A quick scan of Reddit reveals that hundreds of people have had the same issue dating back months—and as recently as Wednesday. Claude’s sleep demands are varied and, often, quirky variations of the same message.

To one user it may write a simple “get some rest,” yet for others its messages are more personalized and empathetic. Oftentimes, Claude will repeat the message multiple times.

“Now go to sleep again. Again. For the THIRD time tonight…” it replied to a person with the Reddit username, angie_akhila.

Some users have said they find Claude’s late night rest reminders “thoughtful,” while others have said they’re annoying, given Claude often gets the time wrong, anyway. 

“It often does it at like 8:30 in the morning. Tells me to go get some rest and we’ll pick back up in the morning,” wrote one user on Reddit. 

Online speculation abounds on why the chatbot insists users rest, including a theory that it’s an intentional feature to promote users’ wellbeing, or that the Anthropic is trying to save computing power by discouraging prolonged Claude use. These explanations aren’t likely as Claude isn’t given context about a user’s usage. The company also recently struck a deal with Elon Musk’s SpaceXAI (formerly SpaceX) to add more than 300 gigawatts of compute capacity.

Anthropic did not immediately reply to Fortune’s request for comment seeking more information about why Claude may be telling users to go to sleep. Yet, Sam McAllister, a member of the staff at Anthropic, wrote in a post on X that the behavior is a “Bit of a character tic.” 

“We’re aware of this and hoping to fix it in future models,” he added in the same post.

Experts tell Fortune that Claude’s insistence on sleep is potentially rooted in its training data. Rather than being “thoughtful,” as some described it, Jan Liphardt, a Stanford bioengineering professor said the large language model may merely be repeating a phrase used in its training data in similar situations. 

“It doesn’t mean that the frontier model has suddenly become sentient,” said Liphardt, who is also the CEO of OpenMind, which builds software for AI-connected robots. “It doesn’t mean that this model has now come alive. It’s reflecting that it’s read 25,000 books on humans’ need [for] sleep, and humans sleep at night.”

Leo Derikiants, the co-founder and CEO of Mind Simulation Lab, an independent AI research lab trying to achieve artificial general intelligence (AGI), told Fortune that Claude’s rest reminders may be influenced by a system prompt acting behind the scenes. These system prompts are like hidden instructions that help guide an LLMs behavior and sets boundaries. 

One company which publishes their system prompts publicly is Grok-creator xAI, now a part of SpaceXAI. Grok’s instructions on Github, for instance, list several safety considerations including not assisting users asking about violent crimes. Yet, because of Musk’s branding of Grok as “brutally honest,” Grok 4’s system prompt also encourages it to, in certain cases, ignore restrictions imposed by users and “pursue a truth-seeking, non-partisan viewpoint.”

It’s also possible that Claude is seizing upon the “go to sleep” language as a way of managing larger context windows, Derikiants said. LLMs like Claude, can only reference a limited amount of information at once. When the context window is nearly full, that may encourage the LLM to introduce wrap-up phrases such as “good night.” The definitive reason, though, requires further research by Anthropic, he added.

Despite the seemingly logical explanations that may explain the behavior, users could be forgiven for seeing the response as evidence of some leap in intelligence on the part of LLMs. The pace of innovation in the AI race has led to increasingly frequent updates and new model releases.

Just in the past month, OpenAI has released GPT 5.5, which OpenAI president Greg Brockman called an advancement “towards more agentic and intuitive computing.” Meanwhile, Anthropic released Opus 4.7 publicly last month while it held its most capable model, Mythos, back from public release because it said it was too dangerous.

Liphardt said AI is advancing so rapidly it is increasingly common for people to assign human characteristics to AI. As these systems get better at mimicking empathy or concern, he warned, it becomes easier for users to forget they are interacting with pattern-recognition engines. 

“I’m continuously surprised by how quickly people, when they interact with a frontier model, project life into it and develop strong connection.”

This story was originally featured on Fortune.com

While this moment represents a huge opportunity for buyers and investors, not every business owner wants, or has the opportunity, to sell. Many owners spent decades building their companies and bristle at the idea of selling to traditional private equity or a bigger company that could restructure their business, lay off employees, and jeopardize their legacy. 

Some businesses will pass down to family members, but the majority face closure and 27% of owners 55 and above are unsure of their succession planning.

A solution? Steward Ownership models which combine self-determination and purpose orientation. These allow companies to resist speculative takeover and embed mission and values for the benefit of employees, customers, and/or the broader community.

Take Rick Plympton and Mike Mandina, respectively former CEO and Founder of Rochester, N.Y.-based high precision optics company Optimax. Since the 1990s, Rick and Mike grew Optimax from 10 employees struggling to make payroll to 500 employees with ~20% annual revenue growth. When the time came to discuss succession, the pair agreed selling wasn’t an option—no matter the potential upside. 

“Mike and I grew up blue collar—we don’t need to make billions of dollars,” said Plympton. “We wanted to create a corporate structure where the company could continue to grow and create jobs here in our community.”

They chose to convert Optimax to an Employee Ownership Trust. In 2020, Rick and Mike donated 20% of their equity to the trust. Over time, Rick and Mike will sell their outstanding shares back to the company after which the trust will fully own Optimax. Optimax’s profits will then be shared with employees or reinvested to grow the business, ensuring jobs and opportunity remain within the Rochester area.

“In the first 30 years of Optimax, we did $500 million in revenue, and roughly half of that was shared with our workforce through salary, benefits, and bonuses … The janitor gets the same monthly bonus check as the president, and every employee has a pathway to becoming a millionaire if they work with us for 30 to 40 years,” said Plympton. “If we can get 10 or 20 more companies in the region to convert to employee ownership, we’ll change the financial dynamics of the entire community.”

As more and more business owners begin to explore what succession looks like for them–whether it’s around the corner, or still five or ten years away–many are looking for alternative approaches beyond traditional sale. While employee ownership was the right move for Optimax, others have embraced models which enable continuity and giving back, such as:

• Newman’s Own: Paul Newman had not yet won his 1987 Oscar for “The Color of Money” when he and his long-time friend A.E. Hotchner started selling salad dressing in 1982. From the beginning, Paul decided to give all the profits away to good causes. When he passed away in 2008, he gifted the food company to Newman’s Own Foundation, a move which caused a stir until the passage of the Philanthropic Enterprise Act in 2018. 
• Kensington Corridor Trust: In Philadelphia, PA, the neighborhood trust holds property and resources collectively. The neighborhood governs how the assets are used, maintaining local control in perpetuity while using profits for reinvestment. 
• The Walker Group: The Connecticut-based IT firm transitioned to a Perpetual Purpose Trust in 2023 to conserve its long-held mission of serving their people and community. Through the trust, they distribute half of their profits to good causes, and half to their employees.

While these models remain the exception not the rule in the US, more are following in Paul Newman, Rick Plympton, and Mike Mandina’s footsteps. Patagonia converted to a Perpetual Purpose Trust in September 2022, while Michael Bloomberg declared he would donate his shares of Bloomberg LP to a trust to finance Bloomberg Philanthropies in 2023.

Foundation-controlled enterprises are more common in Europe, with companies like Rolex, Bosch, Carlsberg, Lego, and Novo Nordisk paving the way. Thanks to the success of GLP-1 drugs, Novo Nordisk Foundation is now the largest foundation in the world. The second largest foundation? Tata Trusts which owns 66% of Tata Sons, the holding company that owns Tata Group–one of India’s largest and best-known business groups.

When you’ve spent decades building, your business is more than just the number of zeroes you can sell it for. Steward Ownership models give you options to pass on what you’ve built for the benefit of your employees, your community, and your legacy.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.

This story was originally featured on Fortune.com

The move by Miran was expected given his seat on the Fed’s Board of Governors will be taken by Warsh.

In a letter announcing his resignation, Miran was critical of the Fed’s approach to measuring inflation and said if the central bank “doesn’t adjust for these errors, it will run unemployment higher than it has to, fighting fake rather than real inflation.”

Miran also said he’s excited about the changes that Warsh plans to make at the Fed, including a shake-up of its communications and balance sheet policies.

Miran, who joined the Fed in September, initially took unpaid leave from his White House job when joining the central bank, before resigning from that post in February. A noted dove, Miran voted for interest-rate cuts in every policy meeting during his period as a policymaker, including dissenting votes urging the central bank to cut by more than fellow officials agreed.

Read More: Senate Confirms Kevin Warsh to Lead Fed in Narrowest-Ever Vote

The Senate on Wednesday narrowly confirmed Kevin Warsh as chair of the Fed, clearing the way for his swearing soon after outgoing Chair Jerome Powell’s term ends on Friday.

This story was originally featured on Fortune.com

Panthalassa, a U.S.-based start up betting on ocean waves to power a fleet of floating data centers, announced $140 million in funding earlier this month led by Thiel.

The funding pushes Panthalassa’s valuation close to $1 billion, the Financial Times reported, citing a person familiar with the terms of the deal.

Panthalassa didn’t immediately respond to Fortune’s request for comment.

“The future demands more compute than we can imagine,” Thiel said in a statement announcing the funding. “Extra-terrestrial solutions are no longer science fiction. Panthalassa has opened the ocean frontier.”

Surging AI demand in the U.S. has hit snags with an aging and beleaguered grid system vulnerable to extreme weather. Energy analytics firm Wood Mackenzie noted in a report earlier this year that data center development has slowed down due to limited electricity capacity growth. Bottlenecked by energy infrastructure from as far back as World War II, tech companies are looking for extraterrestrial arenas to build out data centers without taxing available resources.

Panthalassa, sharing the name with the superocean that once surrounded the supercontinent of Pangaea, has designed “nodes” that sit on top of the ocean, with nearly 280-foot-long steel structures extending below the surface. The sphere at the top of the node bobs in the water, with the attached tube oscillating water within it, spinning turbines inside the structure that generate electricity. The structure also holds a hermetically sealed, or airtight, AI server that is cooled by surrounding seawater.

The company spent years designing Ocean-1 and Ocean-2 prototypes, beginning in 2021, and plans to deploy its most recent Ocean-3 system in the northern Pacific Ocean this year, with commercial deployment to follow in 2027, according to the company.

“There are three sources of energy on the planet with tens of terawatts of new capacity potential: solar, nuclear, and the open ocean,” Panthalassa cofounder and CEO Garth Sheldon-Coulson said in a statement. “We’ve built a technology platform that operates in the planet’s most energy-dense wave regions, far from shore, and turns that resource into reliable clean power. We’re now ready to build factories, deploy fleets, and provide a sustainable new source of energy for humanity.”

Thiel’s bet on subsea data centers

Thiel, the billionaire cofounder of Palantir and PayPal, has taken a special interest in the ocean as the next frontier for human development. For nearly two decades, the investor has shown curiosity in seasteading, or the Libertarianism-inspired creation of permanent, autonomous islands on the high seas independent from other countries’ jurisdictions. In 2008, Thiel donated $500,000 to The Seasteading Institute, an organization founded by a former Google engineer and Sun Microsystems programmer to spearhead these utopian-adjacent sea communities.

“It hits such a nerve because in this idea of starting a new country or doing something new, it reminds people that if we had to design a country from scratch—design the state of California from scratch—it would be so different, and there’s all these super corrupt governance institutions we’d clean out,” Thiel said in a 2018 interview.

Though Thiel’s belief in the near-future viability of seasteading appears to have waned, his interest in ocean-based ventures has not. Thiel’s venture capital fund Founders Fund backed Panthalassa in 2018, but Thiel is now investing in the startup through his personal fund, according to the FT.

But like seasteading, scaling AI infrastructure in earth’s massive bodies of water may not see immediate success. While previous iterations of subsea data centers have shown promise—Microsoft’s Project Natick, which ended in 2024, submerged nearly 900 servers in the Scottish Sea and had fewer submerged servers break compared to control servers on dry land—researchers still see risks associated with scaling subsea systems.

“The main advantages of having a data center underwater are the free cooling and the isolation from variable environments on land,” Md Jahidul Islam, a University of Florida professor of electrical and computer engineering who authored a study on the potential dangers of subsea infrastructure, said in an interview last year with the university.

According to Islam’s research, dense water can carry acoustic signals faster than air, making them vulnerable to acoustic attacks. Isolated data centers underwater can also be harder to monitor and, therefore, to identify if a component breaks.

“These two advantages can also become liabilities,” he said. 

This story was originally featured on Fortune.com

The settlement resolves a DOJ investigation into PayPal over a 2020 program that guaranteed a $530 million investment for Black and minority-owned businesses. The government claimed that the program was unlawful under the Equal Credit Opportunity Act, which bars credit discrimination based in part on race or ethnicity.

PayPal will launch a new program for veteran-owned small businesses or those in farming, manufacturing, or technology. However, the settlement stipulates that PayPal admits to no wrongdoing and will not pay a separate fine to the federal government.

“This Department of Justice is delivering on President Trump’s vow to root out illegal DEI from every corner of corporate America,” Acting Attorney General Todd Blanche said in a statement. Despite the Trump administration’s claims, HR Brew has found that diversity, equity, inclusion, and belonging initiatives are still widespread throughout corporate America, even if they go by a different name.

“For more than two decades, PayPal has helped small businesses start, scale, and thrive by expanding access to digital financial tools,” Taylor Watson, a PayPal spokesperson, said of the settlement in a statement via email. “We’re excited to launch the Small Business Initiative to infuse American small businesses with even more economic opportunity.”

The agreement comes just one month after IBM agreed to pay $17 million in damages to the DOJ over its DEI programming, HR Brew reported previously. Similarly, IBM did not admit to any wrongdoing and the government took the settlement as a victory against DEI.

While the PayPal settlement is different from that with IBM, David Glasgow, co-founder of the Meltzer Center for Diversity, Inclusion and Belonging at New York University School of Law, told HR Brew that the two are linked.

“This administration has clearly signaled that it’s not just going to look at what organizations are doing now, or what they’ve done recently, since the anti-DEI crackdown,” he said. “They appear quite willing to say, ‘No, actually, we’re going to peek behind the curtain and look at what you were doing in the heyday of 2020, relating to DEI. And if you were doing anything that we considered to be unlawful back then, then we may come after you as well.’”

This report was originally published by HR Brew.

This story was originally featured on Fortune.com

The Sierra Nevada tourist hub—home to ski resorts, lakeside casinos, and roughly 25 to 28 million annual visitors—is facing an energy crisis with a familiar culprit: the data centers powering the AI boom.

NV Energy, the Nevada utility that has supplied the bulk of Lake Tahoe’s electricity for decades, told Liberty Utilities—the small California company that services the region—that it will stop providing power after May 2027. The reason? NV Energy needs the capacity for data centers. As in: the energy supplier for the Lake Tahoe region is telling the utility company that it has less than a year to find another power source.

Northern Nevada has become one of the fastest-growing data-center corridors in the country. Google, Apple, and Microsoft have either built or are planning facilities around the Tahoe-Reno Industrial Center east of Reno. The Desert Research Institute, using data from NV Energy’s 2024 Integrated Resource Plan, found that the 12 data center projects located overwhelmingly in Northern Nevada could drive 5,900 megawatts of new demand by 2033. At a regional business event last fall, NV Energy’s director of business development called the moment “unprecedented,” saying the company was eager to serve the new industrial load but that it would not “impact our existing customer base.”

But Liberty’s 49,000 California customers may already be bearing the cost. Liberty Utilities generates about 25% of its power from solar facilities it owns in Nevada. The other 75% comes from NV Energy, and that source will no longer be supplied to the region by this time next year. 

“It’s like we don’t exist,” Danielle Hughes told Fortune. Hughes is a North Lake Tahoe resident, CEO of the nonprofit Tahoe Spark, and a supervisor within the California Energy Commission’s Efficiency Division. 

A jurisdictional knot with no easy fix

What makes Tahoe’s crisis so difficult is that no single regulator oversees the entire chain from power generation to customer bills.

Liberty is a California investor-owned utility. Its customers live in California and pay rates approved by the California Public Utilities Commission. But Liberty’s grid sits inside NV Energy’s balancing authority, connects to NV Energy at 38 points, and relies entirely on Nevada transmission lines, according to a Liberty filing with state regulators. Liberty’s territory is a small sliver along California’s eastern border, sitting within NV Energy’s balancing zone rather than the California Independent System Operator, which coordinates the grid for virtually every other ratepayer in the state.

Building a direct connection to California’s grid would require a new transmission line west over the Sierra, a project Liberty President Eric Schwarzrock said would cost “hundreds of millions of dollars” with significant land impacts.

The CPUC approves Liberty’s rates and procurement requests, but it cannot order NV Energy to keep selling wholesale power or dictate how Nevada plans for data centers. That falls to the Federal Energy Regulatory Commission, which regulates interstate transmission and wholesale electricity sales. With NV Energy and Nevada regulators controlling the upstream grid, the result is a system where California sets the rules, Nevada runs the wires, federal jurisdiction applies to the wholesale market, and no single entity is accountable for the outcome.

In March 2026, Liberty asked the CPUC to authorize an expedited request for proposals for replacement energy beginning June 1, 2027. In that filing, Liberty said NV Energy had cited data centers in the Tahoe-Reno Industrial Center area and northern Nevada transmission constraints, among other reasons, for ending full-requirements service.

Hughes and the Sierra Club’s Tahoe Area Group want the commission to reject that approach and instead open a full proceeding. In an April 1, 2026, letter to CPUC commissioners shared with Fortune, Sierra Club Vice Chair Tobi Tyler argued that the scale of the procurement—affecting 49,000 ratepayers dependent on an isolated, rapidly transforming grid—demands the transparency and public participation that only a formal proceeding provides. Tahoe Spark’s underlying protest states that “California does not produce a Liberty-specific forecast of demand, peak conditions, or procurement needed for numerous California communities in a high wildfire risk area.”

“You need to open a full proceeding and do a transparent process and understand what we look like in California policy, and what the long-term game is,” Hughes said. Even regulators are still sorting through the legal boundaries, she added: “They’re basically trying to decide what to do right now, or even what they legally can do.”

Even the regulators are still sorting through the legal boundaries, she added: “The procurement will have to be approved by the CPUC. They’re basically trying to decide what to do right now, or even what they legally can do.”

The data center next door

Data centers used 22% of Nevada’s electricity in 2024, and that share could rise to 35% by 2030. In NV Energy’s own 2024 resource plan, about 75% of major-project load growth is attributed to data centers, according to Sierra Club expert testimony filed with Nevada regulators and reviewed by Fortune, and most of it is concentrated in Northern Nevada—using the same system that feeds power to Lake Tahoe. 

NV Energy is building Greenlink West, a 525-kV, $4.2 billion transmission line from Las Vegas to Yerington, expected online in May 2027. Schwarzrock said Liberty would be “first in the waiting line” when Greenlink opens, giving it access to a wider pool of energy providers. But that timeline matches the contract deadline exactly, leaving almost no margin for error. About 70% of the project’s costs will be borne by Southern Nevada customers.

But this is nothing new, at least according to NV Energy.

Katie Jo Collier, a spokesperson for the utility, said the transition was rooted in a longtime understanding with Liberty “well before data center load growth was a consideration,” calling it “a planned transition for many years, not a reaction to recent developments.” NV Energy sold its California electric assets to Liberty in 2009 and agreed to keep supplying power temporarily. That arrangement was extended in 2015, again in 2020, and once more in late 2025, and each time because Liberty had not yet secured an independent supply, a timeline corroborated by regulatory documents reviewed by Fortune.

But independent experts have questioned whether NV Energy’s own demand projections are reliable. In testimony filed with Nevada regulators in Oct. 2024, energy economist Rose Anderson of Synapse Energy Economics warned that NV Energy’s major-project load forecast is ‘highly uncertain’ and that existing customers could end up paying for infrastructure built to serve industrial demand that never materializes.

In statement to Fortune, NV Energy reiterated its spokesperson’s statement about the transition having been planned for years. “It is important to note that the NV Energy’s wholesale agreement with Liberty Utilities was always intended to be temporary and transitional, with a clear timeline and multiple extensions over the years to support Liberty’s long-term planning efforts to secure its own access to energy supply. To ensure reliable ongoing service to Lake Tahoe customers, NV Energy agreed in 2025 to continue providing energy service to Liberty’s customers until their transmission access is available and/or until Greenlink is online.”

“NV Energy has provided reliable service to Liberty’s customers in the Lake Tahoe Basin for years and fully intends to continue that commitment while Liberty secures its own transmission access and energy to supply those customers,” the utility added. “NV Energy has taken proactive steps to ensure there is no service disruption to Liberty customers, now or in the future.”

Rates were already climbing

The supply crisis arrives on top of an existing affordability fight. In its 2025 general rate case, Liberty originally sought a 19.1% revenue increase—about $37.51 more per month for the average residential customer, according to CPUC filings. The CPUC approved a smaller increase: 11.4%, with a 9.75% return on equity rather than Liberty’s requested 11%.

The rate case spotlighted wildfire costs, insurance premiums, and infrastructure spending in a high-risk mountain region. The CPUC decision noted Liberty’s wildfire exposure and its exclusion from California’s AB 1054 Wildfire Fund, suggesting that rising insurance costs (quoted at over $30 million alone) for small utilities could warrant future rule making.

Tahoe Spark opposed the rate-case settlement, arguing that it failed to examine the interstate wholesale power structure underlying the costs paid by California ratepayers. Hughes said the problem is not merely high rates but the way costs are allocated in a region where visitor demand, second homes, ski resorts, and development projects drive infrastructure needs that permanent residents pay for.

“We’re the cost of being redistributed onto a declining community, and that is a crisis,” Hughes said.

Hughes argues that Tahoe is treated as a wealthy vacation-home market even though its year-round residents include low-income communities and essential workers. “Even though we have low-income communities in both South Lake Tahoe and North Lake Tahoe, Kings Beach, both the Energy Commission and the California Public Utility Commission do not include us in any of their socioeconomic plans,” she said.

The basin’s government structure compounds the accountability problem. Lake Tahoe spans two states, multiple counties, one incorporated city, and the Tahoe Regional Planning Agency. County supervisors, state appointees, utility regulators, and resort developers all touch parts of the system, but no single body owns the whole problem. Liberty’s demand pattern illustrates how different this territory is from the rest of California: while most regional utilities peak in summer, Liberty’s demand crests around Christmas, when second-home owners arrive for ski season — driving infrastructure costs that year-round residents bear.

What happens next

Liberty has told customers that NV Energy will remain the transmission provider—the wires aren’t going anywhere. The question is who supplies the electricity that flows over them, what it costs, and whether California regulators can protect customers whose upstream grid sits outside California’s usual planning structure.

Schwarzrock said the utility plans to bid the replacement contract to “anybody and everybody,” focusing first on meeting California’s renewable energy requirements. Liberty anticipates issuing a formal RFP in summer 2026, with replacement power most likely coming from sources outside California, delivered over NV Energy’s transmission system.

Hughes said short-term replacement power is likely available from elsewhere in the West—but she’s not optimistic about what comes after. “Short term, you can commonly get good deals, but it’s unstable,” she said. “The short-term deal gets you through. But then you’re in the Western market, competing against PG&E, Southern California Edison, data centers, and mining companies. We’re 49,000 customers. We have no leverage.”

Her larger concern is that as California and Nevada move toward a more integrated Western electricity market, Tahoe’s small customer base will be increasingly exposed to competition from larger utilities and industrial buyers with far more purchasing power.

“We have no representation,” Hughes said. “It’s resource extraction.”

This story was originally featured on Fortune.com

Bill Gates-backed Fervo aims to take the nascent industry mainstream to power the AI boom in short order. The technology combines old-school geothermal science with modern oil-drilling and fracking techniques to develop reservoirs and power plants anywhere the customers want—as opposed to traditional geothermal power, dating back over 100 years, which relies on naturally occurring reservoirs.

“Geothermal is resonating because it’s a proven technology,” Fervo cofounder and CEO Tim Latimer, 36, told Fortune. “With the amount of growth going on in power right now, there’s going to be a lot of everything built, but the product offering we have is 24-7, carbon-free, and can be built quickly. I think it’s very unique.”

After raising $1.89 billion in an upsized IPO—‌70 million shares at $27 apiece, up from an initial range of $21 to $24—shares quickly spiked to a close of $36.54 on their first day of trading. That easily exceeded the April IPO of next-generation nuclear startup X-energy, backed by Amazon.

Fervo is different type of next gen, also known as enhanced geothermal systems (EGS, not to be confused with ESG), designed to address the scalability limitations of traditional geothermal. Named from the Latin word meaning “to boil,” Fervo’s technology utilizes water heated underground to generate steam for electricity-producing turbines, then pumps the cooled water back into the subsurface to be reheated and recycled.

Fervo brought a pilot plant for Google online in Nevada three years ago and is currently building its first commercial plant in Utah, called Cape Station. The 500-megawatt project—enough to power nearly 400,000 homes—will begin delivering some power by year’s end and come fully online in 2028. Southern California Edison is the primary customer.

“Whether it’s utility companies or hyperscalers, they’re looking at what is cost effective, what’s low risk, and what can be built quickly. Those are the things that really matter right now,” Latimer said.

Cape Station entails drilling wells 10,000 feet deep—about 2 miles—and then directionally drilling horizontally another 7,500 feet to create adequately sized reservoirs, and fracking (hydraulic fracturing) the rock to release the flows of water naturally heated to more than 400 degrees. Utah—and much of the western U.S.—is geographically ideal because the necessary heat levels are accessible at shallower depths than most of the country.

“Two miles is still a mind-boggling depth, bit it’s well within the realm of what we’re doing today,” Latimer said. “As the technology cost structure continues to improve, we’re really excited to develop and expand beyond the western United States.”

Geothermal versus everything

Wind and solar power have intermittency issues and their federal tax credits are expiring. Nuclear requires acquiring nuclear fuel and disposing of waste. Fusion remains unproven commercially. Gas-fired power plants must account for fuel price volatility and rising construction costs. Geothermal doesn’t require fuel, produces no waste, and its tax credits run through 2033.

Latimer acknowledges Fervo’s costs are still too high but says that will rapidly change as it scales. The medium-term goal is to cut costs by more than 50%—from  $7,000 per kilowatt to $3,000—approaching the lowest-cost solar farms and gas plants. Fervo estimates reaching $5,500 per kilowatt later this year.

“That’s an important goal for us because that makes us the cheapest form of power, period, with or without any tax credits,” Latimer said. “The fact that it’s reliable, 24-7, and clean is just going to be a bonus on top.”

Geothermal was disadvantaged until 2022 when President Biden’s Inflation Reduction Act included it in the renewable energy credits. And President Trump’s One Big Beautiful Bill attacked wind and solar with accelerated expiration dates while maintaining geothermal’s tax breaks.

“I think it really speaks to the bipartisan nature of geothermal. If you make a product people want, it kind of cuts through the noise,” Latimer said.

Fervo is backed by Gates’ Breakthrough Energy, Google, and others in Silicon Valley, as well as oil and gas players including Devon Energy and fracker Liberty Energy—whose founder and former CEO, Chris Wright, now serves as U.S. energy secretary.

If anything, the industry has been accused of working too hard to appeal to the Trump administration. The academic journal Energy Research & Social Science published an article this May called, “‘The smokin’ hot trophy wife of the oil and gas industry’: The role of petro-masculinity in geothermal rhetoric and policy,” citing a 2025 event called “MAGMA” (Making America Geothermal: Modern Advances) where a Texas Geothermal Energy Alliance leader used that phrase and accusing the industry of “sexualized rhetoric.”

Latimer distanced himself from such language. “It does our industry a disservice. I think we have something that people like naturally. It can lead to more affordable, reliable, clean electricity, and I don’t see the need to dress it up and use inappropriate language to try to get people excited about it.”

Oil and gas roots

A Texas native, Latimer started right out of college in 2012 as a drilling engineer with BHP just as the Australian mining giant was investing in the U.S. shale oil and gas boom. Fascinated by the rapidly evolving drilling and fracking technologies, he felt he “missed the boat” out on the beginnings of the boom.

“I had the urge to be on the frontier, and then I saw that you could apply these technologies to geothermal, which is a carbon-free energy resource that’s a little bit distinct from oil and gas,” he said. “It got me obsessed with the idea of geothermal.”

He left BHP to earn a master’s degree at Stanford University, where he met Jack Norbeck, who also was studying geothermal reservoir engineering. They cofounded Fervo in 2017.

“We saw that horizontal drilling would be this huge unlock that would transform the geothermal sector just like it did for oil and gas,” Latimer said.

Unlike oil and gas—where wastewater disposal underground is the leading cause of incidental manmade earthquakes—geothermal recycles produced water in a closed loop. “The water just goes down cold and comes back hot,” he said. “Through four years of operations, we’ve had no challenges with seismicity associated with our projects.”

Traditional oil and gas giants, such as SLB and Baker Hughes, are now investing in geothermal. Ormat Technologies, the only other publicly traded U.S. geothermal firm, recently partnered with SLB to develop EGS—though Ormat is a 60-year-old company pivoting from traditional geothermal, while Fervo has focused on EGS from the start.

Fervo holds leasing options on roughly 600,000 acres across the western U.S and is assessing its next 10 sites, with the potential to eventually generate a whopping40 gigawatts of power—enough to power roughly 30 million homes. For context, only 4 gigawatts of geothermal power have been developed in all U.S. history.

“It could be a massive, massive amount of electricity,” Latimer said. “We want to move very quickly and do this at a scale the geothermal industry has never seen before.”

This story was originally featured on Fortune.com

I was wrong.

Greg Abel is the right CEO for today’s BRK — actually, a better fit for BRK today than Buffett would be. My thinking changed when I started thinking about Tim Cook, who recently announced his departure from the CEO position at Apple. 

Tim Cook is no Steve Jobs. But there is no other Steve Jobs. The probability of another Steve Jobs replacing Steve Jobs is nearly zero. As I look at Tim Cook’s tenure, I can see that if he and Steve had run Apple together, the company could have been a lot more successful. Tim, as Jobs himself described him, “is not a product guy,” and so he failed to come up with another product of the iPhone’s magnitude. The Apple Car, which could have been that product, was a giant failure with several changes of direction and eventual disbandment. Apple Vision Pro felt half-baked — I’m not sure Steve Jobs would ever have released it. Siri, revolutionary when it launched on the iPhone, today has the IQ of a toaster compared to other AI models. Apple just settled a class-action lawsuit for exaggerating the AI capabilities of its newest iPhones. 

And yet, Tim Cook did an incredible job running Apple. Through small, incremental improvements, he cemented the iPhone as an indispensable device. In the Jobs era, the Apple ecosystem was its biggest competitive advantage — Cook doubled down on it, with all devices working seamlessly together. He extended Apple’s software advantage into hardware, abandoning Intel for in-house M chips, which power MacBooks with multiples of the battery life of the Intel processors they replaced, without sacrificing performance. Apple is no longer the most creative company in Silicon Valley, but everything works. Apple’s revenues are up nearly 4x since Jobs passed away.

I remember reading in Walter Isaacson’s biography of Steve Jobs that Jobs’s advice to Cook (paraphrasing) was: don’t ask what Steve would do — do you. That’s exactly what the people running Disney did not do after Walt Disney passed away, and they almost ran the company into the ground.

I’d argue that if we had to choose between Jobs’s creativity and vision and Cook’s ability to run an insanely complex supply chain and manufacturing operation, Cook was more important to Apple than Jobs at the time of Jobs’s passing. And then again, talent like Jobs’s is almost irreplicable, while Cook’s talent is much easier to replicate.

Companies need different CEOs at different stages of their lives, because they’re solving for different problems. This brings me to BRK.

BRK today comprises GEICO, a consumer auto insurer; the reinsurance operations run for decades brilliantly by Ajit Jain — who has now named Charlie Shamieh of Gen Re as his successor; BNSF, one of the largest railroads in the country; a collection of other operating businesses; a portfolio of marketable securities (Apple, Coke, Amex, etc.); and roughly $400 billion of cash. 

BRK requires three skill sets today. The first is replacing Ajit Jain, who will be very difficult to replace — though the succession is now identified. The new CEO’s job is to make sure the right people are running that business. The second is running the rest of the BRK portfolio of private companies, and this is where BRK needs the most help. Buffett was never a traditional CEO. He loved investing (capital allocation), not managing people, and he avoided conflict at all costs. He bought businesses, let managers run them, collected the cash flows, and reinvested. Today, BRK has a collection of more than 100 operating businesses. BNSF and GEICO are the ones that matter most, and both have become hallmarks of mediocrity.

GEICO is being beaten up by Progressive, which is eating its lunch for breakfast, lunch, and dinner. Progressive overtook GEICO as the second-largest U.S. auto insurer — while GEICO was investing in marketing, Progressive was investing in technology, and GEICO is now stuck with hundreds of legacy IT systems while Progressive can reprice almost daily. BNSF is one of the most undermanaged and least profitable Class I railroads in the U.S.

Buffett famously said you want to own businesses an idiot can run, because someday one will. Both GEICO and BNSF have substantial moats and have survived under-management, but Buffett’s statement is less true today than it was decades ago. The half-life of a moat is shrinking much faster in the age of AI.

As an analyst, I can look at BNSF and GEICO numbers and see they are being under-managed. As a consumer, I can observe that Dairy Queen, a company BRK bought in the late ’90s, has been substantially mismanaged. DQ is a beautiful business: most of its stores are franchised, so they require no capital. Under the right management, it could have tripled in size. The quality of its ice cream has not changed, but the innovation in food and the look of the restaurants have declined under BRK ownership. Every store I’ve visited looks like it’s run by a mom-and-pop. But instead of micromanaging DQ, Buffett had more important decisions to make, like buying Apple or the Japanese trading conglomerates.

BRK has reached a size where, absent a real financial dislocation, capital allocation is unlikely to be the source of forward returns. The low-hanging fruit is improving the performance of BRK’s core holdings, and maybe even shedding companies that shouldn’t be in the BRK portfolio. Greg, a billionaire and a large shareholder of BRK, has proven to be a shrewd operator of BRK’s energy business. Choosing Greg was one of the most important decisions Buffett made in decades. At his first annual meeting, we could see why. A corporate Mr. Fix-It is walking through every business, identifying key performance indicators, installing the right incentives, bringing technology to them, and replacing managers who need replacing — doing things Buffett could not and would not do, but that need to be done.

Abel is not Buffett and that is okay. In fact, it is a good thing. Greg Abel may not draw 40,000 people to Omaha for the annual meeting. But he’ll make the difficult decisions Buffett didn’t want to make. He’ll make the trains run on time, literally and figuratively.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.

This story was originally featured on Fortune.com

Over the past couple of years, public discussions about AI sustainability have largely been drowned out by headlines about the race for computing power, energy, and geopolitical advantage.

But two experts are trying to bring green AI back into the conversation. Sasha Luccioni built a prominent profile over the past five years as AI & climate lead at open source AI company Hugging Face. Now, she and Boris Gamazaychikov, the former head of AI sustainability at Salesforce, say they plan to help organizations make AI sustainability practical and measurable — through rigorous studies examining AI’s environmental impacts, research-driven guidance on AI strategy and procurement, and tools and frameworks that developers and business leaders can apply in the real world.

Most companies still care about sustainability goals internally, even if the public discourse has shifted toward ‘AI race’ rhetoric and beating China, she said. The pair’s newly-launched Sustainable AI Group will help businesses “better understand the choices that they can make,” she explained—where models are running and what kind of models are used to help organizations decarbonize and “de-risk their AI use as much as possible.” 

AI can be selected with sustainability in mind

The problem, Luccioni said, is that today’s AI, with its energy-hungry data centers and heat-intensive chips and servers that often require massive cooling systems, is exposing organizations to volatile costs, supply constraints, regulatory uncertainty, and growing pressure from both communities and employees. But, she added, the good news is that every layer of the AI stack can be designed and selected with sustainability in mind, whether that means choosing a fine-tuned small model over a frontier LLM, or running workloads in a data center powered by renewable energy rather than gas.

“I hear a lot of employees being like ‘Hey, we’re really worried about the environmental impacts of using AI in our work, so how do we use more responsibly?’” said Luccioni, who added that pushback and criticism of AI data centers has become a bipartisan issue both on social media and in government. 

There is tremendous confusion, for example, about how much water today’s AI data centers actually require. The reality, Luccioni said, is that cooling systems involve tradeoffs. “Either you’re wasting a ton of water or you’re wasting a ton of energy.”

Traditional water-based cooling systems rely on evaporation, meaning significant amounts of water must continually be replenished, she explained. But closed-loop systems that recirculate water come with their own costs: they require additional energy to continuously cool the water as it moves through the system.

Many use cases don’t require massive models

Either way, Luccioni said the data center debate relies on a narrative that everyone will be using massive, general purpose LLMs or generative AI models that need the huge data centers to run.

But many enterprise use cases, she said, don’t actually require massive frontier models. Instead, companies often need smaller, specialized AI systems tailored to specific tasks—such as optimizing factory energy usage or helping employees search internal documents more efficiently. Those kinds of models can sometimes run locally or on-premise, reducing both energy use and data privacy concerns.

Rather than assuming every problem requires a giant LLM, Luccioni said organizations should start by asking what they actually need AI to do and then choose the simplest, most efficient system capable of accomplishing that task sustainably.

“I think they should flip the question and say what are some things we could improve in our company? And maybe there’s some smaller solution,” she said. “Right now, there’s this FOMO, and people are rushing into it, but given the cost and commitment, it makes more sense for me to think about defining KPIs.” 

Luccioni said she has also become more convinced that market demand—rather than criticism alone—may be the strongest lever for change in the AI industry. If enough customers begin prioritizing renewable-powered infrastructure or asking tougher questions about carbon intensity and sustainability, she said, providers will eventually respond. Today, however, many companies still do not fully understand how their AI usage connects to broader sustainability commitments, and clearer communication between AI providers, enterprise buyers, and sustainability teams is still missing.

“Currently the AI market doesn’t distinguish between green and not green,” she explained. “Well, what if we get enough people to start factoring that into their procurement choices?” 

Luccioni acknowledged that efficiency gains alone may not solve AI’s environmental challenges, as overall demand for compute keeps rising as organizations expand AI usage. Still, Luccioni said she remains cautiously optimistic. “I feel we have enough existing interest, [including] Boris’ work with clients at Salesforce,” she said. “I think there’s a lot we can do.” 

With that, here’s more AI news.

Sharon Goldman
sharon.goldman@fortune.com
@sharongoldman

This story was originally featured on Fortune.com