TheChek News Flow

As CEO of Databricks, Ali Ghodsi has performed a series of ‘strategic surgeries’ to make his company one of the fastest-growing startups in Silicon Valley

When he was in college in Sweden, Ali Ghodsi read an article about a chief executive who was hired to save a struggling tech company. It reminded him of the surgeries his parents, both doctors, performed while he was growing up.

“I loved the fact that you could think about large corporations as patients, and you could perform surgery on them to make them super healthy and successful,” he said.

Ghodsi is now CEO of a software company called Databricks, which has quietly become one of the fastest-growing startups in Silicon Valley. The 11-year-old company is now valued at $62 billion after securing $10 billion from investors including Andreessen Horowitz and Thrive Capital. The new funding, announced Tuesday, is among the largest in the history of venture capital.

Ghodsi has gotten Databricks to this point thanks to the “strategic surgeries” he said he’s performed to keep the company healthy. After taking the reins in 2016, he grew sales by charging more for software that the company had originally given away for free. Two years ago, when investors demanded efficiency, he slowed hiring and had his engineers build an AI bot called R2-D2 to boost productivity.

Today, data scientists at some of America’s largest companies use Databricks’ software to analyze the large volumes of information they collect—a tool that’s become even more valuable with the rise of artificial intelligence. Walgreens, for example, uses Databricks to help forecast inventory for filling prescriptions, while Rivian uses it to improve the battery life of its electric trucks.

Immigrant drive

Ghodsi was born in Tehran in 1978 on the eve of the Iranian Revolution. Five years later, after his parents became political targets, the family fled to Stockholm, where they initially lived in student dormitories.

He began coding on his home computer in fourth grade and started a business where he charged his classmates to fix their broken computers.

Ghodsi enrolled at a local university called Mid-Sweden University with plans to study computer science. On his first day, his roommate, a business student, joked that he would one day become Ghodsi’s boss. Ghodsi enrolled in a class his roommate was taking and added a second major in business.

While in college, Ghodsi started a consulting firm that built a scheduling system for a local city government, he said.

In 2009, after completing a Ph.D. in computer science, Ghodsi moved to the U.S. as a visiting scholar at the University of California, Berkeley. A year later, while working at AMPLab, a data-analytics lab affiliated with the university, he and six other researchers developed a new piece of code called Spark.

Spark became an instant hit among data scientists for its ability to analyze messy data sets at record speed—at one point setting a world record for sorting through 100 terabytes of data in 23 minutes, beating the old record by over 40 minutes. The seven researchers decided to build a business around it called Databricks.

They asked Ben Horowitz, co-founder of Andreessen Horowitz, for $200,000 to help launch the business. Horowitz encouraged them to think bigger, writing an $11 million check on behalf of his venture-capital firm.

At first, Ghodsi was unsure whether he wanted to become a full-time tech executive. He worked part time at Databricks, where the first CEO was fellow Berkeley computer scientist Ion Stoica. In the fall of 2015, Ghodsi applied to become a computer-science professor at the university.

Then came a call from Horowitz. He asked Ghodsi to become CEO.

In January 2016, Ghodsi succeeded Stoica, who returned to academia. At the time, Databricks charged for a more user-friendly version of Spark but was struggling to find large customers who would pay for it. Many were still downloading the software for free off the internet.

The challenge became clear when a potential customer asked Ghodsi and another co-founder for a selfie before a meeting. When the pair later asked whether he would be willing to pay $10,000 for the software, the customer scoffed. “Why would we ever pay $10,000?” he said. “I’m just going to get it for free.”

Under Ghodsi, Databricks added new features to Spark that were only available to the startup’s paying customers. He then hired hundreds of salespeople to sell it to companies, targeting business giants like Capital One and JPMorgan.

Ghodsi also replaced the executive team and began sharing his board presentations with the startup’s 250 employees—a decision inspired by a business case study he had read in college describing how factory workers performed better when they shared the same goals as their bosses.

In 2016, Horowitz brokered an introduction for Ghodsi to Microsoft CEO Satya Nadella, hoping to strike a deal to integrate Databricks with the tech giant’s Azure cloud-computing platform. Before they were set to announce the deal, some executives inside Microsoft got cold feet about the quality of the young startup’s software. Ghodsi recalled flying to Microsoft’s Redmond, Wash., headquarters and fielding their questions for two hours until they got on board.

Microsoft signed the deal in 2017 and committed to generate $100 million in sales for Databricks.

A few years later, Ghodsi led a crucial push to expand the company’s business. Spark focused on analyzing enormous data sets like server logs, but he wanted to combine that with more organized data stored in tables like sales information. That was a much larger market featuring rising software companies like Bozeman, Mont.-based Snowflake.

Customers said the resulting product saved them hundreds of hours they’d previously spent working across different data formats to gain insights from them. Sales grew even as many companies cut back on software spending during the postpandemic tech downturn.

“You can pry it out of my dead cold fingers,” Dan DeMeyere, chief product and technology officer at the online clothing reseller ThredUp, said of his engineers’ loyalty to Databricks.

In 2022, Ghodsi realized Databricks had grown too bloated and needed another surgery. Nearly a decade after its founding, his startup was doubling the number of new hires each year and losing money, but his investors were now demanding efficiency.

At an annual company off-site held at the Venetian Hotel in Las Vegas in February 2023, Ghodsi told his roughly 5,000 employees that he wanted to make the startup profitable within three years. He had his Chief Financial Officer Dave Conte share a giant chart tracking progress toward this goal. The chart became so widely referenced among employees that it even got its own nickname—the Conte Curve.

Ghodsi began personally approving all new hires while offshoring jobs to countries with cheaper labor such as Costa Rica and India. The company holiday party was cut, as was its annual companywide off-site. And the startup’s engineers began building AI-powered bots—including one named after R2-D2 from Star Wars—to automate internal tasks.

Unlike most big tech companies trying to economize around the same time, though, Databricks didn’t lay off workers.

Since unveiling the Conte Curve, Databricks more than doubled revenue to $2.6 billion while halving its negative operating margin, according to financial documents viewed by The Wall Street Journal.

At the same time, Ghodsi went on a dealmaking spree to try to position Databricks for a new spurt of growth.

In the summer of 2023, he called Naveen Rao after learning that Rao was fundraising for his two-year-old AI startup, MosaicML. Ghodsi offered to buy his company for $1.3 billion instead. Databricks customers now use Mosaic to build AI models that can predict spending habits or determine which promotions to offer their clients.

In the spring, he caught wind that Tabular, a popular data-management startup that was catching on with businesses, was in talks to sell itself to Snowflake. Ghodsi won the deal by offering the startup roughly $2 billion—more than triple what Snowflake proposed, according to people familiar with the deal.

Soon after, Ghodsi began debating whether or not to take Databricks public by the summer of 2025. Instead, he chose to raise another private funding round. The company intends to use some of the new cash to buy out employee shares and the taxes associated with them.

Minister says Eutelsat OneWeb network falling short and Elon Musk’s Starlink ‘not an option’ given his Chinese links

Taiwan’s government is in talks with Amazon’s Project Kuiper subsidiary about co-operating on satellite-based communications, as Taipei broadens its efforts to make its mobile phone and internet infrastructure less vulnerable to a potential Chinese attack.

Wu Cheng-wen, technology minister, told foreign media on Tuesday that the OneWeb network of French satellite operator Eutelsat, which partnered with Taiwan’s state-backed Chunghwa Telecom last year, was falling short in providing sufficient capacity for the country’s needs.

Taipei was now exploring additional international collaboration, Wu said.

“We found that their bandwidth is too small for real applications,” the minister said about OneWeb. “So far as I know, the company is in financial problems at this moment, so they have a [delay] in developing the second generation satellites.

“There are other companies in the Western world, including some from Europe and from North America and Canada, but Amazon Kuiper is the most mature in their development stage so far. So we are discussing with them at this moment if it is possible that we will have a collaboration in the future,” he added.

Asked about Wu’s comments, Eutelsat OneWeb said it was “absolutely not in financial difficulties” and that there was “no delay” in developing the next generation of its constellation, which was progressing as planned.

OneWeb and Kuiper are among the largest of numerous nascent providers of communications services offering or planning networks of satellites in low Earth orbit — the region of space up to about 2,000km above the planet’s surface — in competition to Elon Musk’s Starlink.

Kuiper plans to begin launching its constellation soon, with services from low Earth orbit expected to begin next year.

Eutelsat OneWeb said it was the only LEO network “ready to deliver services in Taiwan” and that it had “the full backing of our shareholders” for its new constellation.

After Russia launched its full invasion of Ukraine in early 2022, Taiwan cranked up efforts to become more resilient against what it fears is a growing risk of Chinese aggression. China claims Taiwan as part of its territory and threatens to annex it if Taipei refuses to submit to its control indefinitely.

The role that Starlink has played in upholding Ukraine’s wartime communications drove the Taiwanese government to pursue LEO communications as a back-up, in case the undersea cables that support its mobile telephony and internet systems are cut.

Taiwanese government officials said separately that Starlink itself was not an option for them, adding Musk’s company would not agree to a joint venture in which Chunghwa Telecom or another Taiwanese entity held at least 50 per cent, as Taipei demands.

Officials said Musk’s extensive business interests in China and past comments on Taiwan’s political status and future were an additional hurdle.

Musk has repeatedly suggested he takes China’s side in the sovereignty dispute. Last year, he said Taiwan was an integral part of China as claimed by Beijing, adding that it was “arbitrarily” outside Beijing’s control because the US military was blocking unification.

Two years ago, he suggested the conflict be resolved by handing at least partial control of Taiwan to China.

Apart from Chunghwa Telecom’s tie-up with OneWeb, Taiwan’s space agency is also working to develop a national provider with its own low Earth orbit satellite constellation.

Wu said the government aimed to pick a site in south-eastern Taiwan for launching its own rockets, with launches expected to start within five years.

Eutelsat is highly leveraged, with net debt in September totalling four times earnings before interest, tax, depreciation and amortisation. Earlier this year, the group was forced to cut revenue guidance for 2024 and 2025 because of uncertainty over the outlook for OneWeb.

Amazon and SpaceX did not immediately respond to requests for comment.

The Takeaway

•China is moving slowly on Synopsys-Ansys antitrust review

•It is reopening Nvidia-Mellanox review in response to U.S. export restrictions

•Numerous other deals could be subject to Chinese review

The Chinese antitrust regulator’s decision a week ago to investigate Nvidia for alleged violations of antitrust rules in relation to its 2020 purchase of Mellanox Technologies is likely to be the first of a series of such reviews by the Chinese government, according to three people familiar with Chinese regulators’ thinking.

The Nvidia-Mellanox investigation was triggered by the U.S. Commerce Department’s decision to tighten restrictions on the export of American chipmaking technology to China, two of the people said. China’s State Council, the equivalent of a cabinet, and its Ministry of Foreign Affairs signed off on using antitrust inquiries as a way of striking back against Washington’s increasing curbs on Beijing’s technological rise, one of the people said.

Already, the State Administration for Market Regulation, which oversees anti-monopoly regulation and other issues, is dragging out the review of chip design software maker Synopsys’ $35 billion planned acquisition of Pennsylvania engineering software developer Ansys. SAMR is also reexamining two past mergers that involve Coherent, formerly known as II-VI, a manufacturer of transceivers that enable fast data flow in chips and artificial intelligence devices, two of the people said.

Whether more investigations are officially announced depends on how tensions evolve with the incoming Trump administration, the person added. President-elect Trump has threatened to impose huge tariffs against China when he takes office.

China’s State Council, SAMR, and the Ministry of Foreign Affairs did not reply to requests for comment. Coherent didn’t immediately comment.

Synopsys said in a statement: “The regulatory approval process is proceeding as expected in relevant jurisdictions, including China.”

Since the second half of 2019, China has conditionally approved eight mergers where American tech companies were the acquirers (and often also the ones acquired), involving a total deal value of more than $150 billion. The biggest was Broadcom’s $86 billion stock-and-cash takeover of VMware in 2023.

Two other deals came from II-VI, a Pennsylvania based maker of advanced engineering materials used in defense, aerospace, and other industries: It acquired optical communications maker Finisar for $2.9 billion in 2019 and purchased Coherent, which sells laser-based tech, for $7.1 billion in 2022. II-VI renamed itself Coherent after the latter deal. Both acquisitions received conditional approval in China, which was the last market to greenlight the transactions. Lately, however, SAMR has been reviewing both deals for potential breaches of the approval conditions, the two people said.

These deals combine two U.S. companies. But regulators in other countries including China have antitrust oversight if the companies have significant businesses in their territories, because the mergers could potentially have an anticompetitive impact on their local markets.

Under China’s anti-monopoly law, SAMR has the power to reopen any past cases with conditional approvals on the grounds that they may have breached those conditions. If found in breach, companies could be fined up to 10% of their annual global revenue or be forced to unwind the merger in China.

“This highlights how China’s regulatory power can extend beyond its borders,” said Angela Zhang, a professor of law at the University of Southern California and an expert on China antitrust issues. “Ultimately, China is using its market access as a tool to exert influence over these transactions.”

Nvidia Probe

Nvidia bought Mellanox, an Israeli company that developed computer networking technology, for $7.1 billion in April 2020. China was the final country to approve the deal.

Mellanox’s InfiniBand technology is used in Nvidia’s Hopper and Blackwell series to allow faster data transfer and quick communication among thousands of chips. It has become one of the ways Nvidia has maintained its lead in the AI chip sector.

SAMR’s approval came with various conditions, including a requirement that Nvidia not bundle any Mellanox equipment with the sale of Nvidia GPUs in China, not discriminate against Chinese customers, and ensure that both companies’ products remain interoperable with those of others. Those conditions are effective for six years.

Since 2022, Chinese companies have complained to SAMR that Nvidia has discriminated against them, in particular relating to access of H20 chips, the only graphics processing units for AI training that Nvidia is allowed to sell to China under U.S. export controls, according to one of the people familiar with Chinese regulators’ thinking.

Chinese regulators also considered as discriminatory the fact that U.S. regulations have prohibited Nvidia from selling its most advanced GPUs to China, the person added. In February, Nvidia disclosed in a U.S. securities filing that “regulators in China have inquired about sales and efforts to supply the China market and our fulfillment of the commitments we entered at the close of our Mellanox acquisition.”

But SAMR didn’t announce the probe until Dec. 9, one week after the Commerce Department released annual updates to its export control regulations, because it wanted to appraise the impact of the latest U.S. rules before taking action, the person added.

As part of the investigation, the antitrust regulator can visit Nvidia’s offices in China, interview its employees, suppliers, and customers, and review and make copies of Nvidia’s books and other internal documents, according to China’s anti-monopoly law.

In a statement, Nvidia said: “We work hard to provide the best products we can in every region and honor our commitments everywhere we do business. We are happy to answer any questions regulators may have about our business.” The company declined to comment on the alleged discrimination against Chinese customers or whether Chinese regulators had inspected its offices.

Dragged-Out Approval

The Commerce Department’s latest export restrictions also prompted Beijing to use the Synopsys-Ansys deal as another card to play.

Sunnyvale, Calif.–based Synopsys is a global leader in developing electronic design automation software that is essential for designing integrated circuits, without which both designers and manufacturers would struggle to create and understand complex chip architectures. Often referred to as the backbone of chip design, EDA software has also been subject to rounds of U.S. export restrictions against China since 2022.

Ansys, meanwhile, makes software used to simulate the physical aspects, like mechanical, electrical and heating, in the design of semiconductors, automobiles, and airplanes.

In May, Synopsys flagged in a securities filing that SAMR had asked it to submit the deal for approval, even though the transaction was below the review threshold set out in China’s anti-monopoly law. In addition to the specific thresholds, the regulator also has the discretion to review any mergers it deems as potentially anticompetitive.

On Dec. 6, Synopsys submitted an application to SAMR. On Dec. 9, the same day the agency announced its Nvidia probe, SAMR told Synopsys it would hold off reviewing the case on the grounds that the application materials were insufficient, according to two people familiar with the regulator’s thinking.

According to Chinese regulation, the agency must make an initial decision on whether to approve or further examine the transaction within 30 days of accepting the application, and it can pause the clock anytime pending more materials. If the agency proceeds with an in-depth review, it could take up to another 90 days with an extension of up to 60 days to reach a final decision.

The back and forth in Synopsys’ application indicates that China could drag out the review process of the $35 billion deal involving two American software makers, just as it delayed Broadcom’s takeover of WMware and Nvidia’s purchase of Mellanox. In both cases, the companies had to extend deal completion timelines while they waited for China’s approval, which was the last to arrive of all the necessary regulatory sign-offs.

Synopsys said it remains confident in a positive resolution of the ongoing regulatory review processes and expects the transaction to close in the first half of 2025.

A gift from survivors of the Titanic to the captain of the R.M.S. Carpathia, the watch was acquired at auction for $1.97 million by the American brand.

In November, an 18k gold pocket watch, a gift from survivors of the Titanic to the captain of the R.M.S. Carpathia, which rescued them, was acquired at auction in the U.K. for $1.97 million, marking the highest-valued piece of Titanic memorabilia ever sold. Initial reports cited an anonymous buyer — until now. The buyer actually was Tiffany & Co., marking a return home for the timepiece.

“For this opportunity to arrive,” said Christopher Young, vice president of creative visual merchandising, events and The Tiffany Archives at Tiffany & Co., noting that the piece had come up at auction a decade ago. “I never would have thought that this would come up again. Never in a million years. There are pieces where you think, ‘Oh, someone’s acquired them for a museum. There’s no way.’ And then suddenly, for whatever circumstances, they become available. It’s just such an emotional return.”

As the story goes, the Tiffany & Co. timepiece was given by three prominent women — Mrs. Madeleine Talmage Astor, Mrs. Marian Longstreth Thayer and Mrs. Eleanor Elkins Widener — as a gesture of thanks to Capt. Arthur H. Rostron, who saved them, along with nearly 700 other passengers, when R.M.S. Titanic sank after hitting an iceberg on its maiden voyage in April 14 and 15, 1912.

A ledger in The Tiffany Archives records that the watch was purchased on May 24, 1912, by Mrs. G. D. Widener for $135. “What was special about it is, the three women who were on the Titanic, which, of course, were very affluent New York Gilded Age society. All three lost their husbands in the tragedy. These were women that were the wealthiest on board,” Young said. “Captain Rostron even provided his cabin for them. He let them take it over on the journey back to New York.

“Once they landed in New York less than a month later, they had conceived of a need for a gesture of thanks, a gesture of gratitude to acknowledge his heroism. And so they came to Tiffany,” he said of the tremendous piece of history, which is engraved with the inscription “Presented to Captain Rostron with the heartfelt gratitude and appreciation of three survivors of the Titanic April 15th 1912.”

The pocket watch is a link to Tiffany’s horological legacy, which dates back to 1847, when the American house began to retail watches and clocks. In 1874, the company opened its own manufacture in the heart of Geneva, to produce timepieces. “Notable is the balance wheel with adjustable screws, and particularly the blocked polished adjustment raquette, allowing for very precise regulation of the movement’s operation,” Young said of the pocket watch. “The escapement bridge is covered by a steel plate finished with blocked polishing, with the escapement wheel’s ruby meticulously set, signifying superior manufacturing quality. The setting of a ruby is indeed a mark of high-quality watchmaking, as these jewel bearings improve the movement’s precision and longevity.”

The American jeweler isn’t keeping the timepiece close to the vest, planning to share it first at LVMH Watch Week in Los Angeles in January with select clients and press and then to the public on a wider scale. “Our intent is with our important store openings we have coming up, we will present it as one of those special moments inside these new stores as an exclusive experience where Tiffany’s history can be celebrated through objects of historic importance,” Young said.

“This watch really celebrates gratitude. It celebrates heroism. This object is really one that also represents the Gilded Age of New York and Tiffany’s role, yet it’s totally relevant to the timepieces we offer today. Everyday, people come to us and trust us with those important moments in their life. They know with Tiffany, you can give something of importance and value that will be cherished for generations.”

Coatue, Lone Pine and Third Point are placing their AI bets on power producers

Hedge-fund firms such as Coatue Management that are known for investing in next-generation technology companies have lately been piling into a sleepier sector.

Old-school power companies such as Vistra VST 0.30%increase; green up pointing triangle, Constellation Energy CEG 0.10%increase; green up pointing triangle and Talen Energy TLN 0.85%increase; green up pointing triangle have become darlings among hedge-fund firms thanks to their starring role in the artificial-intelligence boom. Such companies command the scarcest resource in the generative-AI supply chain: the extra electricity that fuels the data centers needed to train large language models and answer prompts from users of AI chatbots such as ChatGPT.

The phenomenon has made power generators some of the best-performing stocks in the S&P 500 this year—shares of Vistra have more than tripled in price—and has boosted hedge funds’ returns.

Coatue, whose founder, Philippe Laffont, muses about the coming of humanoid AI robots, owned stakes in Vistra and Constellation worth about $2.3 billion at the end of September, according to a securities filing. Coatue’s flagship fund gained about 12% through the end of October, a person familiar with the matter said.

At Steve Mandel’s Lone Pine Capital, a $17 billion investment firm with a large tech and startup portfolio, gains on positions in Vistra, Constellation and Talen accounted for about one-fifth of its main hedge fund’s nearly 22% net return this year through the end of September, according to documents viewed by The Wall Street Journal.

Dan Loeb’s Third Point counts Amazon.com, Meta Platforms and chip maker Taiwan Semiconductor Manufacturing 2330 -0.92%decrease; red down pointing triangle among its top five stock positions as part of a bet on the transformational impact of AI. But the $12 billion firm’s best-performing publicly traded holding is Vistra, helping Third Point’s flagship fund generate a 27% return this year through the end of November.

The total return of the S&P 500 is about 28% through the end of November. Broad hedge-fund indexes are up about 10% in the same span.

Hedge funds that specialize in tech, media and telecommunications thrived in 2020 and much of 2021 when the Covid-19 pandemic supercharged demand for e-commerce and other digital services. They then suffered in 2022 when rising interest rates cooled valuations for tech companies.

It wasn’t long ago that the hedge funds interested in power producers were those that traded in distressed securities. Stagnant power demand, overbuilding and high levels of debt pushed many energy companies, including Vistra’s predecessor, into bankruptcy.

Now, after a decade of flat growth, U.S. power demand is set to soar in the coming years. Some funds started paying attention to power producers after noticing that trends such as the proliferation of electric vehicles and the reshoring of manufacturing would boost demand for energy faster than supply could keep up.

The AI frenzy took that to a whole new level: It takes around six to 10 times as much electricity to process a ChatGPT query compared with a google search, according to Goldman Sachs research. Starting earlier this year, cloud-computing providers such as Amazon.com and Microsoft cut special deals with power generators to secure enough juice for their AI and data-center ambitions at what are thought to be premium prices, making more investors take notice.

Only certain power companies stand to benefit. Those active in deregulated markets, such as Texas and parts of the mid-Atlantic region, can charge the market price for power. Those with so-called dispatchable power, mostly natural-gas and nuclear plants that can quickly and reliably produce electricity when needed, are particularly prized.

Interest in this trade has spread beyond tech-focused hedge funds. Vistra joined Goldman Sachs’s “Hedge Fund VIP” list, a basket of popular holdings among the bank’s clients, earlier this year. Macro hedge funds that bet on big-picture market moves are also large participants in the AI power trade: Rallies in Vistra and Constellation helped Rob Citrone’s Discovery Capital Management gain about 47% through November in its main fund and David Rogers’s Castle Hook Partners, which also has a stake in turbine maker GE Vernova, gain more than 60% in its main fund, people familiar with the matter said.

Recent and future hedge-fund launches are looking to get in on related themes, people familiar with the matter said. Valiant Capital Management, a stock-picking hedge-fund firm, pitched Constellation at the Sohn San Francisco investment conference in October and recently started a special-purpose vehicle focused on companies working on upgrading the U.S. electrical grid. A new BlackRock hedge fund, the Blackrock Power and Energy Navigator fund, is looking to debut early next year.

Meanwhile, some energy hedge funds are exiting positions in certain power companies after the newcomers bid up valuations. Electron Capital Partners, a $3 billion firm that bets on and against utility and renewables stocks, sold its Vistra shares in the third quarter, according to a securities filing. Its flagship fund is up 23% through November, according to an investor document viewed by the Journal.