TheChek News Flow

Wall Street Banks Prepare to Sell Billions of Dollars of X Loans

Banks are hoping to sell the X debt at around 90 to 95 cents on the dollar

Banks are getting ready to sell billions of dollars in debt borrowed by Elon Musk’s X, bringing Wall Street a step closer to exiting the nerve-racking deal that financed the social-media company’s buyout.

Morgan Stanley bankers have reached out to investors ahead of a planned sale next week of up to $3 billion of debt the bank and others such as Bank of America and Barclays lent to complete Musk’s 2022 buyout of the company then known as Twitter, people familiar with the matter said.

The banks hope to sell senior debt at 90-95 cents on the dollar, while retaining more-junior holdings, the people said. The banks just sold approximately $1 billion of debt in a private transaction to several investors, some of the people said.

The debt has been an albatross on the banks since they backed Musk’s $44 billion deal with around $13 billion in financing. The price Musk paid for Twitter was high, even at the time of his purchase, and the company’s rocky performance had knocked down the value. The deal is considered one of the worst that banks agreed to finance since the 2008 financial crisis.

To sell the debt, bankers will have to convince investors that the company’s financials have stabilized. Musk’s recent rise in power and alliance with President Trump have seemed to help change the narrative around X’s fortunes.

Investors have been reaching out to the banks and have indicated interest in buying the company’s debt because they believe that the company’s financials are on an upward trajectory, one of the people familiar said.

In a January email to staff, Musk pointed to the company’s growing influence and power, but said the finances remain problematic.

“Our user growth is stagnant, revenue is unimpressive, and we’re barely breaking even,” he said in the email, which was reviewed by The Wall Street Journal.

Hours after the Journal’s article was published online, Musk posted on X saying that he didn’t send any such email.

The banks never disclosed how they value the loans on their books. Some of X’s equity investors had written down their stakes in the company by as much as 75%.

Banks don’t finance acquisitions intending to hold the buyout debt for long; usually, they will arrange a sale to outside debt investors within months of their commitment to finance the deal. But in volatile periods, when fewer investors are lining up as buyers, banks will opt to hold on to the loans for longer stretches to avoid selling it at a discount, locking in losses.

With the X debt, the banks have waited—and waited—for a moment when both the markets and the company’s financial health would open the window to sell without huge losses.

Meantime, X’s lenders, which also include Mitsubishi UFJ Financial Group, BNP Paribas, Mizuho and Société Générale, have been collecting hefty interest payments. Loans to companies like X typically pay out several percentage points more in interest than the benchmark rate for investment-grade credit.

After Musk bought the social-media company, major advertisers fled and the company’s revenue plunged. However, the company’s financials have been steadily improving as some brands have started spending again on the platform, people familiar with the company said.

Musk in his staff email tipped his hat to the current rising fortunes of the company.

“Over the last few months, we’ve witnessed the power of X in shaping national conversations and outcomes,” Musk wrote. “We are also seeing other platforms begin to adopt our commitment to free speech and unbiased truth,” he added, in a seeming reference to Meta Platforms’ recent decision to roll back fact-checking and adopt a user-driven Community Notes system similar to the one on X.

The stakes couldn’t be higher as President Trump’s nominees and advisers signal the new administration’s intent to roll back financial regulations. If done correctly, financial deregulation can spur entrepreneurship, boost economic growth, and increase wages. But if mishandled, it risks destabilizing the financial system and triggering a crisis on a scale that could dwarf 2008.

To understand the challenge, imagine a tug of war over a bank’s risk-taking. On one side, some executives and board members pull for high-risk, high-return strategies, enticed by the large payouts if those investments succeed. On the other side, investors with significant exposure to losses pull for prudence. But what happens when those advocating for caution let go of the rope? The result is excessive risk-taking, with dire consequences for financial stability.

Why would influential investors with significant exposures stop advocating for caution? The reason stems from a concept economists call “moral hazard.” When decision makers share the upside of risky bets but are shielded from the downside, they have little incentive to act prudently. In a healthy financial system, investors who stand to lose from a bank’s failures constrain excessive risk-taking. However, when these investors believe government bailouts will rescue them when things go wrong, they stop pulling for caution and join the side advocating for more risk.

Unfortunately, moral hazard has become entrenched in the U.S. financial system. Over the years, federal regulators, including the Federal Reserve, have repeatedly bailed out investors—sometimes even uninsured or nondeposit investors in large financial institutions. The Dodd-Frank Act was passed after 2008 to prevent another similar financial crisis. Instead, the law has unintentionally worsened the risk-taking dynamics in the financial sector. By codifying the Federal Reserve’s responsibility for all “systemically important” financial institutions, the act has effectively extended the protective bailout net, encouraging ever-riskier behavior.

The collapse of Silicon Valley Bank in March 2023 underscores this problem. Although many of the banks’ deposits exceeded the threshold for government guarantees, the authorities opted to make all depositors whole. That decision again signaled that even uninsured investors can expect to be bailed out the next time a bank fails. This only deepens the moral hazard problem, incentivizing banks to take greater risks, knowing that others will shoulder the losses.

The current regulatory approach attempts to counteract the powerful risk-taking incentives unleashed by government bailouts by constructing a vast regulatory and supervisory apparatus. But this comes with enormous costs. The apparatus burdens financial institutions with significant compliance costs that are ultimately passed to the public, and it impedes the efficient allocation of credit, slowing economic growth.

While deregulation seeks to alleviate these adverse effects, it must be approached cautiously. Rolling back regulatory constraints on risk-taking without addressing the root cause of moral hazard—bailout expectations—would only exacerbate the problem. It would loosen the grip of regulators pulling for prudence without replacing it with the steady hands of private investors duly incentivized by having their wealth on the line. A similar mismanagement of the timing of financial deregulation contributed to the 2008 crisis and numerous other financial crises worldwide.

What’s needed is a credible strategy to ensure that decision makers within financial institutions have genuine “skin in the game.” This means reforming policies that shield influential investors and executives from the consequences of their actions, exposing them to potential losses. Such reforms would realign incentives, restoring the natural tension that constrains excessive risk-taking.

Credibility is key. Regulators and policymakers can’t simply promise not to bail out uninsured investors. Investors won’t believe them because officials have repeatedly shown that they will bail out just about everyone when things go awry. If investors don’t believe they are exposed to potential losses, they won’t pull for prudence. They will continue to advocate excessively high-risk strategies, expecting the government to bail them out if those gambles fail. Without credibility, the tug of war over risk will remain dangerously imbalanced.

By aligning incentives before loosening restrictions, the administration can create a robust and dynamic financial system. Research consistently shows that well-executed deregulation can boost entrepreneurship, drive job creation, and improve living standards—especially for the middle class. But the sequencing matters. If deregulation is rushed without addressing moral hazard, the consequences could be catastrophic.

The choices made in the coming months regarding the sequencing of financial deregulation will shape the U.S. economy for decades. Getting it right could encourage growth and prosperity. Getting it wrong risks repeating a key mistake that underpinned the 2008 crisis, with devastating consequences for economic and social stability.

Tech leaders gathered at the White House this week to announce an ambitious new joint venture named the Stargate Project, aimed at building new artificial-intelligence data centers in the U.S. The effort promises investments of $100 billion in the first year, and $500 billion over four years.

But details remain sparse.

SoftBank CEO Masayoshi Son will handle the financing, CEO Sam Altman of OpenAI will contribute his overall AI expertise, and Oracle executive chairman Larry Ellison will oversee the buildout. All three joined President Trump on Tuesday for the Stargate announcement.

MGX, a new AI investment firm in the United Arab Emirates with ties to the government there, will be a financing partner. Microsoft, Nvidia, and Softbank-owned Arm Holdings were named as technology partners.

There are several unanswered questions about the project, the most important being where the money is going to come from. SoftBank has about $38 billion in cash and short-term investments on its books, and $80 billion in long-term debt. The firm raised $100 billion for its Vision Fund venture capital fund, but struggled with a follow-up Vision Fund 2. Softbank’s stake in Arm was worth over $140 billion as of Wednesday, and it could choose to sell shares to help finance Stargate.

Oracle has about $11 billion cash, to go with $80 billion in debt. OpenAI is a start-up, still dependent on investor cash. It last raised $6.6 billion in October.

MGX may shoulder much of the financial burden, but it isn’t clear what resources it can muster. MGX has three current investments, according to FactSet: OpenAI; Elon Musk’s xAI; and another richly valued start-up, Databricks.

Mubadala, the U.A.E. sovereign-wealth fund, is an MGX investor. According to its website, Mubadala has about $300 billion of assets under management, suggesting Stargate would be a big lift, even for a large sovereign-wealth fund.

Among other partners listed by MGX are BlackRock, Silver Lake, Bain Capital, and Vista Equity Partners.

Those are rich pockets, but $500 billion rich? We don’t really know, and the Stargate partners aren’t providing information.

SoftBank wouldn’t comment to Barron’s on the sources of Stargate funding. Nvidia and Arm offered general support for the initiative in their replies, but didn’t clarify their roles in the partnership. OpenAI’s own press release announcing the project didn’t offer financing details.

Elon Musk quickly cast doubt on the whole initiative. “They don’t actually have the money,” Musk, a key Donald Trump ally, wrote on X. “SoftBank has well under $10B secured. I have that on good authority.”

OpenAI CEO Sam Altman defended Stargate’s funding in a social media post: “Wrong, as you surely know,” he said in a reply to Musk on X. “This is great for the country. I realize what is great for the country isn’t always what’s optimal for your companies.”

On Wednesday, Arm CEO Rene Haas didn’t provide details on financing, but told CNBC: “I would say, rest assured, that you would not have the individuals you saw yesterday and the president all together with such an announcement if there wasn’t a lot of backing from a financial standpoint.”

The debate and questions about funding highlight the significance of the expenditures and the complex nature of AI model building and deployment, and the infrastructure needed to support it.

For a sense of Stargate’s proposed scale, Microsoft offers a good comparison. Microsoft is projecting $80 billion in capital expenditures in fiscal 2025, ending June—an 80% increase from 2024, which was a 58% rise from the year before that. Stargate wants to ramp from nothing to $100 billion, 25% larger than Microsoft’s planned outlay, all in one year. It’s a big ask, considering that Microsoft had $73 billion in annual free cash flow and Stargate currently has none.

Microsoft, Nvidia, and Arm were left out of the equity discussion, suggesting their balance sheets won’t contribute to the effort.

Microsoft has been OpenAI’s biggest investor. Its Azure cloud is the home to OpenAI’s signature ChatGPT service.

The new dynamic suggests bigger AI ambitions for both Microsoft and OpenAI.

“It is entirely plausible that the Stargate news was not forced on Microsoft, but that Microsoft viewed exclusively financing OpenAI no longer made sense from a risk and capital allocation standpoint,” said UBS analyst Karl Keirstead in a note to clients. “Put another way, OpenAI’s massive scale ambitions…may have been running up against Microsoft’s desire to exclusively serve them.”

Microsoft referred Barron’s to a Tuesday blog post that briefly alluded to Stargate, but went on to redefine its close relationship with OpenAI. Most of their agreement through 2030 remains, the blog post said, but Microsoft’s Azure cloud platform will no longer have exclusivity on new OpenAI computing demand, though it retains the right of first refusal on new cloud business.

While the partnership remains, there are cracks forming. They’re likely to reverberate across the tech landscape.

High Interest Rates Are Hammering Investors. What Lies Ahead Could Be Worse.

Rising rates would be bad news for bondholders and borrowers of all stripes, particularly the U.S. government. They cast a shadow over stocks, too.

For both bondholders and borrowers, the pain of higher interest rates may have only begun.

President Donald Trump will have to decide if he wants to be part of the problem—or part of the solution. Many of the political promises that got him elected, including tariffs, tax cuts, and heavy government spending, could push interest rates higher.

The bond market has gone through one of its most wrenching periods in history, capped by plummeting prices late last year and in early January. For investors, the rise in yields from an unprecedented zero percent to near 5% has meant the worst fixed-income returns in almost 90 years.

Housing has suffered a double whammy as new homes have become unaffordable for many buyers facing 7% interest rates on mortgages, which take their cue from bonds. And prospective sellers are frozen in place rather than trading their old 3% loans for much costlier ones.

The U.S. Treasury, the world’s biggest borrower, has felt the pressure of higher bond yields most acutely. After running budget deficits previously only seen during wars or recessions, the U.S. government’s total debt has grown larger than the U.S. economy. When it could borrow at rates under 1%, this debt appeared manageable. At current borrowing costs, the government’s interest tab is the fastest-growing part of its budget, surpassing defense expenditures in 2024.

To bond investors, who are Uncle Sam’s lenders, this burgeoning debt points to still higher interest yields and lower bond prices in the years ahead. That growth is driven by ongoing deficits totaling 6% of gross domestic product, which in turn are boosted by surging interest expense on existing and new debt.

High rates are hammering borrowers. Lower-quality companies, especially those that rely on bank loans and private credit, are seeing increased credit stress. Commercial real estate is in a tightening vise as old, low-cost loans have to be refinanced at higher rates while sectors such as office buildings remain depressed.

Then there’s the stock market. Equities will suffer if bond yields keep heading higher. Late last year and into early January, stocks stumbled as the yield on the benchmark 10-year Treasury note shot up from 3.6% to 4.8%. Then equities rallied as the Treasury yield fell back by a quarter of a point on better-than-expected inflation news. The S&P 500 index hit another record this past week.

Interest rates are also front and center for President Trump. In a Thursday video address to the World Economic Forum in Davos, Switzerland, Trump said he would “demand that interest rates drop immediately,” in a public signal to the Federal Reserve.

The problem is that the bond market may not follow the central bank’s lead this time. When the Fed cut its federal-funds target by 100 basis points late last year, the 10-year Treasury yield did the exact opposite, rising to its recent peak.

Regardless of Trump’s dictates to the Fed, the bond market “is likely to price in more uncertainty because of the inherent contradictions among many of Trump’s policies, the volatility around their application (for example tariffs), and still-too-large budget deficits,” writes Steven Blitz, chief U.S. economist of TS Lombard.

Blitz thinks 10-year Treasury yields could climb to 6%, a level not seen since mid-2000. That is especially true if the Fed ends its rate cuts and lifts the fed-funds rate to 5% to 5.25% in two years, from the current target range of 4.25% to 4.5%.

A 6% Treasury yield could translate into a 7% yield on corporate bonds. “This yield level can be problematic for the equity market,” Blitz concludes. “Institutional equity market investors will likely reassess allocations, and 7% yields could accelerate the natural bias for a net outflow from the equity market among aging baby boomers.”

The current near-5% Treasury yields only look high compared with the 1.5% to 2.5% yields levels that prevailed in the decade following the financial crisis of 2008-09, notes Richard Sylla, who literally wrote the book on the subject, A History of Interest Rates.

One of the lessons of this history is that long-term interest-rate cycles typically run for 20 to 40 years, Sylla tells Barron’s. Yields bottomed around 0.3% during the Covid-19 market panic of March 2020, says Michael Hartnett, chief investment strategist at BofA Global Research. Although Hartnett predicts that bond and stock markets will rally later this year as the 10-year yield dips under 4%, he sees rising bond yields as the No. 1 long-term risk faced by investors.

Until the post-Covid updraft, yields had been tumbling. Interest rates fell over four decades, starting in 1980, with staunchly anti-inflationary monetary policies, smaller government and deregulation, and broad globalization that saw the fall of the Soviet Union, China’s emergence in the world economy, and the North American Free Trade Agreement, or Nafta, Hartnett says.

After the financial crisis, that free-market orthodoxy was flipped, with interventionist fiscal and monetary policies. Central banks drove policy rates to zero (or below, in Europe and Japan), which was augmented by massive purchases of bonds to inject liquidity, known as quantitative easing. That bond buying accommodated deficit spending by governments while lifting asset prices.

At the same time, Hartnett says, populism took hold internationally, culminating with Donald Trump’s first presidential victory in 2016 and his reelection last year. Fiscal policies turned to “extreme excess,” with the U.S. budget deficit exceeding 6% of GDP in fiscal 2024. At $36 trillion, the national debt exceeds the size of the U.S. economy.

All this makes one wonder whatever happened to the “bond vigilantes” of a previous era, says Martin Barnes, a former chief economist at BCA Research. In the 1980s and 1990s, these bond investors would rebel when fiscal and monetary policies diverged from practices that had brought interest rates down from the stagflationary 1970s. They would drive bond prices down and yields up.

Some hint of the vigilantes came with the abnormal rise in bond yields after the Fed cut last year. Sylla sees that as reaction to easing monetary policy while inflation still ran close to 3%, above the central bank’s 2% target.

Eventually, Barnes expects bond vigilantes to return, forcing the government to pay a fiscal risk premium. It’s unknown when they will exact that price. He notes that economists Carmen Reinhart and Kenneth Rogoff posited after the financial crisis that a debt-to-GDP ratio of 90% would cause economic problems. Yet U.S. total government debt exceeds 120% of GDP and hasn’t set off alarm bells.

Hartnett thinks the new administration could surprise with its spending curbs. He sees both a “devilish Trump and an angelic Trump,” with the latter slowing the deficit’s path.

Treasury Secretary-designate Scott Bessent has set an ambitious “3-3-3” plan, consisting of halving the deficit to 3% of GDP, real economic growth of 3% annually, and boosting U.S. domestic oil production by three million barrels a day. But BTIG Washington analyst Isaac Boltansky is skeptical, saying “deficit hawks are the most endangered species in Washington.”

The Department of Government Efficiency, or DOGE, led by Elon Musk is unlikely to make a dent in the deficit. Consider that in 2024 almost 60% of the $6.75 trillion in federal spending went for Medicare, Medicaid, and Social Security; while 13% went for interest on the national debt, and about the same amount for defense expenditures. Discretionary nondefense spending, the likely target for cuts, accounted for only 13% of expenditures.

The Congressional Budget Office projects the deficit to total 6.2% of GDP in 2025, shrink to 5.2% by 2027, then climb again to 6.1% by 2033. “Since the Great Depression, deficits have exceeded that level only during and shortly after World War II, the 2007-09 financial crisis, and the coronavirus pandemic,” the CBO said.

The CBO’s outlook embodies optimistic assumptions. It assumes no recessions in the coming decade. It is based on current law, notably the expiration of the Tax Cuts and Jobs Act of 2017, the signature tax measure of the first Trump administration. Renewal of the tax law, a major objective of Trump’s, would add about another $4 trillion to the debt over the next decade. And that is before adding goodies like exempting tips and taxes from overtime pay, as he promised on the campaign trail. No one knows how much tariffs will offset the shortfall at this point.

The CBO also assumes bond yields will be lower over the course of the next decade. It forecasts a 4% average 10-year Treasury yield for 2025, falling to 3.9% in 2026-29, then 3.8% for 2030-35.

Some form of fiscal discipline may be the path to fiscal sustainability. The key is for interest rates to be lower than the growth of nominal GDP. The federal debt was pared from a peak of 106% of GDP in 1946 after World War II to just 23% by 1974, in large part because of the strong growth in postwar America but also because rates were suppressed artificially.

More likely, interest rates will stay high. Bond investors are apt to demand a higher yield premium to compensate for the risk of persistent deficits, and the resulting inflation they could produce. The big warning flag for BofA’s Hartnett would be investors fleeing both Treasuries and the U.S. dollar in a global vote of no-confidence in America.

Nobody knows if, or when, that so-called riot point comes. But it is the major long-term risk facing investors in both bonds and stocks, with major consequences for anybody needing to borrow money.

Blackstone’s Schwarzman stirs London art market with record purchases

Private equity chief spends millions on 18th century society portraits by Joshua Reynolds and Thomas Gainsborough

US private equity executive Steve Schwarzman is shaking the London art market by paying record prices to acquire paintings by Joshua Reynolds and Thomas Gainsborough, the leading 18th century society portrait artists.

Schwarzman, who was given an honorary knighthood last year by the UK government, is restoring Conholt Park, a 17th century 2,500-acre estate in Wiltshire that he bought for £80mn in 2022. As he makes plans to furnish the house, he is rapidly building a collection of masterpieces by making private offers to owners through dealers.

He recently acquired Reynolds’ “Portrait of Lady Worsley” from the family of David Lascelles, Earl of Harewood, for about £25mn. The portrait, which has been at Harewood House in West Yorkshire since the 18th century, shows a society heiress who was involved in a notorious court case after eloping with her lover.

Schwarzman, co-founder of private equity group Blackstone, has also bought Gainsborough’s portrait “Lady Bate-Dudley”, a 1787 painting which has been on long-term loan to London’s Tate Gallery. It was withdrawn from the Tate collection last year after the owner sold it privately through Simon Dickinson, a London gallery.

The billionaire declined to comment this week but people close to the transactions confirmed that he was the buyer of both works. The Lascelles family decided to sell the Reynolds’ portrait to help cover the upkeep of Harewood House after receiving an anonymous offer through Christie’s private sales department.

The £25mn price is a record for Reynolds, apart from the £50mn paid by the National Portrait Gallery and the J Paul Getty Museum in Los Angeles two years ago for the exceptional “Portrait of Omai”, a Polynesian islander. The sale price of the Gainsborough is not known, but it is thought to be well in excess of the artist’s auction record of £8.2mn in 2019.

Representatives of Schwarzman have made approaches to other owners of rare 18th century paintings, according to people with knowledge of the London market. Sales have been made easier by the fact that he is not applying for export licences for the works, so there is no risk of any deal being blocked by the UK government.

London art advisers said there had been rumours for months of an anonymous buyer bidding for works at high prices. The market in 18th century paintings is thin because few tend to be offered for sale, and it often takes a buyer willing to offer a knockout price to families that have held them for decades.

Schwarzman also owns Miramar, a Gilded Age mansion in Newport, Rhode Island that he has restored and filled with art. He said last year that it would become a private art museum on the death of him and his wife Christine. His honorary knighthood was partly in recognition of his philanthropy in the UK, including a £185mn gift to Oxford university.

Reynolds’ portrait of Seymour Fleming, who became Lady Worsley by marriage in 1775, is an iconic painting of the stepdaughter of the first Lord Harewood. Her husband Sir Richard Worsley brought a case of criminal adultery against the military officer with whom she had an affair, provoking a scandal that was dramatised by the BBC in 2015.

How small Chinese AI start-up DeepSeek shocked Silicon Valley

Hedge fund billionaire Liang Wenfeng builds model on tight budget despite US attempt to halt China’s high-tech ambitions

A small Chinese artificial intelligence lab stunned the world this week by revealing the technical recipe for its cutting-edge model, turning its reclusive leader into a national hero who has defied US attempts to stop China’s high-tech ambitions. 

DeepSeek, founded by hedge fund manager Liang Wenfeng, released its R1 model on Monday, explaining in a detailed paper how to build a large language model on a bootstrapped budget that can automatically learn and improve itself without human supervision.

US companies including OpenAI and Google DeepMind pioneered developments in reasoning models, a relatively new field of AI research that is attempting to make models match human cognitive capabilities. In December, the San Francisco-based OpenAI released the full version of its o1 model but kept its methods secret. 

DeepSeek’s R1 release sparked a frenzied debate in Silicon Valley about whether better resourced US AI companies, including Meta and Anthropic, can defend their technical edge.

Meanwhile, Liang has become a focal point of national pride at home. This week, he was the only AI leader selected to attend a publicised meeting of entrepreneurs with the country’s second-most powerful leader, Li Qiang. The entrepreneurs were told to “concentrate efforts to break through key core technologies.”

In 2021, Liang started buying thousands of Nvidia graphic processing units for his AI side project while running his quant trading fund High-Flyer. Industry insiders viewed it as the eccentric actions of a billionaire looking for a new hobby.

“When we first met him, he was this very nerdy guy with a terrible hairstyle talking about building a 10,000-chip cluster to train his own models. We didn’t take him seriously,” said one of Liang’s business partners. 

“He couldn’t articulate his vision other than saying: I want to build this, and it will be a game change. We thought this was only possible from giants like ByteDance and Alibaba,” the person added. 

Liang’s status as an outsider in the AI field was an unexpected source of strength. At High-Flyer, he built a fortune by using AI and algorithms to identify patterns that could affect stock prices. His team became adept at using Nvidia chips to make money trading stocks. In 2023, he launched DeepSeek, announcing his intention to develop human-level AI.

“Liang built an exceptional infrastructure team that really understands how the chips worked,” said one founder at a rival LLM company. “He took his best people with him from the hedge fund to DeepSeek.”

After Washington banned Nvidia from exporting its most powerful chips to China, local AI companies have been forced to find innovative ways to maximise the computing power of a limited number of onshore chips — a problem Liang’s team already knew how to solve.

“DeepSeek’s engineers know how to unlock the potential of these GPUs, even if they are not state of the art,” said one AI researcher close to the company. 

Industry insiders say DeepSeek’s singular focus on research makes it a dangerous competitor because it is willing to share its breakthroughs rather than protect them for commercial gains. DeepSeek has not raised money from outside funds or made significant moves to monetise its models.

“DeepSeek is run like the early days of DeepMind,” said one AI investor in Beijing. “It is purely focused on research and engineering.”

Liang, who is personally involved in DeepSeek’s research, uses proceeds from his hedge fund trading to pay top salaries for the best AI talent. Along with TikTok-owner ByteDance, DeepSeek is known for giving the highest remuneration available to AI engineers in China, with staff based in offices in Hangzhou and Beijing.

“DeepSeek’s offices feel like a university campus for serious researchers,” said the business partner. “The team believes in Liang’s vision: to show the world that the Chinese can be creative and build something from zero.”

DeepSeek and High-Flyer did not respond to a request for comment.

Liang has styled DeepSeek as a uniquely “local” company, staffed with PhDs from top Chinese schools, Peking, Tsinghua and Beihang universities rather than experts from US institutions.

In an interview with the domestic press last year, he said his core team “did not have people who returned from overseas. They are all local . . . We have to develop the top talent ourselves”. DeepSeek’s identity as a purely Chinese LLM company has won it plaudits at home. 

DeepSeek claimed it used just 2,048 Nvidia H800s and $5.6mn to train a model with 671bn parameters, a fraction of what OpenAI and Google spent to train comparably sized models.

Ritwik Gupta, AI policy researcher at the University of California, Berkeley, said DeepSeek’s recent model releases demonstrate that “there is no moat when it comes to AI capabilities”.

“The first person to train models has to expend lots of resources to get there,” he said. “But the second mover can get there cheaper and more quickly.”

Gupta added that China had a much larger talent pool of systems engineers than the US who understand how to get the best use of computing resources to train and run models more cheaply.

Industry insiders say that even though DeepSeek has shown impressive results with limited resources, it remains an open question whether it can continue to be competitive as the industry evolves.

Returns at High-Flyer, its big backer, lagged behind in 2024, which one person close to Liang blamed on the founder’s attention being mostly focused on DeepSeek.

Its US rivals are not standing still. They are building mega “clusters” of Nvidia’s next-generation Blackwell chips, creating the computing power that threatens to once again create a performance gap with Chinese rivals.

This week, OpenAI said it was creating a joint venture with Japan’s SoftBank, dubbed Stargate, with plans to spend at least $100bn on AI infrastructure in the US. Elon Musk’s xAI is massively expanding its Colossus supercomputer to contain more than 1mn GPUs to help train its Grok AI models.

“DeepSeek has one of the largest advanced computing clusters in China,” said Liang’s business partner. “They have enough capacity for now, but not much longer.” 

Supercomputers: the new superpower status symbol

US president’s endorsement shows data-centre dominion is now a national priority

Almost a hundred years ago, New York’s skyline became the scene of a bitter battle. The prize: displacing the Woolworth Building as the world’s tallest property. Such superlatives mattered to a growing, industrialised nation keen to flex its dominance. The winner was 40 Wall Street, swiftly pipped by the Chrysler Building, which in turn lost the title to the Empire State Building.

Today, there is a new superpower status symbol: data centres. The latest bid for digital supremacy is Stargate. This mammoth project led by artificial intelligence company OpenAI, and announced by US President Donald Trump, pledges to spend up to $500bn over four years to construct world-changing AI infrastructure — basically warehouses filled with clusters of microchips — on an unprecedented scale.

Trump’s endorsement, with a reassurance that no government money is involved, shows data-centre dominion is now a national priority. That is no surprise. New regimes love big gestures. Think of the “10 great buildings” with which Chairman Mao peppered Beijing in 1959, or The Line, a 106-mile sliver of a city under construction at the behest of Saudi Arabia’s Crown Prince Mohammed Bin-Salman.

Stargate is a trophy too. But it is also an arms race, because OpenAI is in effect building a weapon. Co-founder Sam Altman’s goal is to achieve “artificial general intelligence”, the kind that outflanks human beings. That this is now a White House-backed initiative reflects a superpower reality: whichever country reaches AGI first poses a significant threat, commercially and militarily, to the rest.

What does $500bn buy? First, assume the money can actually be raised; departed OpenAI co-founder and Tesla head Elon Musk is sceptical. The latest Blackwell chips from Nvidia could cost about $50,000 apiece, suggesting Altman’s budget could cover 10mn of them. The biggest AI supercomputer in the world today, Musk’s Colossus, has about 100,000. Size alone does not guarantee AGI success, but it is an important factor.

At its most mundane, OpenAI is following other big technology firms. So-called hyperscalers Amazon, Meta Platforms, Microsoft and Alphabet are collectively on course to invest $260bn in capital expenditure next year, according to Visible Alpha. Stargate’s plan of an average $125bn a year is big, but not substantially greater than theirs.

But Altman’s scheme is undeniably different. Alphabet, Microsoft, Meta and Amazon are building computing power customers can use, with a path to sizeable near-term revenue. OpenAI’s foray is more of a moonshot. Succeed, and vast profit awaits. Fail, and a surfeit of servers could push down prices and profit for the whole industry.

In the real world, meanwhile, traditional projections of power remain in fashion. China is planning 118 skyscrapers higher than London’s Shard, to add to the 100 or so it already has, according to the Council on Tall Buildings and Urban Habitat. Saudi Arabia, keen to be an AI hub, is also eyeing what will be the world’s tallest building, the Jeddah Tower. It’s just that in the AI age, there are more ways to reach for the heavens.

China’s deflating housing market offers an interesting parallel

Cryptocurrency proponents celebrated Donald Trump’s presidential victory, seeing in him a kindred spirit. The price of bitcoin, the original and most prominent cryptocurrency, has surged since his re-election in November. Under Trump, the crypto industry seems set to get what it wants — legitimacy provided by government oversight and light-touch regulation. It is a toxic mix for both the financial system and investors.

The switch in Trump’s views on cryptocurrencies — from sceptic to vocal advocate — does not mask the reality that nothing has changed in the fundamentals of this asset class, including its lack of intrinsic value. But the ethos of his administration lines up well with bitcoin’s libertarian aspects. 

Shortly before this week’s inauguration, both Trump and his wife Melania launched meme coins. It is remarkable for a government official, let alone the leader of the free world, to create and promote a purely speculative financial product from which they can profit. The financial grab left some crypto investors worrying that Trump might even undercut the mainstream acceptability of cryptocurrencies by reinforcing perceptions that they are all essentially scams.  

Trump has since righted the ship somewhat. He issued an executive order supportive of the crypto industry and directed the government apparatus to set up a regulatory framework to promote its activities.

The new president wants America to become the crypto capital of the planet, floating a proposal to create an official US bitcoin reserve. Establishment of such a reserve would give bitcoin an official imprimatur. But it makes little sense. Instead it would result in the government taking on the risks associated with bitcoin’s price volatility. Even if it generated paper profits, selling a significant share would cause bitcoin’s price to plunge, lowering the value of the rest of the government’s holdings.

Still, it is clear which way the winds are blowing. See the nominations of crypto enthusiasts Scott Bessent as Treasury secretary and Paul Atkins as head of the Securities and Exchange Commission. David Sacks, now the White House crypto tsar, will also be a forceful advocate for the industry.

True believers in decentralised finance built on bitcoin’s blockchain technology must be distraught. The notion that a government should be involved in the creation, dissemination, and usage of bitcoin contravenes the very principles under which it was created.

At least their digital wallets are getting fatter, which may soften the blow. 

Financial regulators will now undoubtedly ease restrictions on the issuance, use and trading of cryptocurrencies and crypto-related financial products. Crypto creators, promoters and exchanges will be able to operate more freely, while banks and investment managers will face fewer constraints in dealing with the assets. These changes will boost the broad adoption of crypto by both retail and institutional investors. 

The mainstreaming of crypto and the benign attitude of regulators will also spur closer connections between the industry and traditional financial institutions such as commercial banks and investment management firms. These connections will expose the conventional financial system to risk spillovers.

Meanwhile regulatory agencies and top administration officials are legitimising crypto assets, despite their highly speculative nature and the perils of exposing unsophisticated retail investors to their volatility.

Investors should be free to invest as they please, no matter how risky the asset class. But when a US president and his top officials speak favourably about an industry, investors could well let down their guard. History shows that such government boosterism often ends badly, with retail investors and taxpayers bearing the financial burden. 

China’s housing market bubble, which is deflating with painful consequences, provides an interesting parallel.

For many years the Chinese government relied on the property sector to drive its economy while touting it as a way for households to build their own wealth. State-owned banks provided loans to property developers and mortgages to households. Local governments, which rely on land sales as a key source of revenue, further stoked the property boom. Now that the property bubble is bursting, the burden is falling heavily on the lower-income households who locked up a large share of their savings in property or scraped together down payments that are now stuck with failed developers. 

The Chinese housing boom was at least related to real, physical assets. Bitcoin, by contrast, has no intrinsic value. Price volatility renders it an unviable medium of exchange and its value is based purely on scarcity, a characteristic that is arguably shared by gold.

There is nothing wrong with digital gold or with investors willing to roll the dice, unless the president and government officials are the ones who are hawking it. 

Trump and his government’s implicit endorsement of bitcoin and other cryptocurrencies means that the ultimate losers — if and when the bubble pops — will be the US taxpayers.

The ‘secret’ back door to the world’s biggest ski area

Orelle is so small it has only just got its first hotel — yet it offers high-speed access to the endless pistes of the Trois Vallées. Has its moment finally come?

Secret is one of travel writing’s most abused words — but humour me for a moment. Even if you’ve never skied, you’ve probably heard of Courchevel: the world’s most lavish ski resort, with five Palace hotels (only Paris has more). Perhaps you’ve heard of its neighbour in the next valley, Méribel, founded by a Scottish colonel and long a rowdy favourite of British skiers. Up and over the ridge to the west and you’re in Val Thorens, thriving in these less snow-sure times given its distinction as the highest ski resort in Europe.

Their three valleys, linked by ski lifts and pistes, make up another superlative: the world’s biggest ski area. Les Trois Vallées has 162 lifts, and 600km of pistes, all fully connected. It is so well known that it’s easy to be blasé about those remarkable statistics. Consider that the biggest ski area in the US, Park City, has just 40 lifts; a big hitter like the Espace Killy, comprising Val d’Isère and Tignes, can only offer 78, half the tally of the Trois Vallées.

The secret, then, or at least the place you’re much less likely to know, is Orelle, a community of 330 people over the next ridge — the famous ski area’s unadvertised fourth valley. Orelle, actually not one village but a collection of 10 hamlets scattered at the bottom of the steep-sided valley, hides in the shadow of its neighbours’ reputations, and is as modest as Courchevel is ostentatious. Where Courchevel boasts a dozen Michelin stars, the only restaurant in Francoz, Orelle’s main hamlet, is an ageing pizza van, parked beside the only bar (pointedly called “Les 4 Vallées”).

I arrived with my family just after Christmas, as Orelle marked a key milestone — the opening of its first hotel. We travelled by train, leaving London at 8am, connecting in Paris and Lyon, reaching the little town of Saint-Michel-de-Maurienne at 5.30pm, from where it was a 10-minute taxi to the hotel, the Hob Orelle. Though Francoz is at an altitude of just 890 metres above sea level, the streets were deep in snow, the dark forests rising up the valley sides smothered in white. The kids ran out to find a slope to slide down, resisting calls to put on jackets and gloves as they gleefully froze their hands making snowballs.

In the morning, we set out to try Orelle’s key asset, installed in 2021 — what must rank as one of the world’s great ski lifts. For skiing with kids, convenience is everything and from the door of the Hob’s boot room, we walked less than 100 metres on a snowy garden path to reach a little square, around which cluster the village’s one ski hire shop, the bar, pizza van, a tiny souvenir shop, the ticket office and the lift station.

The 10-person gondola climbs above the tree line, in about 12 minutes, to the Plan Bouchet, a mountain mid-station at 2,350 metres. There you can either get out, hop on a chairlift and start skiing, or continue, changing to the second stage of the new gondola which rises all the way to the Cime de Caron at 3,193 metres. It is a discombobulating change of perspective — to go in 20 minutes from sleepy village streets to a high alpine peak, looking across to the bulk of Mont Blanc, the rock towers of the Aiguilles d’Arves and the white expanse of the Vanoise glaciers.

That 2,300-metre vertical ascent dwarfs many of the most celebrated lifts on the planet — Jackson Hole’s “tram”, for example, graces T-shirts, caps, fridge magnets and even has a band named after it, but only manages 1,261 metres — and yet like everything about Orelle, this lift comes with little fanfare.

We headed over the far side of the ridge, skiing down to Val Thorens where, it being New Year’s week, the runs were busy but the snow, this high up, the best the kids had ever experienced. You could, if fast and determined, ski all the way to Courchevel. We contented ourselves with a few runs before a long lunch at a piste-side restaurant, the Chalet de la Marine, then a slightly anxious rush to catch the last lift up to the ridge in order to make it back to Orelle.

Miss it, and get stuck in the wrong valley, and you are looking at a two-hour, €450 taxi to get home. Catch it by the skin of your teeth, like we did, and you’re rewarded with being the last ones on the mountain, skiing alone down a long, blue run as the setting sun turns the snow pink — I removed my glove and gleefully froze my hand filming the kids as we went.

The reputation of the Orelle gondola and the Cime de Caron might soon start to grow. On Tuesday this coming week, a new summit station opens, with a restaurant and the highest wine bar in the Alps. Next week it hosts the Freeride World Tour, the first time the extreme skiing competition has come to the Trois Valleés.

There’s also a new covered walkway joining the Orelle gondola with the older, existing cable car up from the Val Thorens side. Just as with the new Matterhorn Alpine Crossing that connects Zermatt and Cervinia by cable car, part of the idea is to diversify away from the snow-dependent ski market and allow non-skiers and summer visitors to also make the journey over the mountains from Orelle to Val Thorens, stopping at the top for lunch and to take in the view. Unfortunately, the launch has been marred by an accident involving the Val Thorens–Cime de Caron cable car during pre-season maintenance in November; eight workers were injured and the lift will be out of action all winter.

Nevertheless, combined with the new hotel, plus a new beginner’s area and “magic carpet” lift at Plan Bouchet, the message, at least according to the tourist board’s bumpf, is clear: “Orelle is ready to play in the big leagues.”

The surprising thing is that, given its privileged access to some of the Alps’ highest slopes, the village has remained “the best kept secret in the Trois Vallées” for so long. The first lift opened in 1996 (skiers could get over to Val Thorens using a chairlift from Plan Bouchet to a lower point on the ridge). Yet accommodation options remained limited to some privately rented houses until, in 2007, a self-contained complex of cheap and cheerful holiday apartments with its own restaurant and small shop, Le Hameau des Eaux d’Orelle, opened just up the hill above Francoz. Even today, counting every apartment, private let and the new hotel, Orelle can muster only 1,300 tourist beds, not one per cent of the Trois Vallées’ total.

At the end of the day, almost everyone catches the gondola down from Plan Bouchet to the village but, though there’s no piste, it is possible to ski. I tried it on our first afternoon, following a snow-covered road that winds through the forest and ends up at Bonvillard, highest of Orelle’s 10 settlements. From there, basically lost, I alternated bits of skiing, skirting gardens and dodging brambles, with walking, skis on shoulder, though a succession of the rustic hamlets, past pretty stone chapels, a still-in-use communal bread oven, and along streets silent except for the drip of melting snow from the roofs.

If not a ghost town, Orelle is much quieter than it once was. In the 1920s there were more than 1,200 residents, the economy driven by an electrochemical factory down by the river. Its decline, before final closure in 1991, caused depopulation but the commune has been cautious about embracing tourism as an alternative, the mayor, in office since 1993, keen to preserve its traditional character.

So the opening of the Hob, built and owned by the Mairie (so in effect the village itself) and leased to an operating company, is a huge deal. The concept seems a good one: low cost, inclusive, fitting the village and much needed in a sport that risks becoming the preserve of the very wealthy. The name Hob is a contraction of hôtel and auberge (inn or hostel) — there are six double rooms, six family rooms and eight dormitories for up to six people. The dormitories — or dodos — have curtained bunks with their own light and storage, and the whole room can be “privatised” by a group of friends or a family. The prices are, for the Trois Vallées, startlingly low: doubles from €94, family rooms from €114, dorm beds from €32.

Initial impressions were good. The building is bright and colourful, the open-plan restaurant and bar has floor to ceiling windows looking out on a snowy terrace. The mattresses were comfortable, our family room spacious, with a balcony and big bathroom.

Sadly, there were, it has to be said, issues. Our visit was only a fortnight after opening and the staff had yet to find their feet. We asked for towels, finding we had only two for the four of us, and were promised more would be sent up. They were, eventually, but only after four more days of requests and empty assurances. Service was slow and random — order a glass of Mondeuse, receive a pint of lager — and I fell asleep and woke every day to the sound of the slamming door between kitchen and restaurant, exactly positioned one floor below my bed.

The main problem for us was at dinner: though Orelle and the Hob are in many ways ideally suited to families, there was no children’s menu, little or sometimes no choice of dishes on our half-board menu, and no chance of eating early. We sat down when the restaurant opened at 7pm but it could be 9.30pm before pudding arrived.

These are quibbles — and of course you can’t expect five-star service on a hostel budget. By now, I’d hope the staff have upped their game, and surely they’ve fixed the soft-close on the kitchen door?

On the last day there was fresh snow. The village was dark and under a thick layer of clouds but the lift took us up above them, so in sunshine we looked down over a cotton-wool sea. Previous days had taken us on forays to Val Thorens and beyond to Méribel, but today we stayed in Orelle, gradually working up to taking the kids to the highest point of the whole Trois Vallées at 3,230 metres. From there, we set out on their first proper off-piste run, a glorious long powderfield just beyond one of the pistes.

Then it was back down for a drink in the village square and one of the tourist office’s regular animations, tonight a performance by an elderly man with a flat cap and a barrel organ. For those seeking nightclubs, shops, glamour and society, Orelle is probably the worst resort in the Alps. For anyone looking to combine big skiing and little village life, it could be rather special.