The following is an excerpt from the book “The Nvidia Way: Jensen Huang and the Making of a Tech Giant,” written by Tae Kim, our senior technology writer, and published by WW Norton & Company on December 10th. . This excerpt is from a chapter about Starboard Value, an activist hedge fund founded by Jeff Smith.
In early 2013, Nvidia shareholders were starting to get restless. The stock price has been roughly flat for four years, and business results have been mixed. In its most recent quarter, which ended in January, sales rose 7% year over year, but profits fell 2%.
Nvidia had a strong balance sheet with approximately $3 billion in net cash, a significant asset at a time when the company’s overall market value totaled $8 billion. However, the growth rate was only in the single digits, and the price-to-earnings ratio (P/E) was only 14 times. After writing down NVIDIA’s cash reserves, Starboard believed the company was significantly undervalued and its core assets still have room for growth. The funds pounced, and the hedge fund accumulated 4.4 million shares of Nvidia stock, worth about $62 million, in the quarter ending June 2013, according to a 13F filing with the Securities and Exchange Commission.
Some Nvidia executives were reluctant to have Starboard join as an investor. One NVIDIA executive said the company’s board is very concerned that activist funds will force the company to reorganize, install its own board of directors, and that NVIDIA will reduce its investment in CUDA. This is the kind of fundamental restructuring that NVIDIA is attempting with Darden. following year. Another Nvidia executive said Starboard wanted a board seat, but the board opposed it.
Still, the relationship never became too adversarial. “I don’t think we’ve reached what’s called a crisis stage. Are you familiar with DEFCON 1?” one Nvidia executive said, referring to the warning system used by the U.S. military for nuclear war. DEFCON 5 indicates peace and DEFCON 1 means nuclear war is imminent. “I went up 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 Starboard primarily advocated an aggressive stock repurchase program and a reduced emphasis on non-GPU projects such as phone processors. Starboard declined to press further after the meeting. Hedge funds finally got their wish with share buybacks. In November 2013, NVIDIA made two announcements. One is a commitment to buy back $1 billion in stock by fiscal year 2015, and the other is authorization for an additional $1 billion in stock repurchases. The stock price rose about 20% in the months that followed, and Starboard sold its Nvidia position by March of the following year.
Far from a contentious relationship, Nvidia and Starboard seemed to work well together in this short period of time.
“We were very impressed with Jensen,” Smith said.
Jensen recalls meeting with Starboard, but doesn’t remember specifically what was discussed. Before I knew it, Starboard was no longer an investor. But Starboard’s influence on the chip industry and Nvidia didn’t end there.
The company Mellanox was founded in 1999 by several Israeli technology executives led by Eyal Waldman, who became CEO. Mellanox quickly became an industry leader with its high-speed networking products for data centers and supercomputers based on the InfiniBand standard. The company’s sales have grown from $500 million in 2012 to $858 million in 2016. However, profit margins are thin due to high R&D costs.
In January 2017, Starboard purchased an 11% stake in Mellanox. He sent a letter criticizing Waldman and his team’s disappointing performance over the past five years. Mellanox’s stock price fell even as the Semiconductor Industry Index rose 470 percent in value. Operating profit margin was half the average for other companies in the same industry. “Mellanox has been one of the worst-performing semiconductor companies over a long period of time,” Starboard’s letter said. “The time for peripheral changes and marginal improvements is long past.”
After a series of lengthy discussions with the board of directors, Starboard and Mellanox reached a compromise in June 2018. If Mellanox fails to meet certain undisclosed financial targets, Mellanox will appoint three Starboard-approved members to its board of directors and provide the hedge fund with additional future rights. Even with these concessions in hand, Starboard still has the option of pursuing a proxy fight to replace Waldman. Alternatively, Mellanox could choose to sell to a company that could generate a higher return on assets than as a standalone company. The foundation was laid for what would become one of the most significant deals in the history of the chip industry.
In September 2018, Mellanox received a non-binding acquisition offer from an outside company for $102 per share. This is a nearly one-third premium to the current stock price of $76.90. Mellanox was in full effect. The company solicited other bidders from investment banks, eventually expanding its list of potential buyers to a total of seven companies.
Another NVIDIA executive said Jensen wasn’t considering acquiring Mellanox when it became available. But he quickly realized the strategic importance of the asset, decided NVIDIA had to win the auction, and joined the search in October.
Ultimately, the list was narrowed to three potential bidders. Nvidia, Intel, and Xilinx, which manufactures chips primarily for industrial use. The three potential buyers engaged in a months-long bidding war, with Intel and Xilinx winning the highest bid at around $122.50 per share. Nvidia rose slightly to $125 per share. The company won the bidding war on March 7, 2019 with an all-cash offer of $6.9 billion.
A few days later, Nvidia and Mellanox announced the deal 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 will increase, meaning that workloads such as AI, scientific computing, and data analytics will require significant performance improvements, and that it will require faster computing with GPUs and more. We talked about what can only be achieved through good networking. He believes that AI applications will ultimately require tens of thousands of servers to be interconnected and work together, and that Mellanox’s market-leading networking technology is essential to making that possible. I explained.
“Emerging AI and data analytics workloads require data center-scale optimization,” he said. Jensen predicted that computing would move beyond a single device and that entire data centers would become computers.
Jensen’s vision was realized just a few years later. In May 2024, Nvidia revealed that the former Mellanox portion of the company generated $3.2 billion in quarterly revenue. This is more than a seven-fold increase compared to the final quarter of early 2020, when Mellanox was reported as a publicly traded company. The former Mellanox business, which paid a one-time fee of $6.9 billion to Nvidia, was generating more than $12 billion in annual revenue and growing at triple-digit rates just four years later.
“Frankly, Mellanox was a great thing that was thrown into our laps by activists,” said a senior Nvidia executive. “When you talk to AI startups today, Mellanox’s networking technology, InfiniBand, is critical to scaling their computing power and making everything work.”
Brian Venturo, co-founder and CTO of CoreWeave, a leading GPU cloud computing provider and Nvidia customer, said InfiniBand technology has the ability to minimize latency, control network congestion, and speed up workloads. We argue that we still have the best solution to run it efficiently.
Mellanox was, in some ways, a lucky accident for Nvidia. Jensen was off from the start. However, once Nvidia identified and understood the opportunity, it decided to aggressively pursue Mellanox. This was great, but the outcome depended on Nvidia’s ability to execute once the new business became part of the company. In that sense, Mellanox was a typical Nvidia achievement. The company jumps where others don’t, and Mellanox helped Nvidia dominate in AI.
“This will definitely go down in history as one of the best acquisitions of all time,” said Jay Puri, head of global field operations at Nvidia. “Jensen realized that data center-scale computing requires very good, high-performance networking, and that Mellanox is the best in the world.”
Seeing all that Nvidia has achieved over the past decade, Starboard Value’s Jeff Smith also summed up his thoughts:
“We should never have left that position.”
