The following is excerpted from the book “The Nvidia Way: Jensen Huang and the Making of a Tech Giant,” written by Tae Kim, a senior technology writer at Barron’s, and publishing Dec. 10 by W.W. Norton & Company. The excerpt is from a chapter about activist hedge fund Starboard Value, founded by Jeff Smith.

Early in 2013, Nvidia’s shareholders were getting restless. The stock price had been roughly flat for four years, and the financial performance was mixed. In its latest quarter ending in January, sales were up 7 percent year-­over-­year, but earnings were down 2 percent.

Nvidia had a strong balance sheet of about $3 billion in net cash, which was a significant asset when the overall market value of the company was $8 billion total. However, its growth rate was only in the single digits, which resulted in a price-­to-­earnings (P/E) multiple of just 14 times earnings. After backing out Nvidia’s cash on hand, Starboard believed that the company was severely undervalued, and its core assets had far more room to grow. The fund pounced: according to Securities and Exchange Commission 13F filings, the hedge fund accumulated a stake of 4.4 million shares in Nvidia, worth about $62 million, during the quarter ending in June of 2013.

Some executives at Nvidia weren’t excited about having Starboard as an investor. One senior Nvidia executive said the company’s board was very worried that the activist fund would force a reorganization of the company, install its own board, and make Nvidia cut back on its investments in CUDA—­the kind of drastic reshaping that it would attempt with Darden the following year. Another Nvidia executive said Starboard wanted a board seat, but the board had pushed back.

Still, the relationship never became too antagonistic. “I don’t think it ever got to what I would call a crisis stage. You know DEFCON 1?” one Nvidia executive said, referring to the alert system used by the U.S. military for nuclear war. DEFCON 5 indicates peace, while DEFCON 1 means nuclear war is imminent. “It got to DEFCON 3.”

The Starboard team met several times with Jensen and other Nvidia leaders to discuss strategy. Looking back on the investment years later, Smith said that Starboard primarily advocated for an aggressive stock buyback program and a de-­emphasis on non-­GPU projects such as phone processors. Starboard refrained from applying additional pressure after the meetings. The hedge fund eventually got its wish on the buybacks. In November 2013, Nvidia made two announcements: a commitment to buy back $1 billion of stock by fiscal 2015 and the authorization of an additional $1 billion stock buyback. The stock price rallied about 20 percent in the ensuing few months, and Starboard sold its position in Nvidia by March the following year.

Far from a contentious relationship, Nvidia and Starboard seemed to work well together in this brief period.

“We were incredibly impressed with Jensen,” said Smith.

For his part, Jensen recalls the meetings with Starboard but doesn’t particularly remember what was discussed. Before he knew it, Starboard was no longer an investor. But that wasn’t the end of Starboard’s influence on the chip industry, and on Nvidia.

A company called Mellanox was founded in 1999 by several Israeli technology executives, led by Eyal Waldman, who became its CEO. Mellanox provided high-­speed networking products for data centers and supercomputers under the “InfiniBand” standard and soon became an industry leader. It had impressive revenue growth, going from $500 million in 2012 to $858 million in 2016. However, its high research and development spend left it with very thin profit margins.

In January 2017, Starboard bought an 11 percent stake in Mellanox. It sent a letter criticizing Waldman and his team for their disappointing performance over the prior five years. Mellanox’s share price had fallen even though the semiconductor industry index had risen in value by 470 percent. Its operating margins were half of the average of its peer companies. “Mellanox has been one of the worst performing semiconductor companies for an extended period of time,” read Starboard’s letter. “The time for fringe changes and marginal improvements has long passed.”

After a long series of discussions with the board, Starboard and Mellanox reached a compromise in June 2018. Mellanox would appoint three Starboard-­approved members to its board and give the hedge fund additional future rights if Mellanox didn’t meet certain undisclosed financial targets. Even with those concessions in hand, Starboard retained the option of waging a proxy fight to replace Waldman. Alternatively, Mellanox could choose to sell itself to a company that could generate better returns on its assets than it could as an independent company. The groundwork was laid for what would be one of the most consequential transactions in the history of the chip industry.

In September 2018, Mellanox received a nonbinding purchase offer from an outside company at $102 per share—­a premium of almost a third over its current stock price of $76.90. Mellanox was now fully in play. It solicited an investment bank to seek other bidders and eventually expanded its list of potential buyers to seven in total.

Jensen wasn’t thinking about acquiring Mellanox when it became available, according to another Nvidia executive. But he quickly saw the strategic importance of the asset, decided Nvidia had to win the auction, and joined the hunt in October.

Eventually, the list was narrowed down to three serious bidders: Nvidia, Intel, and Xilinx, which made chips primarily for industrial uses. The three potential buyers got into a multi-­month bidding war, with Intel and Xilinx topping out around a bid of $122.50 a share. Nvidia went just a little bit higher, at $125 per share. It won the bidding war on March 7, 2019, for an all-­cash offer of $6.9 billion.

Days later, Nvidia and Mellanox made the deal public and held a conference call with analysts and investors.

“Let me tell you why this makes sense for Nvidia and why I’m excited about it,” Jensen said. He talked about how the demand for high-­performance computing would rise—­how workloads including AI, scientific computing, and data analytics required enormous performance increases, which could only be attained through accelerated computing with GPUs and better networking. He explained how AI applications would eventually require tens of thousands of servers connected to one another and working together in concert, and the market-­leading networking technology from Mellanox would be critical to make that possible.

“Emerging AI and data-­analytics workloads demand data-­center-­scale optimization,” he said. Jensen was predicting that computing would move beyond one device—­that the entire data center would become the computer.

Jensen’s vision came true just a few years later. In May 2024, Nvidia disclosed that the portion of the company that was formerly Mellanox had generated $3.2 billion in quarterly revenue, up more than seven times from the final quarter in early 2020 in which Mellanox reported as a public company. After just four years, the former Mellanox business, which had cost Nvidia a one-­time fee of $6.9 billion, was generating more than $12 billion in annualized revenue and growing at triple-­digit rates.

“Mellanox was frankly a wonderful thing thrown in our lap by activists,” a senior Nvidia executive said. “If you talk to AI start-­ups today, InfiniBand, Mellanox’s networking technology, is incredibly important to scale the computing power and make everything work.”

Brian Venturo, cofounder and CTO of CoreWeave, a leading GPU cloud-­computing provider and a customer of Nvidia’s, argues that InfiniBand technology still has the best solution to minimize latency, control network congestion, and to make workloads perform efficiently.

Mellanox was a happy accident for Nvidia in some respects. Jensen wasn’t on top of it from the start. But once Nvidia identified and understood the opportunity, it made the decision to pursue Mellanox aggressively. It was a great deal, though the outcome depended on Nvidia’s ability to execute once the new business became part of the company. In those ways, Mellanox was a typical Nvidia achievement: the company pounced when others didn’t, and Mellanox helped power Nvidia’s rise to dominance in the AI space.

“It’s absolutely going to go down in history as one of the best acquisitions ever,” Nvidia’s head of global field operations, Jay Puri, said. “Jensen realized that data-­center-­scale computing requires really good high-­performance networking, and Mellanox was the best in the world at that.”

After seeing Nvidia achieve all that is has over the past decade, Jeff Smith of Starboard Value had one summarizing thought, too.

“We never should have exited the position.”