TheChek News Flow

Staff ‘forced out’ and independence curtailed at the Nuclear Regulatory Commission

Donald Trump’s attack on the independence of the US nuclear safety watchdog has accelerated severe “brain drain” at the agency, raising the risks of future accidents, former officials have warned.

Almost 200 people have left the US Nuclear Regulatory Commission since the president’s inauguration in January, and the pace of executive departures shows little sign of slowing with the resignation of the agency’s director of nuclear security and its general counsel.  

Nearly half of the agency’s 28-strong senior leadership team has been installed in an “acting” capacity, and only three of five NRC commissioner roles are occupied. Trump sacked commissioner Christopher Hanson in June and Annie Caputo resigned unexpectedly last month.

“It is an unprecedented situation with some senior leaders having been forced out and many others leaving for early retirement or worse, resignation,” Scott Morris, the former NRC deputy executive director of operations who retired in May, said in an interview.

“This is really concerning for the staff and is one of the factors causing many key staff and leaders to leave the agency they love . . . creating a huge brain drain of talent,” he added.

Morris said any move to replace experienced nuclear safety professionals with politically motivated individuals would be a “dangerous game” that could result in problems being discovered years in the future. Almost two dozen new reactors are at present under development.

Several former NRC staff and commissioners told the Financial Times that the exodus began with a surge in private sector job opportunities due to the nuclear energy boom. But it accelerated drastically following the Trump administration’s attacks on the agency’s independence, they said.

Since taking office Trump has moved to assert presidential control over independent regulatory agencies, which were set up by Congress in part to shield them from executive interference.

In February he signed an order directing “so-called” independent agencies to submit all proposed and final significant regulatory actions for review. He has fired senior officials at multiple agencies, including the NRC, Equal Employment Opportunity Commission and National Labor Relations Board.

Earlier this week, Trump said he wanted to fire top Federal Reserve official, Lisa Cook, in a move that has raised fears that he is trying to demolish the independence of the world’s most important central bank.

The president claims the NRC suffers from a culture of “risk aversion” that has slowed the rollout of nuclear technologies. In May he directed a fundamental restructuring of the agency in consultation with a team from the so-called Department of Government Efficiency (Doge).

Trump has directed the NRC to set fixed deadlines for evaluating and approving licences, conduct a full review of its regulations and reconsider radiation safety limits.

Allison Macfarlane, who served as NRC chair from 2012 to 2014, said Trump’s actions at the NRC — when there were so many new entrants to the industry — raised safety risks.

“These are tech bros, and they are using a start-up model, move quickly and break things. And that doesn’t work in nuclear,” she said. “A lot of these folks want to sell their reactors overseas, but who is going to trust the NRC anymore?”

An NRC spokesperson said experienced, technically qualified managers were appropriately performing roles as acting office directors and related positions. “The NRC is aligned with our inter-agency partners in accelerating progress while upholding the highest standards of public health, safety, and environmental protection,” he said.

Some industry participants, such as Oklo, a small modular reactor developer, have welcomed the administration’s reform efforts. “The biggest challenge isn’t physics; it’s policy inertia,” said Jacob DeWitte, Oklo’s chief executive.

Oklo was refused a licence by the NRC in 2022 due to “significant information gaps” in its application and is reapplying.

But many executives are concerned the retreat from the agency could delay the nuclear renaissance.

“They are under an enormous amount of pressure, they’re losing people to private industry, [and] then more is being asked of them,” said James Walker, chief executive at Nano Nuclear Energy, a developer of small modular reactors. “There does need to be some sort of big investment . . . so they can grow commensurately with private industry taking off.”

Democrats warn that Trump’s “hostile takeover” of the NRC jeopardises nuclear safety and years of bipartisan progress to develop nuclear energy.  

“Hollowing out the agency will make us ill-equipped to usher in a nuclear renaissance and undermine our national safety and security, our economic growth, and our global leadership in the nuclear industry,” Sheldon Whitehouse, the top-ranking Democrat on the Senate environment and public works committee, told the FT.

Last week three members of the House energy and commerce committee wrote to Chris Wright, US energy secretary, asking him to clarify whether a Doge representative had told the agency to act as a “rubber stamp” to the department’s approval of new reactor licences.

A Department of Energy spokesperson said “unlocking nuclear energy” was critical to fuelling the artificial intelligence race and the administration was working with the NRC to return focus to the core mission of protecting human health and safety.

In May 2022, two rival U.S. battery startups announced similar coups.

Group14 Technologies, a developer of silicon anodes that promised greater driving range and faster charging in electric vehicles, said that within two years—in 2024—its product would be powering EVs made by Porsche. It was big news because, despite years of trying, no Western battery startup had managed to break into any major EV model.

Weeks later, Group14’s main U.S. competitor, Sila Nanotechnologies, said Mercedes-Benz would start using its silicon anodes just a bit later—in 2025.

But now both deals have been scotched, and Western-made next-generation batteries seem little closer to powering a major EV: On Monday, Porsche, already a year late to use Group14’s anodes, effectively shut down the division that would make the batteries. The automaker laid off 200 workers at battery-making subsidiary Cellforce Group, leaving only a skeletal research and development team. Sila CEO Gene Berdichevsky told me Mercedes’ plans were also off the board.

In separate statements, both German automakers blamed poor EV sales in the U.S. and China for their decisions; Porsche said it was also responding to U.S. tariffs.

But in abandoning their first-mover advantage in powerful new batteries, Porsche and Mercedes—long known for engineering-first cultures—reflect a Western auto industry that has generally gone cold on its strength: raw innovation. The moves illustrate how Western EV and battery companies are ceding ever-increasing ground and momentum to Chinese rivals, which have led the industry’s most recent technological advances and dominate production.

“As Western companies stagnate, Chinese firms are turning the page.…Options for U.S. battery manufacturing are narrowing,” Katherine He, an investor with TDK Ventures, wrote on Linkedin this week.

The EV and battery industries are in a malaise. Major carmakers have pushed back their timelines for adding next-generation batteries to their EVs. And most battery startups are struggling financially as venture funds have dried up for companies lacking commercial revenue.

Last week, for instance, Group14 said it had raised $463 million in a Series D round. But the round, led by South Korea’s SK, an industrial group that has a battery-making division, appears to have included in-kind assets: In the announcement, Group14 said that in addition to the fundraise, it had obtained full ownership of a South Korean battery plant in which it already held a 25% stake. It bought the remaining 75% from SK. I asked Group14 CEO Rick Luebbe whether the $463 million included the in-kind value of that 75% share. He told me the companies were not releasing the terms of the acquisition.

Explain how Apple is positioning itself in the AI talent war against other major tech companies.

Ironically, the current industry gloom seems in large part the result of a massive miscalculation. Starting about 2020, investors and the EV and battery industry’s main players announced combined spending in the hundreds of billions of dollars to prepare for the consumer EV bonanza that everyone (aside from Toyota) expected.

Starting in 2022 and 2023, however, Ford, General Motors, Volkswagen and the battery makers that hoped to power their EVs noticed pronounced consumer resistance to the new EV models, and sales that did not take off as hoped.

That’s when the narrative of an industry “slowdown” took hold, and it’s been that way ever since.

But that narrative isn’t quite right: It should never have become industry dogma that consumers would pay thousands of dollars—in some cases, tens of thousands—more for an electric version of basically the same car. I previewed the fuzzy math behind such EV forecasts as far back as early 2021.

The truth is, consumer appetite for EVs in the West hasn’t been genuinely tested yet. That test will start in two or three years, when prices are lower. Most average consumers won’t need $20,000 or $25,000 EVs to look more longingly at EVs, but they won’t pay more for one—they will fork over more or less the same regardless of whether a car is electric, combustion or hybrid.

And if it’s an EV, it will have to compete. By then, what Chinese EV makers are offering will be widely known.

The signs thus far are that Western carmakers are just as complacent as they were in the 1980s and 1990s, when high-quality Japanese competition arrived in the U.S. and ate Detroit’s lunch. Today, it’s Chinese carmakers like Xpeng, Nio and Geely that have barreled ahead with batteries that charge in five minutes and power long-range EVs with heated massage seats, minirefrigerators, high-grade leather upholstery, voice-controlled electronics and superb sound systems.

So far, Western automakers haven’t gotten the memo. China’s BYD, for instance, currently offers autonomous advanced driver assistance systems as standard equipment in 21 models. But Stellantis recently killed its effort to develop hands-off-the-wheel, eyes-off-the-road ADAS, Reuters reported this week, saying consumer demand is too low. That was a surprising assertion from Stellantis given the wild popularity of ADAS in U.S. consumer surveys. The reality is that automakers everywhere will likely have to match BYD on free ADAS if they intend to compete.

Against this backdrop, Sila Nano’s Berdichevsky said that while Mercedes’ EV was off the table, he was tailoring silicon anodes for another major automaker, which he declined to name. Sila is almost finished building a production plant in Moses Lake, Wash., capable of manufacturing those anodes at large scale. Given qualification timelines, Sila’s anodes would be in this new automaker’s EVs by 2028 at the earliest, he said. Those anodes will contain low percentages of silicon; in terms of high-silicon anodes—the ones that will significantly boost range and charging time—that’s not likely until the early 2030s, he said.

Berdichevsky is holding out hope for a wild-card event that could accelerate the timeline for silicon’s commercialization—for example, if China overplays its stranglehold on battery supplies and cuts exports of graphite, the conventional anode material silicon hopes to replace.

“Imagine having a gigafactory and suddenly you can’t get graphite, but you can get silicon carbon anodes domestically,” he said. “That could drive adoption and road-map acceleration.”

Short of that, look for next-gen EV performance in a little over five years.

• Top AI researchers are receiving tens of millions of dollars in compensation, turning them into the equivalent of professional athletes.
• Companies are acquiring entire teams and leaders through complex, multi-billion-dollar deals and investments to gain influence and expertise.
• A distinct hierarchy has emerged among AI labs, with OpenAI, Anthropic, Google’s DeepMind, and xAI at the top, while Meta’s position is more tenuous.

A central question looms over artificial intelligence: How is the intensifying the war for talent impacting the race for artificial general intelligence (AGI)?

We asked our new Deep Research tool and got the following answer:

• Unprecedented Compensation: The compensation packages being offered are staggering. While OpenAI CEO Sam Altman has claimed Zuckerberg offered as much as $100 million in signing bonuses to lure away his talent, even more conservative reports point to multiyear packages worth tens of millions of dollars.1,2 These “eye-watering offers” have turned top AI researchers into the equivalent of professional athletes, commanding massive paydays.1
• Strategic ‘Acqui-hires’ and Investments: Companies are moving beyond individual hires to acquiring teams and leaders through complex deals designed to bypass lengthy antitrust reviews. Meta’s $14.3 billion investment for a 49% stake in Scale AI to hire its CEO, Alexandr Wang, as Chief AI Officer is a prime example.3 Similarly, Meta hired former GitHub CEO Nat Friedman and Safe Superintelligence CEO Daniel Gross, with plans to partially buy out their venture fund for over $1 billion.1 These moves are about acquiring not just individuals, but entire nodes of influence and expertise.
• A Clear Pecking Order: A distinct hierarchy has emerged among AI labs. At the very top are OpenAI, Anthropic, and Google’s DeepMind outfit, which have the clout, capital, and access to cutting-edge research to attract nearly anyone. Elon Musk’s xAI has vaulted up to join them near the top. Meta is sometimes discussed in this class, but its reputation has taken a hit since the disastrous launch of its Llama 4 model, making its position more tenuous despite its immense spending.1

• Internal Resentment: The influx of highly paid new talent is causing friction within existing teams. At Meta, some employees feel the emphasis on new superstars labels them as mediocre by comparison. At OpenAI, the massive offers from rivals have led some researchers to wonder if they should “throw a hissy fit” to secure a retention bonus.1 In response to Meta’s poaching, OpenAI is reassessing its compensation and has begun paying out substantial bonuses, some in the millions, to nearly a third of its staff to retain them.4
• The Investor Pitch: The pressure is so intense that even major investors are being deployed to keep talent in place. In a notable move, Josh Kushner, founder of major OpenAI investor Thrive Capital, gave a presentation to OpenAI staff arguing that leaving for a rival startup would be a poor economic decision due to factors like equity dilution.5 This kind of direct intervention from an investor to stanch employee churn is highly unusual and underscores the severity of the talent drain.
• The “Missionaries vs. Mercenaries” Debate: The talent war has taken on an ideological dimension. Sam Altman has publicly called Meta’s recruiting efforts “somewhat distasteful” and framed the battle as one where OpenAI’s mission-driven culture will ultimately triumph over mercenary motivations.6 This narrative aims to position OpenAI as the more authentic and dedicated player in the pursuit of AGI, a powerful non-monetary recruiting tool.

The Takeaway

• Analysts raise concerns about SoftBank’s spending

• SoftBank could influence portfolio companies to buy Intel-made chips

• SoftBank’s dealmaking likely won’t stop

In May, a Nomura Securities equity analyst asked SoftBank Chief Financial Officer Yoshimitsu Goto a pointed question on an earnings call: How could he justify SoftBank’s recent $6.5 billion purchase, using borrowed money, of Ampere Computing, a failing developer of data center chips? The startup had just $16 million in revenue and had lost hundreds of millions of dollars the year before the sale.

Goto didn’t answer the question directly. “I believe the valuation falls within a reasonable range,” he said, according to a transcript of the call, adding that SoftBank followed procedures carefully for a deal its board approved. The acquisition is still awaiting U.S. government approval. The exchange reflected simmering worries among investors about SoftBank’s spending even as the company’s investment in OpenAI is lifting its stock price.

SoftBank executives have made it clear that the whirl of dealmaking isn’t likely to stop. In February, executives at the Japanese conglomerate showed investment bankers a slide showing roughly two dozen AI and infrastructure-related deals it either had done or was considering, according to a banker who attended the call.

The latest, and one of its highest profile, is its $2 billion investment last week in Intel, the struggling chip manufacturer now partly owned by the U.S. government. The investment potentially provides SoftBank with a missing piece of its portfolio: chip manufacturing. SoftBank was most interested in the deal because of Intel’s foundry business, which is looking to compete with chipmaking giant Taiwan Semiconductor Manufacturing Co. in making chips for other companies, a person close to SoftBank said. In the past, Intel has concentrated on making its own chips, but TSMC’s rise to chip dominance has highlighted the weaknesses of that approach.

The Intel investment came the same week SoftBank bought an electric vehicle factory in Ohio from Foxconn to make AI computer servers. SoftBank will be installing those servers in future data centers as part of its planned—and delayed—$500 billion Stargate project. And in addition to purchasing Ampere, SoftBank also bought a smaller, struggling British AI chip firm, Graphcore, last year.

SoftBank CEO Masayoshi Son appears to have sufficient investor support to keep the money flowing. The company’s stock has hit record highs this year, up 79% year to date, driven in part by an increase in value for its huge bet on OpenAI, as well as the stock price of Arm Holdings, which has increased more than 160% since SoftBank took it public in the fall of 2023, although it’s only up 9% this year. Arm sells chipmaking architecture to companies like Apple and incorporates its technology into Nvidia chips.

Time and time again, SoftBank’s high-wire investment strategy has hinged on the idea that its constellation of investments could complement each other. Sometimes the strategy flops. It bet on ride-hailing companies like Uber and Grab with the idea that it could create a global transportation network—an idea that never came to pass. It wanted to stock office buildings operated by WeWork with other startups it was backing, a strategy it couldn’t pull off.

“Softbank’s approach to investing has always been an index-like approach—own every layer in the stack that feeds into a megatrend, then have all those players give business to each other to strengthen each other,” said Kevin Xu, founder of Interconnected Capital, an investment fund focused on AI infrastructure stocks.

Xu, whose investment firm owns Intel shares, said the chip deals seemed to tie together in a few ways. SoftBank’s investment in Intel, for instance, could indicate SoftBank’s willingness to influence its companies to commit to using Intel’s chipmaking foundry, a crucial sign of support for the struggling business. Intel CEO Lip-Bu Tan said in a note to employees last month that Intel’s investment in its next-generation chipmaking process, 14A, will hinge on confirmed customer commitments. It is expected to mass-produce chips by 2027 if it can secure enough customers and ensure efficient production.

Access to that chip production would also be useful if SoftBank pulls off a long-shot effort to develop its own chip. Its purchases of smaller firms like Ampere and Graphcore would give Son a bigger stable of chip designers, who “are as scarce as the AI researchers getting the big packages from OpenAI and Meta,” Xu said.

Valid Concerns

Still, some Wall Street analysts have started to raise questions about whether the heavy spending is too risky. David Gibson, an equity analyst at MST Financial, wrote in the spring, “Son’s plans for chipmaking I think remains a big risk in the short term.” He added that the purchase of Ampere “raises valid concerns that Masa is spending excessively like he did with WeWork.”

But SoftBank may have bigger ambitions. Arm reported last month that a growing portion of its revenues are from selling licenses and providing design services for SoftBank, which still owns about 87% of Arm.

That indicated SoftBank itself was building an AI chip using Arm’s technology, even if it has not announced a product. (If the effort is successful, it would provide some cold comfort for SoftBank on missing out on the meteoric stock rise of dominant chip provider Nvidia, which agreed to buy SoftBank-owned Arm in 2020 before the deal collapsed due to regulatory scrutiny.)

“Our business relationship with SoftBank has expanded to help them build towards their greater, broader AI vision,” said Arm CEO Rene Haas on the chip designer’s earnings call last month. Haas has become a close confidant of Son and sits on the board of Stargate.

The fates of SoftBank and Arm are deeply intertwined in other ways. As of last quarter, the value of SoftBank’s stake in Arm made up 48% of SoftBank’s roughly $200 billion asset value, a crucial metric that allows it to borrow more money. Earlier this year, SoftBank took out $8.5 billion of loans, adding to the tens of billions of dollars of debt it already had. Its executives have said they will maintain debt levels around or below 25% of the value of its assets.

The company has also been selling stakes in telcos Deutsche Telekom and T-Mobile to finance its AI investments, and anticipates it will get more cash from initial public offerings of startups it owns, like Japanese payments firm PayPay.

And despite anxieties among investors about SoftBank’s spending, Son continues to find enough banks to finance his AI ambitions. For Japanese and European banks that have worked closely with Son for many years, SoftBank is an extremely important customer, because few other clients in their countries repeatedly borrow large amounts as SoftBank does.

Gibson, the MST Financial analyst, estimated the company would have about a $20 billion shortfall over the next few years, which it could cover in part by selling more of its stakes in Deutsche Telekom and T-Mobile.

Plenty of questions remain about Son’s ability to make all his deals come together. But Xu noted, “None of this has to make sense at the end of the day either, because we are dealing with Masa’s whims and desires to build empires, so everyday business logic may not matter.”

Gapping down
In reaction to earnings/guidance:

Other news:

Gapping up
In reaction to earnings/guidance:

Other news: