WSJ : Cutting-Edge AI Was Supposed to Get Cheaper. It’s More Expensive Than Ever

Cutting-Edge AI Was Supposed to Get Cheaper. It’s More Expensive Than Ever.
With models doing more ‘thinking,’ the small companies that buy AI from the giants to create apps and services are feeling the pinch

As artificial intelligence got smarter, it was supposed to become too cheap to meter. It’s proving to be anything but.

Developers who buy AI by the barrel, for apps that do things like make software or analyze documents, are discovering their bills are higher than expected—and growing.

What’s driving up costs? The latest AI models are doing more “thinking,” especially when used for deep research, AI agents and coding. So while the price of a unit of AI, known as a token, continues to drop, the number of tokens needed to accomplish many tasks is skyrocketing.

It’s the opposite of what many analysts and experts predicted even a few months ago. That has set off a new debate in the tech world about who the AI winners and losers will be.

“The arms race for who can make the smartest thing has resulted in a race for who can make the most expensive thing,” says Theo Browne, chief executive of T3 Chat.

Browne should know. His service allows people to access dozens of different AI models in one place. He can calculate, across thousands of user queries, his relative costs for the various models.

Penny wise, pound-foolish
Remember, AI training and AI inference are different. Training those huge models continues to demand ever more costly processing, delivered by those AI supercomputers you’ve probably heard about. But getting answers out of existing models—inference—should be getting cheaper fast.

Sure enough, the cost of inference is going down by a factor of 10 every year, says Ben Cottier, a former AI engineer who is now a researcher at Epoch AI, a not-for-profit research organization that has received funding from OpenAI in the past.

Despite that drop in cost per token, what’s driving up costs for many AI applications is so-called reasoning. Many new forms of AI re-run queries to double-check their answers, fan out to the web to gather extra intel, even write their own little programs to calculate things, all before returning with an answer that can be as short as a sentence. And AI agents will carry out a lengthy series of actions based on user prompts, potentially taking minutes or even hours.

As a result, they deliver meaningfully better responses, but can spend a lot more tokens in the process. Also, when you give them a hard problem, they may just keep going until they get the answer, or fail trying.

Here are approximate amounts of tokens needed for tasks at different levels, based on a variety of sources:

• Basic chatbot Q&A: 50 to 500 tokens

• Short document summary: 200 to 6,000 tokens

• Basic code assistance: 500 to 2,000 tokens

• Writing complex code: 20,000 to 100,000+ tokens

• Legal document analysis: 75,000 to 250,000+ tokens

• Multi-step agent workflow: 100,000 to one million+ tokens

Hence the debate: If new AI systems that use orders of magnitude more tokens just to answer a single request are driving much of the spike in demand for AI infrastructure, who will ultimately foot the bill?

Ivan Zhao, chief executive officer of productivity software company Notion, says that two years ago, his business had margins of around 90%, typical of cloud-based software companies. Now, around 10 percentage points of that profit go to the AI companies that underpin Notion’s latest offerings.

The challenges are similar—but potentially more dire—for companies that use AI to write code for developers. These “vibecoding” startups, including Cursor and Replit, have recently adjusted their pricing. Some users of Cursor have, under the new plan, found themselves burning through a month’s worth of credits in just a few days. That’s led some to complain or switch to competitors.

And when Replit updated its pricing model with something it calls “effort-based pricing,” in which more complicated requests could cost more, the world’s complaint box, Reddit, filled up with posts by users declaring they were abandoning the vibecoding app.

Despite protests from a noisy minority of users, “we didn’t see any significant churn or slowdown in revenue after updating the pricing model,” says Replit CEO Amjad Masad. The company’s plan for enterprise customers can still command margins of 80% to 90%, he adds.

Some consolidation in the AI industry is inevitable. Hot markets eventually slow down, says Martin Casado, a general partner at venture-capital firm Andreessen Horowitz. But the fact that some AI startups are sacrificing profits in the short term to expand their customer bases isn’t evidence that they are at risk, he adds.

Casado sits on the boards of several of the AI startups now burning investor cash to rapidly expand, including Cursor. He says some of the companies he sits on the boards of are already pursuing healthy margins. For others, it makes sense to “just go for distribution,” he adds. Cursor didn’t respond to requests for comment.

One solution: dumber AI
The big companies creating cutting-edge AI models can, at least for now, afford to collectively spend more than $100 billion a year building out infrastructure to train and deliver AI. That includes well-funded startups OpenAI and Anthropic, as well as companies like Google and Meta that redirect profits in other lines of business to their AI ventures.

For all of that investment to pay off, businesses and individuals will eventually have to spend big on these AI-powered services and products. There is an alternative, says Browne: Consumers could just use cheaper, less-powerful models that require fewer resources.

For T3, his many-model AI chatbot, Browne is beginning to explore ways to encourage this behavior. Most consumers are using AI chatbots for things that don’t require the most resource-intensive models, and could be nudged toward “dumber” AIs, he says.

This fits the profile of the average ChatGPT user. OpenAI’s CFO said in October that three-quarters of the company’s revenue came from regular Joes and Janes paying $20 a month. That means just a quarter of the company’s revenue comes from businesses and startups paying to use its models in their own processes and products.

And the difference in price between good-enough AI and cutting-edge AI isn’t small.

The cheapest AI models, including OpenAI’s new GPT-5 Nano, now cost around 10 cents per million tokens. Compare that with OpenAI’s full-fledged GPT-5, which costs about $3.44 per million tokens, when using an industry-standard weighted average for usage patterns, says Cottier.

While rate limits and dumber AI could help some of these AI-using startups for a while, it puts them in a bind. Price hikes will drive customers away. And the really big players, which own their own monster models, can lose money while serving their customers directly. In late June, Google offered its own code-writing tool to developers, completely free of charge.

Which raises a thorny question about the state of the AI boom: How long can it last if the giants are competing with their own customers?

FT : Donald Trump moves to scrap $4.9bn in already allocated US foreign aid

Donald Trump moves to scrap $4.9bn in already allocated US foreign aid
US president makes new effort to assert power over Congress with bid to cut ‘woke’ funding

Donald Trump has ordered the cancellation of $4.9bn in foreign aid spending already approved by Congress, making use of an obscure legal mechanism in an effort to bypass lawmakers and assert his authority.

The president wrote to House Speaker Mike Johnson on Thursday night informing him of the rescission of funds earmarked for development assistance, peacekeeping and international organisations.

“The Trump administration is committed to getting America’s fiscal house in order by cutting government spending that is woke, weaponised and wasteful,” the White House said in a statement on Friday.

The move to claw back funds already approved by Congress is the latest step by the administration to test the limits on presidential authority. It comes as Trump looks to extend his influence over the Federal Reserve, taking the unprecedented step this week of attempting to sack a governor.

It marks the first time in 50 years that a president has sought to use executive powers under the Impoundment Control Act to deploy a so-called pocket rescission to circumvent lawmakers on Capitol Hill.

Congress can take up to 45 days to consider a rescission request from the president. But by issuing the directive with less than that amount of time until the end of the fiscal year on September 30, the White House is betting it can lock up the funds before they can be spent.

Trump’s directive drew swift criticism from Democrats and his fellow Republicans in Congress, who argued it was an illegal attempt to circumvent lawmakers’ constitutional control over spending. 

“Any effort to rescind appropriated funds without congressional approval is a clear violation of the law,” said Susan Collins, Republican chair of the Senate Committee on Appropriations on Friday. 

Patty Murray, the top Democrat on the committee, blasted what she said was an attempt by the White House “to do an end run around Congress”.

“No lawmaker should accept this absurd, illegal ploy to steal their constitutional power to determine how taxpayer dollars get spent,” she said. 

Trump’s directive would lock up $3.2bn in development assistance funding that the administration described as “endless handouts that allow recipient governments to eschew responsibilities to their own citizens”.

It would also slash $393mn in funding for US peacekeeping missions and $521mn in previously allocated contributions to international organisations, including the World Trade Organization, the International Labour Organization, the OECD and Unesco.

Secretary of state Marco Rubio said on Friday that the spending “violates the president’s America First priorities”.

“None of these programmes are in America’s interest, which is why the president is taking decisive action to put America and Americans first,” he wrote on X.

The non-partisan Government Accountability Office has previously said that pocket rescissions were against the law. Russell Vought, director of the White House Office of Management and Budget, considered deploying the measure during Trump’s first term, before deciding not to proceed.

The move is likely to inflame tensions between Republicans and Democrats in Congress, increasing the risk of a government shutdown if a funding package for the next fiscal year cannot be agreed by October 1.

“With this illegal power grab, Donald Trump and Russell Vought are driving us towards a government shutdown,” said Brendan Boyle, the top Democrat on the House Budget Committee.

FT : John Malone held talks with Rupert Murdoch over Warner Bros-Fox merger

John Malone held talks with Rupert Murdoch over Warner Bros-Fox merger
‘Cable cowboy’ reveals talks as he reorganises media empire and prepares to give most of his $10bn fortune to good causes

John Malone held talks last summer with Rupert Murdoch over merging Warner Bros Discovery with Fox, with the US “cable cowboy” highlighting the difficulty of one company owning both Warner’s CNN and Fox News as one reason the talks did not progress.

Malone, Warner’s chair emeritus and an adviser to the company, revealed in an interview with the Financial Times that he and Murdoch, his fellow chair emeritus at Fox, had “serious discussions about merging Fox into Time Warner, into Discovery” at the Sun Valley conference in July 2024.

The deal “probably would have happened if we thought that Fox News and CNN could live under one umbrella,” Malone said.  

In his autobiography, set to be published next week, Malone charts the formation of his sprawling business empire — from the creation of a cable giant at TCI and helping launch pay-TV in the US, to his stakes today in Warner, and in Formula 1 and Live Nation via Liberty Media.

After more than 50 years at the vanguard of dealmaking, the 84-year-old is still exploring deals for various parts of his empire, the proceeds of which he said will underpin plans to give most of his $10bn wealth to philanthropic causes after he and his wife pass away.

With little sign that his renowned “urge to merge” — as he puts it in his book — is fading, Malone signalled an openness to a future deal involving Formula 1, owned by Liberty Media.

As for Fox Broadcasting, separate to the news division, Malone said “a deal would potentially give Warner’s US sports position a huge strengthening” if Murdoch, whom he counts as a friend, were open to selling.

Other people close to the discussion last summer said Warner chief executive David Zaslav and News Corp chair Lachlan Murdoch were also present. Fox insiders downplayed the significance, saying a possible deal was raised but not pursued.

Malone supports a proposed split of Warner, in which he owns a 0.7 per cent stake, into two publicly traded companies: a Warner Bros focused on its studios, film, and streaming, and a Discovery Global holding its traditional cable networks.

The deal would unwind a merger that has led to significant shareholder losses. The combined company’s shares have halved, wiping about $30bn off its market capitalisation, as Americans ditch cable TV in favour of streaming.

“Because it was a deal between two heavily regulated companies, there was no real opportunity to do due diligence,” Malone said.

“When we acquired Warner . . . we didn’t see the deterioration. We knew it was a transformation, but we didn’t know that we would start in [such a] hole . . . have I taken a personal financial beating up til now? I can afford it, I guess.”

With HBO’s best shows being brought back to its own streaming service, HBO Max, Malone sees a brighter future for the separated companies, saying Zaslav will have a “clean shot” with a streaming company boasting the “best studio on the planet [and] the biggest library on the planet”.

“Am I glad that we did it? Maybe in the end, it will work out,” Malone said of the merger. “Who knows, big tech might step up and take us out.”

Malone, who retains a seat at the top table of the rapidly changing media industry, is predicting a fresh wave of mergers. As well as big tech swooping on the traditional US media groups he had a hand in building, Malone expects consolidation among “sub-scale” streaming services, especially those owned by legacy media groups.

While he believes “exclusive, high-quality entertainment content” gives legacy media a valuable edge, he said big tech will increasingly “outbid everybody for sports . . . we know that it moves customers [and] it sells advertising”.

The scale of the threat from big tech means consolidation between legacy media was also conceivable, Malone added, pointing to the “possibility of consolidation of a Disney and a Warner, or a Disney and a Paramount or an NBC Universal with a Warner”.

A restructuring of Liberty Media — in which Malone is chair and the shareholder with the most voting rights — means its 30 per cent stake in Live Nation will be separated into a newly formed company this autumn.

The paperwork for the split has been filed, Malone said, which would leave Liberty Media holding motorsports Formula 1, MotoGP, and various smaller private assets. This follows several other spin-offs, including the Atlanta Braves baseball team and SiriusXM, a satellite radio operator.

The simplified structure would make it easier for Malone to pursue a deal involving Formula 1, for which Liberty has already received approaches.

TKO, the group behind Ultimate Fighting Championship and World Wrestling Entertainment that is run by Ari Emanuel, has been speculated as a potential partner. Malone said that “obviously Ari would love to have that discussion” and added that a combination of this sort could create “a lot of synergies”.

“Everything will be spun off, and then let’s see what new things we can create,” Malone said, adding that GCI — an Alaska-based telecoms company in which he will retain a shareholding — “perhaps could be a vehicle to do something with”.

Malone was instrumental in creating modern broadband infrastructure. He rolled up scores of networks and pushed for upgrades in the 1980s so that they could carry hundreds of cable TV channels, and then upgraded them again for high-speed internet, which ultimately paved the way for Netflix and others to upend the media industry.

Now, he is reorganising some broadband assets by merging holding company Liberty Broadband, in which he is a shareholder, with Charter Communications. The deal is scheduled to complete when Charter finalises its own $34.5bn merger with Cox Communications.

The listed telecoms holding group, Liberty Global, which owns stakes in operators including Virgin Media O2 in the UK, is also slimming down and spinning off assets.

Despite his close bonds with contemporaries such as Murdoch, Ted Turner and Barry Diller, Malone describes in his book, Born to be Wired, how he sometimes found it difficult forming personal relationships over the course of his swashbuckling 50-year career.

Parts of the book read like a “thank you” letter to those who advised and mentored him, notably Bob Magness, who brought him in to rescue TCI, his US cable company, in the 1970s. Malone eventually sold TCI to AT&T for $48bn in 1999. He also seeks to credit the managers under him who had a “helluva” job making his ideas work: “I could come up with ideas faster than anybody could execute.”

While Malone misses the “instant gratification” of making decisions as a chief executive, he says he is happy not to have to deal with an “army of lawyers and regulators” anymore.

Malone is watching the Murdoch family’s succession battle closely. Rupert’s attempt to change his family trust to give control to his eldest son, Lachlan, was blocked by a court in Nevada in December.

He says it would be “very disruptive” if Rupert’s other son, James, ended up controlling Fox News. “I know Rupert would hate the thought,” he said, pointing to a threat to the network’s centre-right political position under the younger Murdoch son.

Malone will not be handing his remaining businesses to his children. Instead, he plans to give most of his $10bn wealth, roughly half of which is made up of real estate, to charitable causes, such as medical research, education and preserving some of his vast land holdings as public open space.

However, as he rearranges his sprawling empire, some assets he will not countenance selling. One is the Atlanta Braves, in which he says he would never sell his stake as long as his friend and Braves fan Ted Turner is alive. “It’s not about money on those kind of things. This is about relationships and history.”

Another asset with protected status is Malone’s Castlemartin Estate in County Kildare, Ireland, which includes a stud farm and restored medieval church.

When Mark Zuckerberg, who built his social media empire off the back of the high-speed broadband networks Malone pushed to create, offered to buy the estate earlier this year, Malone gave him the cold shoulder.

“I told him no, sorry Mark,” smiles Malone. “He said: “I hear it’s the prettiest place in Ireland”. And I said: “Maybe in the world, but it ain’t for sale.”

Barrons : Luxury blingflation creates opportunity for cheaper challengers

Luxury blingflation creates opportunity for cheaper challengers
Second-tier brands and retailers are starting to stretch their tendrils upwards

The best things in life may be priceless. But over the past few years some of them have become extortionate. As the luxury industry has created its own kind of high-end inflation, price rises have left some customers behind. That has created an opportunity for upwardly mobile brands, even mass-market challengers.

After a series of sharp price increases, luxury’s unaffordability has turned from a feature to a bug. Sales declined 1 per cent in the last quarter, according to Bernstein, with those at industry bellwether LVMH falling by 4 per cent, and at troubled Gucci-parent Kering by 15 per cent.

But second-tier brands seem to be stepping in. Burberry, for instance, is retrenching after a failed attempt to go even more upmarket — and investors have rewarded it with a share price bounce of more than 90 per cent in a year. Those reluctant to spend more than £6,000 to bag themselves a Louis Vuitton Capucines can opt for niche leather goods producers such as Polène and DeMellier.

It is not just new luxury brands that are stepping in. Retailers such as Zara owner Inditex, with a broader customer base, can also stretch their tendrils upwards.


Zara has been investing in larger, glitzier stores, offering some higher-priced products, including a £700 leopard print jacket from last year’s collaboration with model Kate Moss. Its average selling price per item is already higher than its usual competitors — $34 compared with $26 at fast-fashion retailer H&M and $14 at Shein, according to McKinsey analysis.

Unusually for a mass-market retailer, Zara also manages to project an aura of scarcity. Since it makes a large proportion of its own goods, it can rapidly revise its offering. A Bernstein analysis found that over four weeks it changed about 40 per cent of its online collection, twice as much as H&M. In practice, if consumers see a product they like, there’s extra reason to snap it up sharpish, which gives Zara extra pricing power.

None of this suggests that Zara is going to go all-out luxe. But it does add an extra layer to its equity story. Already, Inditex grows faster than rival H&M, and has over twice the ebitda margin. That comes with a market capitalisation equivalent to 22 times forward earnings, a discount of less than 10 per cent to LVMH. Upward mobility provides a fetching trim.

Barrons : Aptiv Is Breaking Up—and Investors Should Buy the Stock

Aptiv Is Breaking Up—and Investors Should Buy the Stock
Shares have sunk because it is perceived as an auto-parts supplier, but the planned Aptiv spinoffs have growth potential.

Auto-parts supplier Aptiv is splitting up into two companies, one selling safety and software solutions, the other supplying electrical distribution systems.

Breaking up will allow both companies to better allocate capital to growth opportunities. It also could unlock shareholder value as investors assign higher multiples to non-automotive industrial businesses that demonstrate growth and margin improvement.

The company’s recent performance has been strong, with a feared slowdown in automotive production not materializing. That has Wall Street analysts increasing 2025 and 2026 earnings estimates.

If you break it up, they will come. That’s the bet auto-parts supplier Aptiv is making—and investors should consider going along for the ride.

Aptiv hasn’t had an easy time of it in recent years. The company began life as Delphi Automotive, the auto-parts business that was spun out of General Motors in 1999, before declaring bankruptcy in 2005, crushed by a combination of high labor costs, high debt, and an underfunded pension. After a restructuring, Delphi spun off Aptiv in 2017, with the former keeping the powertrain technologies such as fuel injectors, valve actuators, and sensors, and the latter taking the higher-growth businesses such as vehicle electrification and safety.

Aptiv was a great stock through 2021, benefiting from investor excitement for electric vehicles. But when the EV boom went bust, so did Aptiv’s stock—it dropped 55% from a late-2021 peak of almost $180 to a recent $79.88.

The problem isn’t earnings. Aptiv is expected to earn $7.48 this year, according to FactSet, up from $2.61 in 2021, a compound annual growth rate of 30%.

Instead, the problem is one of perception. At its peak, shares of Aptiv fetched about 36 times forward earnings, but once investors decided it was just an auto-part supplier again, its valuation slumped to 10 times. To change the perception problem, Aptiv plans to split into two companies: one that will focus on its slower-growing electrical-distribution-systems business, the other on faster-growing safety and software. If all goes well, the spin, which is scheduled to wrap up during the first quarter of 2026, should allow both parts of the company to succeed—and investors to profit.

“Aptiv’s bold spinoff isn’t just a corporate shuffle, it’s a strategic shift,” says spinoff specialist and Edge Research founder Jim Osman.

It isn’t hard to see why management is pursuing a breakup. The EDS business had annual sales in 2024 of $8.3 billion, with profit margins of 9.5% on earnings before interest, taxes, depreciation, and amortization, or Ebitda. Safety & software, meanwhile, had sales of $12.2 billion with Ebitda margins of 18.8%—and the potential to reach customers outside the auto business. “The [Delphi] spin was about making Aptiv a better automotive supplier,” writes Baird analyst Luke Junk. “This is more orienting the company toward higher return opportunities overall.”

The new Aptiv, or RemainCo in Wall Street jargon, will sell sensors, software, and electronics that enable higher levels of autonomous functions and electronic communication for cars, planes, and other machines. It is also looking to expand beyond autos with strategic acquisitions—a transformation that has already begun. Aptiv bought communications software provider Wind River in 2022.

If all goes well, it will begin to look more like industrial companies such as TE Connectivity and Amphenol, which fetch 16 times earnings and 22 times estimated 2025 Ebitda, respectively. Aptiv isn’t either of those companies, of course; TE and Amphenol benefit from high growth in artificial-intelligence data center spending. Post-spin, about 25% of Aptiv’s business will be nonautomotive, estimates Junk. The numbers for TE and Amphenol are about 55% and 80%, respectively. But Aptiv could trade like Sensata, a maker of sensors and electrical components for the car, aerospace, construction, and manufacturing industries. It trades for about nine times estimated Ebitda, one and a half points above Aptiv’s 7.5. Sensata gets 45% of its business from nonautomotive companies.

There’s room for improvement in the EDS business, as well. For starters, the outlook for automotive suppliers is better than feared, says RBC Capital Markets analyst Tom Narayan, with a tariff-induced slowdown not visible in the numbers. That means stronger production from auto makers and more parts for auto suppliers. The independent EDS business should be able to boost margins from high-single digits to low-double digits using a combination of automation and changes to its manufacturing base.

Finally, there is consolidation. EDS will “have buyers,” adds Narayan, with other suppliers looking for ways to grow while lowering costs. He has an Outperform rating and $92 price target on Aptiv stock, up 15% from recent levels.


But there’s reason to be more optimistic. Getting an automotive-like multiple on the electric business and a Sensata-like valuation on the RemainCo means Aptiv shares could hit $100, up about 25% from recent levels. The biggest risk is that investors continue to view both spinoffs as car companies, which would cause the stock to fall back toward $70, or about 15% below Tuesday’s close.

But a well-managed spin has the potential to create value—especially where none was seen. Fuel-systems provider Phinia was supposed to face challenges after its 2023 spin from BorgWarner amid the transition to battery-powered personal transportation, says Osman. Shares have roughly doubled over the past two years, while BorgWarner stock has gained about 30%. When General Electric split up in 2024, GE Aerospace was supposed to be the hotter stock, and while it has doubled since the spin, shares of GE Vernova, which were thought to have a tougher road, quadrupled.

Ultimately, the success or failure of the spin will depend on management’s execution and the fickleness of the markets. But with Aptiv’s history, we like our chances.

The Technical View
Aptiv stock’s rally started during the week ended May 6, when shares jumped 10%, breaking a series of lower highs dating back to the fourth quarter of 2024. Aptiv has shown excellent follow through after bouncing off a double bottom and breaking through resistance at $72.53. Now, the 50-week moving average is starting to slope higher, suggesting that this turnaround is real. There’s a good chance the stock will hit $100 by the end of the year. If the stock breaks below $69, consider selling. — Doug Busch

Barrons : Nvidia Is Still Poised for a Blockbuster Year. And the Stock Remains C

Investors shouldn’t get distracted by the noise about China. Nvidia is positioned to benefit from its most important and largest product ramp in its history. The signs are there.

Late Wednesday, the chip maker delivered another impressive, eye-watering report, with unprecedented growth rates for a company its size. Revenue for the July second fiscal quarter was up 56% year over year to $46.7 billion, ahead of expectations. Nvidia’s outlook was solid. For the current quarter ending in October, the company provided a revenue forecast range with a midpoint of $54 billion, which was above analysts’ consensus of $53.4 billion.

But Nvidia shares initially fell modestly after the results, which likely has more to do with some profit-taking after the stock rose about 35% over the prior three months. Uncertainty over its business in China may have been another factor.

To be sure, China remains Nvidia’s biggest problem. While some analysts had hoped for a quick rebound in sales of Nvidia’s China-specific H20 graphics processing unit, or GPU, it hasn’t happened. In fact, on Wednesday, Nvidia said there were no H20 sales from China-based customers in the second quarter at all. The company also hasn’t assumed any H20 shipments to China in its guidance for the current quarter.

“We continue to work through geopolitical issues,” Nvidia CFO Colette Kress said on the earnings call.

It’s worth noting that the company’s fiscal third-quarter outlook still beat Wall Street’s current estimates, even without the benefit of H20 sales.

Nvidia’s Chinese sales have become a casualty of the continuing back and forth between the U.S. and China. In April, the U.S. government moved to effectively ban sales of Nvidia’s H20 chips to China by tightening export licensing requirements. In July, the U.S. reversed course and said it would approve those H20 licenses. But China reportedly grew frustrated after disparaging remarks from U.S. Commerce Secretary Howard Lutnick and told companies not to order H20s due to security concerns.

Last week, Nvidia CEO Jensen Huang said he was surprised by the latest development given that China’s government had asked Nvidia to secure H20 export licenses from the U.S. government. “Hopefully, the response that we’ve given to the Chinese government will be sufficient. We are in discussions with them,” he said. Huang has denied the existence of any kind of security backdoor built into Nvidia’s H20 chips.

Despite all the noise, there’s a good chance things will get resolved given China had originally asked Nvidia for H20 licenses and with the high demand for Nvidia GPUs in the Asian country. The latest developments are likely more theatrics than a substantive change in views.

The H20 was designed to comply with restrictions that prohibit China’s access to the most advanced AI hardware. Despite those limitations, Chinese companies still prefer the H20 to run AI inference workloads reliably, as Nvidia’s software is more stable than anything created in China.

If the two economic superpowers come to terms, the benefit to Nvidia would be significant. Back in May, Nvidia said it could have sold $8 billion of H20 chips in the second quarter, absent regulatory hurdles.

Huang also said sales of a newer Blackwell chip for China is a “real possibility” in the future, but that it was too soon to know whether that would be approved.

This area of the company’s business is unpredictable and could change at any moment. The bottom line is that the China revenue for Nvidia is all upside from here.

Ultimately, Nvidia investors should focus on the company’s main AI chips, which will drive positive results no matter what happens with China. The company is still poised to benefit from its most important product releases in its history: the 72 GPU rack servers, called the GB200 NVL72 and the GB300 NVL72.

The NVL72 systems incorporate 72 GPUs, linked together inside one server rack, up from eight GPUs in the previous version. It’s an unprecedented density of computing power that is crucial to train the latest AI models.

Huang has confirmed that the new racks are in full production and that output is ramping up after initial manufacturing obstacles. Each rack costs several million dollars.

“It delivers an exceptional generational leap,” Huang said on the earnings call. Nvidia’s NVL72 “rack-scale computing is revolutionary, arriving just in time as reasoning AI models drive order of magnitude increases in training and inference performance requirements.”

There is evidence the rack server ramp is under way. Nvidia’s networking revenue segment surged 46% quarter on quarter in the July quarter. That’s a good sign, as NVL72 systems require more networking hardware per rack.

Over the course of the latest earnings season, we’ve learned that Big Tech firms continue to increase their capital expenditure budgets while emphasizing that demand for AI is outstripping capacity. Nvidia’s NVL72 rack servers are the product needed by start-ups and enterprises as they clamor for more AI computing resources.

With the production of Nvidia’s AI rack servers just ramping up, several quarters of strong financial performance should be on the way. While Nvidia stock has doubled from April’s lows, it isn’t overpriced. Shares trade at 32 times forward earnings—almost cheap, given the 50% earnings growth forecast by Wall Street analysts for the current fiscal year.

In this column, I have repeatedly said that Nvidia shareholders are best served by taking a long-term view on the stock and not selling in response to volatility unless there is a significant change in Nvidia’s long-term outlook. That advice has paid off, with Nvidia touting a world-topping $4.4 trillion market value.

Nvidia’s underlying fundamentals remain as strong as ever. Hold on.

Barrons : AT&T Can Provide a Dividend—and a Higher Stock Price, Too

AT&T Can Provide a Dividend—and a Higher Stock Price, Too

There’s little that’s more frightening to an income investor than a company deciding to spend billions of dollars on a deal—dollars that could be going to the dividend. But in the case of AT&T, which just spent $23 billion to acquire spectrum from EchoStar, the acquisition could make the dividend safer.

Spectrum—the radio frequencies your wireless calls travel over—is the finite resource that every mobile company needs. Just as you can’t have two radio stations in the same city broadcasting on the same frequency, companies like AT&T, Verizon Communications, and T-Mobile US need spectrum for their customers’ calls and data. It’s all managed by the U.S. government via auctions and licenses.

Tuesday’s announcement was a big deal. AT&T said it had an agreement to purchase about 50 MHz of low and mid-band spectrum from EchoStar, adding to the company’s “solid” portfolio. Wireless carriers control hundreds of MHz of spectrum, and AT&T listed $127 billion worth of license assets in its 2024 annual report. The telecom giant touted the deal as covering “virtually every market across the U.S. and positioning AT&T to maintain long-term leadership in advanced connectivity across 5G and fiber” in a press release announcing the deal.

The deal was enormous for EchoStar, which was at risk of losing the spectrum it wasn’t using, and the satellite and wireless communications company saw its stock jump 70% on the deal. Some EchoStar bonds jumped 25%—they had been yielding north of 50% before the agreement and now yield about 14%, with the cash helping to pay off some of its $26 billion in debt.

The acquisition should also be good news for the three top telecoms, notes Citi analyst Michael Rollins. By buying the spectrum, AT&T keeps it out of the hands of a potential new, fourth competitor. “We view this as a positive for the longer-term trading multiples for AT&T, Verizon, and T-Mobile,” he says.

The large outlay doesn’t appear to affect AT&T’s financials in a way that threatens the dividend. The added debt looks manageable, and after the company announced the deal, it maintained its financial guidance—it still expects to earn $2.02 a share in 2025—and its share repurchase plan, which involves buying $20 billion in stock from 2025 through 2027. Business was strong to begin with—the company added 401,000 wireless subscribers and 243,000 fiber internet subscribers in the second quarter—and dividend coverage looks solid, with AT&T returning roughly half of its annual free cash flow to investors.

Citi’s Rollins already had a 90-day short-term catalyst call on the stock, which he rates Buy. He likes the purchase, too. AT&T will use the excess spectrum to beef up its wireless services to the home. “A very logical path for AT&T,” adds Rollins, who notes that the company wants to bundle home and wireless services to boost growth.

He has a $32 price target for AT&T stock, up 10% from Wednesday’s close of $29.06. Combining that with AT&T’s 3.9% dividend yield would result in a 14% total return over the next 12 months.

Rollins is also a fan of Verizon, which he also rates Buy. A combination of low valuation and sustainable growth underpin his views. He also likes Verizon’s pending acquisition of Frontier Communications, which will beef up the company’s fiber internet offerings. Verizon trades for about nine times estimated 2026 earnings, while AT&T shares fetch 13 times.

While his $48 target for the stock suggests a gain of just 8.4 %, adding Verizon’s 6.2% dividend yield implies a total return of almost 15%, similar to AT&T. Verizon is the eighth highest-yielding stock in the S&P 500. AT&T is the 60th, with its yield sliding from 5.6% a year ago due to the stock’s outsize strength.

It was only a little more than two years ago, in the summer of 2023 that AT&T was trading below $15 a share on concerns that copper wires sheathed in lead might need replacing to mitigate health concerns. Since then, AT&T has returned about 120%, with a fifth of that return coming from dividends. Verizon stock has returned about 50%, just ahead of the S&P 500 over the same span, with 40% of the total return coming from dividends.

The AT&T spectrum buy should just be another catalyst to boost the share performance of both stocks in the coming year. Income investors should enjoy their dividends—and their capital gains, too.

Barrons : Palantir CEO Alex Karp Sells More than $60 Million in Stock

Palantir CEO Alex Karp Sells More than $60 Million in Stock

Palantir Technologies CEO Alex Karp unloaded more than 400,000 shares in the data-analytics company last week, according to a filing with the Securities and Exchange Commission.

A Form 4 filed with the agency on Friday showed that Karp sold 409,072 shares of common stock, at prices ranging from $142.46 to $157.56 a share, on Wednesday and Thursday. The sales were executed through automatic transactions to cover tax-withholding obligations following the vesting of restricted stock.

Following the transactions, Karp owned about 6.43 million shares of Palantir, which were valued at more than $1 billion as of Monday afternoon.

For most of the year, Palantir stock had seemed untouchable, continuing to rise regardless of concerns about its lofty valuation and high volatility. Shares have more than doubled this year, but fell 1% to $157.17 on Monday. The iShares Expanded Tech-Software Sector exchange-traded fund, which counts Palantir among its major holdings, was off 0.7%.

Palantir’s rise has been challenged as of late. Shares of the artificial-intelligence heavyweight fell for six consecutive trading sessions through Wednesday, marking their longest losing streak since April 2024. The decline came after strong second-quarter results caused the stock to surge earlier this month.

Other big names in tech, notably AI plays such as Nvidia, have come under pressure as investors question how long the current red-hot enthusiasm will last. OpenAI CEO Sam Altman’s remarks that investors were “overexcited” about the technology, and an MIT report casting doubt on the revenue-generating potential of generative AI, appeared to be compounding the downturn.

Given all that, scrutiny is mounting ahead of Nvidia earnings later this week. The chip maker is seen as a bellwether for the broader AI and semiconductor industries. Commentary from CEO Jensen Huang is expected to shed light on the state of AI demand.

Barrons : Hedge Funds’ New Secret Weapon

Hedge Funds’ New Secret Weapon
New companies are collecting any kind of business and consumer economic activity they can get their hands on and selling it to investors.

If you took a road trip in the past year, chances are good that you drove by one of Ryan Joyce’s cameras. They’re everywhere: hanging from highway streetlights, protruding from the walls of warehouses, and sitting on the roofs of car dealerships.

Joyce is the CEO of Genlogs, a start-up that collects data on the movements of America’s semi-trailer trucks. The company operates nearly a thousand cameras that monitor the nation’s roads, snapping photos of 18-wheelers as they drive past. Joyce says Genlogs can tell its clients where every delivery truck in the country is located on a given day.

Most of Genlogs’ customers are logistics and freight brokerage companies, which use the start-up’s data to reduce theft and optimize delivery routes. But recently, Joyce says he has been receiving lots of interest from a new set of potential clients: investors. Hedge funds want Genlogs’ data because it can tell them the volume of truck traffic in the U.S., a useful clue to the health of the American economy. Genlog cameras could even help investors bet on specific publicly traded stocks, like Walmart or trucking giant J.B. Hunt Transport Services, Joyce says.

Welcome to the world of “alternative data.” It’s a fast-growing market where companies collect information on everything from credit-card transactions to the number of travelers passing through security at airports. Investors then buy the data, hoping to glean insights beyond those provided by traditional company filings and government reports.

The use of alternative data by investors is nothing new. For decades, various funds have incorporated unusual data sources—including measures of electricity consumption and counts of the number of cars parked outside retail stores—into their research processes.

What has changed since the pandemic, insiders say, is the volume, sophistication, and variety of the data being collected and used. Thanks to recent advances in computer vision technology, language processing models, and web scraping tools, even small firms like Genlogs can harvest far larger and more complex data sets than ever before.

It isn’t just trucks. Data vendors now tell investors how many customers are clicking on a particular product on an e-commerce website. They track how many pedestrians are entering a specific bricks-and-mortar store. They monitor blockchain activity, Reddit and X feeds, and satellite imagery.

Not surprisingly, as the quality of alternative data has improved, a rapidly growing share of the investment community has come to embrace it. “The hedge fund part of our business is growing 100% year over year,” says Craig Fuller, CEO of FreightWaves, a company that tracks 85% of global cargo volume as it moves around the world.

Investment firms will spend at least $3.3 billion on alternative data sets in 2025, according to Neudata, a consultancy. Spending has grown by 21% a year since 2020, and could climb to more than $8.6 billion by 2030 (estimates go as high as $39.9 billion). A poll released in February by law firm Lowenstein Sandler found that two-thirds of institutional investors now use alt data, double the share compared with just two years ago.

Richard Lai, Bloomberg’s global head of alt data, says investors’ appetite for data has dramatically exceeded expectations: The number of firms using Bloomberg’s alt data analytics platform is more than four times the company’s estimate of the market size in 2023, he says.

Prices for alt data vary widely. More than a quarter of data sets tracked by Neudata have a sticker price of $25,000 or less, while only 3% of data sets sell for $500,000 or more. But being on that upper end pays huge dividends. The industry’s biggest players can rake in tens of millions of dollars in annual sales from a single data set, according to Neudata.

Stock market investors aren’t the only ones gobbling up data: private-equity firms use alt data to assess potential acquisitions; insurance companies use it to set premiums; big-box retailers use it to pick new store locations. But sales to investors are a growing share of the pie, according to Daryl Smith, Neudata’s head of research.

An Arms Race
Fund managers are tight-lipped when discussing how alt data informs their investment strategies. But one thing isn’t a secret: Data sources once considered unconventional are now part of the day-to-day investment process at most big asset managers. And securing new data sets, sometimes at significant cost, is one of the key ways that elite funds stay ahead of the pack.

“There has been a technology arms race within the quant world to find and engage with data,” says David Easthope, a senior analyst at Crisil, a research firm focusing on the financial industry. “Investors who are seeking superior returns are always looking for an edge.”

To make their trades, hedge funds’ quant models rely on clean, high-frequency data sets that go back for years or even decades. Until recently, the number of data sets that met quant funds’ strict criteria was relatively limited. But as data sets mature, quant investors say the number of data sets digestible by their models is growing.

Artificial intelligence is opening up new frontiers, as well. It’s allowing companies to analyze new types of data, like Genlogs’ camera footage or pdf files of government documents. It’s also helping vendors improve the quality of their existing databases.

“The growth opportunity isn’t just in the data itself,” says Carolyn James, director of strategic sales at ImportGenius. “It is the ability, with AI models that really do deep learning, to actually link data to other sources.” Thanks to AI, ImportGenius says it can provide its clients with container-level specificity of every product entering and exiting the U.S. by boat.

“AI is definitely a tailwind,” says Topher Haddad, CEO of start-up Albedo Space. Albedo plans to launch satellites into very low earth orbit, where they will capture images of the earth’s surface at a resolution previously reserved for military spy satellites and drones. As computer vision improves, he says, “You can actually pull out and extract more intelligence, in an accurate way, from any given image resolution.”

AI is also making it easier for non-quantitative investors to use alternative data sources. Analysts can use AI models to analyze the text of social-media posts or the audio of investor calls, without spending millions on developing in-house models. They can also buy data directly from a growing number of AI-powered data vendors, or use information published on Bloomberg’s alt data analytics platform.

Thanks to improved accessibility, many data businesses say they’re now seeing strong interest from the “long tail” of investors: the small hedge funds, family offices, wealth advisors, and retail traders who previously lacked the resources to make use of large volumes of alternative data.

Company filings and government economic reports simply aren’t good enough to stay ahead anymore, says Jay Hatfield, who runs investment firm Infrastructure Capital Management. For one, they’re too slow: Earnings and government statistics are only published quarterly, while alt data can deliver insights in real time. For another, public data is often unreliable: Reports published by the U.S. Bureau of Labor Statistics have been getting less accurate for years, even before President Donald Trump fired the head of the bureau after a particularly weak July jobs report.

To supplement traditional information sources, Hatfield’s firm has turned to alt data. Analysts use information from Placer.ai, a start-up that harvests anonymized geolocation data from tens of millions of mobile apps, to see how many customers are walking into stores. They also use data from Similarweb, a company that tracks how many users are downloading apps and visiting specific webpages.

At Bloomberg, part of Richard Lai’s role is to identify and add new data sets to Bloomberg’s alt data platform. The task is “never-ending,” Lai says. As information becomes widely adopted by investors, its ability to deliver market-beating returns can dissipate, causing hedge funds to look for even better information. “This is something that will never stop,” he says. “Data is going to get better, it’s going to get faster, it’s going to get more accurate.”

A Better World?
Cesar Orosco, head of alpha equity investments within Vanguard’s quantitative equity group, says the prevalence of alt data is changing the art of investing. Decades ago, information was scarce, and researchers spent much of their time trying to find reliable data. Now, information is abundant—and the real edge comes from interpreting, rather than obtaining, data.

Many data providers and investment professionals argue that alt data is making financial markets more efficient. Orosco isn’t so sure. As investors wade through oceans of data, much of which sends noisy or contradictory signals, they can easily draw incorrect interpretations. Speculative bubbles and meme stocks endure, he notes. “Are the markets more or less efficient? I think the jury is still out,” he says.

Privacy remains a big concern. Every data provider interviewed by Barron’s stressed that the data they analyzed and sold was anonymous at every stage, and couldn’t be used to identify individuals. Data platforms such as Similarweb and Placer.ai say the information they collect from partners is already anonymized when they receive it. To protect privacy, Genlogs says its cameras instantly delete images of private vehicles.

“Trying to balance the privacy risk with the benefits is going to be a major question over the next five years or so, as AI capabilities become more robust,” says Patrick Schmid, chief insurance officer at the Insurance Information Institute, a research group for insurers. “Trying to mitigate personally identifiable information associated with the bits is going to be critical.”

Easthope says the biggest privacy concern comes from investors collecting information on executives and employees. Data about where CEOs are traveling—obtained, for example, by tracking their private jets—can provide useful investment information, while posing a risk to privacy and safety.

“There’s a voracious appetite for this data,” Easthope says. “And where the line is, I frankly don’t know. I think that line will get pushed.”

Barrons : Citadel’s Ken Griffin on Markets, the Fed, and Building His Firm for t

Citadel’s Ken Griffin on Markets, the Fed, and Building His Firm for the Next Century
His business handles one out of every four stock trades. Our deep dive into the Wall Street firm of the future.

The most successful companies on Wall Street have been built by those with relentless ambition—and a strong wind at their back. Decades of deregulation helped growth-minded CEOs turn firms such as Morgan Stanley and Bank of America into behemoths. More recently, financial entrepreneurs have leveraged booming private markets to create the likes of Blackstone in private equity, Bridgewater Associates in hedge funds, and Andreessen Horowitz in venture capital.

Now a somewhat stealthier trend, fueled by the explosive growth of technology in finance, is behind a new wave of digital-first powerhouses such as Interactive Brokers Group
IBKR, Susquehanna International, Jane Street, and especially Citadel—a burgeoning Wall Street empire controlled by billionaire Ken Griffin.

Citadel is a two-headed business beast consisting of Citadel LLC, a $68 billion hedge fund operation best known for its top-performing multistrategy flagship Wellington, and Citadel Securities, a sprawling market maker that facilitates and engages in the trading of stocks, derivatives, and increasingly bonds. Citadel—Griffin chose the name because it “denotes a place of strength and protection,” he says—is defining the prototype of a next-gen, bulge-bracket firm, except that Citadel isn’t that bulgy, with just a fraction of the employees and overhead of the traditional Wall Street firm.

As of now, the boss is happy. “It’s incredibly satisfying to run one of the world’s most successful hedge funds and to witness the transformative impact of Citadel Securities on the capital markets around the world,” Griffin says. “Yes, I’m proud of that.”

Taken separately, first, Citadel the hedge fund is a huge deal even as a stand-alone. Wellington (not related to Vanguard Wellington Fund or Wellington Management) has been a superstar, producing annual average returns of 19.2%, net of fees, since inception in 1990—nearly two times the market. A million dollars invested in the fund back then is worth $452 million today.

Even though Griffin says he spends the majority of his time working on the hedge fund’s investment portfolio, arguably the hotter ticket is Citadel Securities. CitSec, as it is known, is a complex, sometimes controversial, ever-evolving endeavor—which, Pac-Man-like, is on a seemingly inexorable march across capital markets, gobbling up market share in options, equities, Treasuries, and corporate bonds, and now expanding to Europe and Asia. Example: CitSec recently bought Morgan Stanley’s U.S. equity-option market-maker business, leaving no major banks in the market-making business.

Citadel Securities says it now trades 25% of all U.S. equities, including 35% of retail flows plus 45 billion options quotes a day, and is a top-three trader in U.S. Treasuries and swaps. In total, it executes $652 billion in notional trades a day. The goal is clear. “Building the capital-markets firm of this generation is a vision that is increasingly becoming a reality,” says Citadel Securities CEO Peng (pronounced “pung”) Zhao, a Ph.D. in statistics from the University of California, Berkeley. (Griffin serves as nonexecutive chairman of CitSec.)

“This is a very different company than most people understand,” says Alfred Lin, a partner at venture-capital firm Sequoia Capital who sits on Citadel Securities’ board (and whose brother happens to be the head of global fixed income and macro at Citadel the hedge fund). “Citadel is taking math and their distribution and technology power to price risk using techniques not traditionally used on Wall Street.”

While technology has been absolutely critical to Citadel’s success, so too has Griffin’s strategy. “You don’t want to go up against the Wall Street incumbents,” he says. “Instead, you want to understand where the market is heading and position yourself there before the incumbents arrive.” The incremental rise of electronic trading allowed Citadel to move step-by-step ahead of names like Goldman Sachs Group and Morgan Stanley
MS in a number of trading businesses.

Griffin also makes all kinds of headlines. A Republican who has given hundreds of millions mostly to GOP candidates and causes, Griffin has praised and criticized Trump. Just this past week, Griffin told Barron’s: “I hope President Donald Trump appreciates that while he can score political points by attacking Jay Powell, ultimately the independence of the Fed is of the utmost importance to the American and global economy.”

Meanwhile, Griffin, 56, has been adding to his personal portfolio at a stunning pace, including buying over a billion dollars worth of real estate (in New York City, the Hamptons, London, St. Tropez, Hawaii, and multiple properties in South Florida, including $400 million in Palm Beach, according to The Wall Street Journal). His art collection, estimated to be worth close to $1 billion, includes works by Picasso, Van Gogh, and Warhol, and he has bought priceless historical American documents, including rare copies of the U.S. Constitution, the Bill of Rights, the Emancipation Proclamation, and the 13th Amendment (signed by Abraham Lincoln). Then there’s the $45 million stegosaurus.

Griffin’s philanthropy, now directed through an entity called Griffin Catalyst, exceeds $2 billion and includes funding vaccine development during the pandemic and helping create Operation Warp Speed. On the hipper side, he recently donated $2.255 million to Mr. Beast’s water philanthropy after the YouTube superstar called him out on the Today show. “It seems very important to Ken that the world knows how wealthy he is,” says a business associate. “A lot of other people go to great lengths the other way.”

The totality of Griffin’s world is dizzying. The billions upon billions of hedge fund investments, market-making activities, and personal assets are markers of a man with great aptitude and perhaps even greater ambition.

Along the way, Griffin and his companies have encountered friction, false starts, falling outs, fines, and failure. The hedge fund dropped 55%—losing $9 billion of clients’ money—during the global financial crisis, and was at death’s door. CitSec tried and failed to get into investment banking. The market maker has had a number of run-ins with regulators. Yet now, after more than 30 years in the business, the tumblers have been falling into place.

A decade ago, Griffin was worth $6.1 billion, according to Bloomberg; today his net worth has ballooned to $48.3 billion, making him the world’s 31st-richest person, by dint of hedge fund payouts and an 85% stake in that business, plus his 80% piece of the market maker—the latter being valued at $22 billion three years ago after Griffin sold a 5% stake to Sequoia. Never mind the real estate and art.

Wealth accumulation by longstanding lieutenants such as Griffin’s right-hand man and chief operating officer of the hedge fund, Gerald Beeson, who joined in 1993 as an intern out of DePaul University, and Zhao, who joined in 2006, is likely significant as well. Not that making a career at Citadel is easy. Both the hedge fund and CitSec look to hire what Griffin calls “great athletes,” and are known to be workplaces where elbows are sharp and fools aren’t suffered gladly—or really at all.

“I thought we ran hard at Goldman,” Pablo Salame, the co-chief investment officer of the hedge fund, who came from Goldman Sachs in 2019, said to a colleague. “And then I showed up here, and I realized there’s a whole different gear.” (Griffin is co-chief investment officer and CEO of the hedge fund.) “I would tell my team, you’re playing for Real Madrid,” says a former Citadel executive. “You don’t get to keep your spot on the field if you’re not producing.”

“Ken would call you on a Sunday night at 11 p.m. and he might be screaming and yelling at you, but he was working,” says another former employee. “He demanded 150%, and nothing else in your life should matter.”

“Ken’s not yelling at you, he’s yelling with you,” a person close to Citadel joked.

“Citadel Securities and hedge fund Citadel are highly performant places to work,” says Matt Culek, CitSec’s COO, who has worked at the firm for 13 years. “This is a place to come if you have a lot of confidence in your ability and want to challenge yourself.” It’s true that Citadel hardly has an issue attracting aspiring masters of the universe. “We had over 100,000 applicants for our summer intern program, and our acceptance rate was 0.4%,” says Culek. “People are falling over themselves to demonstrate they deserve to be here just for the summer, and then some subset of them get to come back full time.”

Griffin was STEM-smart and an achiever from his earliest days, growing up in Boca Raton, Fla., where he was president of his high school’s math club. He graduated from Harvard University with an economics degree in three years. Chapters of Griffin’s origin story have become the stuff of Wall Street legend, like convincing Harvard to let him install a satellite dish on the roof of his dorm so he could trade convertible bonds.

After Harvard, Griffin moved to Chicago, where Frank Meyer, a pioneering hedge fund investor, mentored him and helped seed Griffin’s hedge fund. “The best advice I’ve ever gotten was from Frank Meyer, who in the early days of my career really pushed me to think big,” says Griffin. (Meyer also pressed Griffin to build a multistrategy platform instead of a single-strategy fund.) Griffin stayed in Chicago until three years ago, when he says he became disgruntled with the crime and business environment and relocated his companies to Miami—a triumphant return of South Florida’s most successful financier native son.

As he built his businesses, Griffin leaned heavily into his math and computer science background, applying evolving technology along the way. “The term HFT [high-frequency trading] was coined here,” says Zhao, who adds that a physicist working at the firm came up with the name for a strategy he created. Though Citadel Securities used to be known mostly as a high-frequency shop, Zhao says that today HFT is “part of the core competency you need to have as a modern-day market maker, but it’s becoming less and less a business model.”

After the hedge fund’s meltdown during the financial crisis, Griffin rallied the troops, and by 2012 he had led the fund back to its high-water mark. Performance has been strong. Over the past five years, Wellington has had annualized net returns of 23.6%, according to the Global Investment Report, versus a 14.5% annualized total return for the S&P 500. The hedge fund has returned more than $25 billion to investors over the past eight years—a mixed blessing for the funds’ limited partners, who then have the problem of where to reinvest that money.

As a multistrat hedge fund, Citadel employs a podlike structure, where individuals or groups of money managers pursue their own thesis strategies but with strong central controls. The firm isn’t wedded to a single asset class or investment strategy and style, which allows Salame and Griffin and the management committee to shift the emphasis of the fund. Instead of a typical 2%-of-assets management fee, Citadel “passes through” its direct costs, such as employee salaries, which can amount to up to 7%. And it charges 20% of upside as well. Citadel is unique in that it is publicly rated investment grade by S&P Global Ratings, which, as of May 1, rated it BBB-, one notch above junk, and “stable.”

“We want to generate returns that are diversifiable and where there is liquidity,” says Salame, an Ecuadorean who’s an aficionado of Japanese whiskeys. “You have to constantly review whether your capital allocation and portfolio shape is optimal today, even though it was optimal a day ago.” As a former Goldman colleague says, Salame was “demanding and one of the smartest people I’ve ever worked with.”

The fact that Griffin owns the two Citadels makes for a singular and potentially conflicted structure. Executives on both sides of the house, as they describe it, stress that there is zero communication between the two, and that no trading information from the market-maker business passes over to the hedge fund. The two firms do occupy the same office buildings in some cities, such as New York, where both are slated to move into a proposed 1,600-foot-tall Park Avenue tower, sharing some back office functions and even a logo. Furthermore, Citadel Securities does, in fact, use trading information from its market making business, not in the hedge fund, but in Citadel Securities itself.

Like other market makers (also known as nonbank liquidity providers), such as Jane Street and Virtu Financial, CitSec receives orders from retail and institutional clients. It processes customers’ trades by executing them mostly on an exchange or sometimes on an alternative trading system such as a dark pool. Citadel can make money off the spread between the bid and offer, but more important, it can use those securities in myriad strategies including hedging, mitigating risk, or trading for its own account. It’s a hideously complex, esoteric business, but a very profitable one. In the first quarter of this year, CitSec had $3.4 billion in net trading revenue, up some 45% from the same period last year, while net income climbed 70% to $1.7 billion, according to Bloomberg.

“Ken tells an amazing story about when he developed conviction on a trade, which he was right on,” says Billy Hult, CEO of Tradeweb Markets, a bond trading platform of which Citadel is a customer. “When Ken went to take the trade off, he realized how much money the market maker was making by unwinding his trade, and he said, ‘Wow, there are two great businesses to be in—the business of investing, and the business of market making—and they’re both really lucrative.’ ”

CitSec’s market-making business also entails the somewhat controversial business of payment for order flow, much scrutinized during the GameStop imbroglio in 2021, where Citadel pays retail brokerage firms like Robinhood Markets and Charles Schwab to process their trading orders. In the first quarter of 2025 alone, Citadel paid brokers $388 million—much of that for options contracts, according to Global Trading —more than any other market maker. Citadel says it gets the most orders because it has the best execution, and that the money it pays to the brokers is essentially rebating back part of the spread. Those payments allow brokers to offer rock-bottom-priced or even free trading to investors—which begets more trading, which benefits Citadel.

Trading data, which is nonproprietary, can be fed into Citadel’s massive computing stack, where 260 Ph.D.s have access to some 100 petabytes of data—more than the entire collection of the Library of Congress. Citadel data scientists and traders can use that information to identify inefficiencies between securities and arbitrage opportunities to obviate risk and make trades. Some are longer-term trades, says Zhao, while an HFT strategy, in which time is measured in 500 millionths of a second, makes thousands of small trades. Profiting from the spread on customers’ trades, mitigating risk, and trading for its own account are closely intertwined. Revenue from those activities “can’t be separated in a rational way,” according to Zhao.

Market conditions recently have been ideal for Citadel Securities. Citibank says that year-to-date average daily volumes through July are 17.2 billion shares, a record high and up over 2.5 times from 2017. So far in 2025, stocks below $5 have represented 31.7% of total volumes, reflecting the latest meme-stock craze. And zero-day-to-expiration options are booming as well. All that trading means more volume for Citadel.

“The equity market is robust in pricing, but also reflects that the U.S. government is running a very large fiscal deficit, which is incredibly stimulative for the economy and healthy for corporate profits,” Griffin says. “That, combined with a weaker dollar, has created tailwinds. There is one salient issue in the equity market now: How much of the hype of AI will translate into the reality of a more productive, more prosperous future?”

Going forward, Citadel has limitless possibilities, says Sequoia’s Lin. “They can invest. They can trade and make a market. They can use math to help trade, absorb, and price risk better or help clients do that. They can provide advanced analytics and technology. They can provide intelligence to banks and to customers.”

“We’re still in early innings, even for some of our core products,” says Jim Esposito, the president of Citadel Securities, who came over from Goldman last year. “In equities, we’re skewed to the U.S.; in fixed income, we’re just getting started. I’m spending time with Ken and Peng thinking about geographic and product expansion. We’re everybody’s favorite strategic partner to chat with, from technology platforms like Google to private-equity firms like Apollo [Global Management].”

Espo, as he is known, is amping up client-facing efforts at this tech-first company as it pushes more into bond trading and European operations, which can require high-touch relationships with senior clients and government officials.

“We’re seeing higher levels of uncertainty in the world,” says CitSec COO Culek. “That’s going to increase the need for price discovery, which will lead to more market activity.” Culek also points to strength in retail trading, event-based derivatives (trading on elections, economic data, or policy decisions), and crypto. “We’re bullish on the growth of that set of opportunities,” he says. “We see revenue others are already making that we can capture over time, and we can envision entirely new revenue in the future, some of which we are going to proactively create ourselves.”

Griffin continues to be thoroughly engaged. “Some people who have Ken’s success at his age start doing things like getting married in Europe,” says a top executive at a major financial institution. “But Ken is uniquely still driven. That’s a really important factor in why they succeeded and why they’ll continue to succeed.”

While one generation of Wall Street entrepreneurial titans, including the likes of Bridgewater’s Ray Dalio and Blackstone’s Steve Schwarzman, is winding down, another, with Griffin at the fore, is still gearing up. “Where we are today represents a phenomenal starting point for where these businesses will go over the next century,” says Griffin.

With ambition like that, and a digital wind in his sails, what sounds audacious might just be plausible.