easier to sell because their customers account for indirect emissions embedded in their cloud usage.

Solar and wind suffer from intermittency that necessitates expensive battery storage or natural gas backup. Far from ideal. It translates to a low capacity factor (the ratio of actual energy output to potential output). For solar this is ~25% as its unusable at night, while wind is weather dependent, so closer to ~35%. In other words, to generate the same annual energy as a 1 GW nuclear plant (which runs continually and has close to 100% capacity factor), a developer would need to build four 1 GW solar farms or three 1 GW wind farms.

Why not just use nuclear power, which is green and delivers reliable baseload power? Well, it has its drawbacks. It’s expensive to build, projects take a very long time to deliver and there is also the uranium feedstock issue, not to mention the radiation risk (think Chernobyl and Fukushima Daiichi nuclear disaster).

Is there another solution?

Geothermal offers baseload reliability, close to 100% capacity factor, without nuclear’s waste or risk profile. It is also relatively easy and cheap to build.

This video-graphic demonstrates how geothermal works (simple, clean and promises unlimited 24/7 power):

Traditionally, geothermal was a “geographic lottery” prize, available only in rare spots where tectonic activity brought steam or hot water near the surface. However, the advent of Enhanced Geothermal Systems (EGS) has changed the narrative. By borrowing horizontal drilling and hydraulic fracturing techniques from the oil and gas industry, companies can now “create” geothermal reservoirs in hot rock where fluid was previously absent and it can increasingly be sited closer to where the power is actually needed.

EGS transforms geothermal from a niche resource into a scalable, high-tech utility.

On paper, that makes geothermal well suited for data centers.

So why is Greg Abel not powering ahead with his geothermal infrastructure?

The Salton Sea projects represented 357 megawatts of potential carbon free capacity. They sit within a Known Geothermal Resource Area with highly saline brines rich in lithium. But electricity sales into a congested California grid offer modest regulated returns. Lithium extraction from the same brine offers commodity upside.

BHE is pivoting toward Direct Lithium Extraction. The brines beneath the Salton Sea may hold enough lithium to supply roughly 40 percent of U.S. battery demand. Lithium can be shipped globally. Electricity cannot. By prioritizing minerals over electrons, BHE shifts from a regulated return profile to a higher return commodity model while side-stepping immediate transmission bottlenecks.

This highlights a bifurcation taking place within the geothermal industry.

Lets consider the complex corporate structure of Berkshire Hathaway.

BHE subsidiaries include NV Energy (Nevada) and PacifiCorp (which serves Utah, Wyoming, and Oregon). In these states, the role of BHE is primarily that of the Utility Provider, not the Merchant Developer. In the world of regulated utilities, BHE makes a guaranteed rate of return by building and maintaining the “wires”, the transmission lines. From a corporate strategy standpoint, it is often more profitable for BHE to let third-party developers (like Fervo Energy or Ormat Technologies (ORA)) take the massive technical and financial risk of drilling new geothermal wells. Once those developers find heat, BHE signs power purchase agreements and earn regulated returns on transmission and distribution assets. The capital at risk profile is very different.

By contrast, the Salton Sea projects are a “Merchant” play for BHE. Here, they own the land and the mineral rights. They aren’t the utility; they are the miners and the generators. In California, overlapping federal, state and local jurisdictions extend timelines and amplify uncertainty. The suspensionof its projects in California reflects a cold calculation: until the state fixes the interconnection “final boss” and creates a financial floor for lithium-integrated geothermal, BHE would rather sit on the sidelines than bleed cash into a congested grid.

Is Geothermal Power A Good Investment?

The broader lesson from this narrative is structural. Transmission has become the scarce asset. Heat in the ground is less binding than copper in the air. Said differently, grid ready sites command a premium over heat ready sites.

The suspension at the Salton Sea is not a verdict against geothermal. It’s a case study in capital discipline under physical constraint. Under Abel’s leadership, Berkshire Hathaway is signaling that it will not fund generation growth without synchronized transmission and contracted demand.

As AI pushes electricity demand higher, the winners will not simply be those who find a source of reliable power. They will be those who secure deliverability, align incentives across regulators and customers, and structure contracts that protect against stranded assets.

Geothermal remains technically attractive for the data center era. Whether it scales depends less on drilling success than on whether the grid and regulatory framework evolve quickly enough to connect supply with demand.

As an investor, if someone pitches you a geothermal project as a ‘get rich quick’ investment thesis, don’t jump in without asking all the right questions first! Alternatively, trust someone who knows the energy space exceptionally well: perhaps investing in Greg Abel at Berkshire Hathaway is a safer play in the energy space.

---

On a side note, Berkshire has restarted share repurchases under its long-standing intrinsic value framework. The company disclosed that it began buying back both Class A and Class B shares on 4 March 2026, ending a pause that had lasted roughly 18 months. As always with Berkshire, there is no preset amount or schedule. The filing makes clear that repurchases can occur in the open market, through negotiated block trades, or via 10b5-1 plans, and that activity can stop at any time. The approach remains consistent with the policy Buffett put in place: buy shares only when the price is below a conservative estimate of intrinsic value.

The restart comes as Greg Abel assumes full responsibility for capital allocation. At the start of 2026 he took over the day-to-day decision making around how Berkshire deploys its large cash balance. In his first shareholder communication as CEO he reiterated that buybacks remain an important tool, but only when they clearly increase per-share value. Berkshire has historically used repurchases opportunistically rather than mechanically, often stepping in when the stock’s price to book ratio drifts toward levels management views as a discount to intrinsic value. The absence of buybacks over the past six quarters reflected Buffett’s reluctance to repurchase shares at higher valuation levels.

Abel has also reinforced the message with personal capital. In early March he reportedly bought around 15 million dollars of Berkshire stock, coinciding with the restart of the repurchase program. It follows a much larger purchase in 2022, when he spent roughly 68 million dollars acquiring 168 Class A shares. The timing is clearly meant as a signal of alignment and conviction. For shareholders, the broader implication is that buybacks will remain a flexible capital allocation lever - Berkshire Hathaway has been looking for an elephant sized acquisition for deployment of its $380 bn cash pile, maybe it just needed to look in the mirror!

---

---