Eyal Waldman, co-founder and CEO of Israel-based Mellanox, who earlier this month sold his company to California-based chipmaker Nvidia for $6.9 billion, got no phone call from Prime Minister Benjamin Netanyahu, an ardent nationalist who is known to congratulate Israeli entrepreneurs who make a financial exit.
Waldman does not shy away from controversy in his country. In the midst of one of Israel’s most divisive election campaigns, Waldman took heat from hardline activists after reminding the Israeli public that Mellanox has dozens of Palestinian employees in the occupied West Bank and even in the isolated Gaza Strip. “It’s a win-win situation—we need the talent and it helps the Palestinian economy,” said Waldman.
Yet this is the least of Waldman’s worries: the more daunting challenge he faces is integrating the company he founded in 1999 with Nvidia and clearing regulatory hurdles—particularly in China. With these uncertainties in mind, the deal is expected by its chaperones to close at the end of 2019. But not everyone is comfortable with the risk: New York-based activist investment fund Starboard decided to sell its entire holding in Mellanox in February, ahead of the deal’s announcement.
With 24% of Mellanox’s $1.1 billion in 2018 revenue earned in China, Mellanox faces the idiosyncrasies of Chinese anti-monopoly regulators. In the politicized atmosphere of the US-China trade spat, it’s even harder to anticipate if China will allow a deal in the sensitive semiconductor industry.
By joining forces, Nvidia and Mellanox are betting that together they are better positioned to win over large corporate data centers, especially those working on AI and mega databases. Nvidia’s cash cow in the past several years has been selling graphics processing units to video gamers and miners of cryptocurrency, but in the fourth quarter of 2018 Nvidia’s gaming unit took a big hit with revenues sinking 45% year-on-year. As this business slows, Nvidia CEO Jensen Huang is betting $7 billion that the synergy will set Nvidia on a new course.
He told Bloomberg that data centers in the future will increasingly be built as though they are single giant computers. These computers process reams of data for artificial intelligence that power anything from image recognition to self-driving cars “with tens of thousands of compute nodes,” requiring Nvidia’s coprocessors in cloud installations and Mellanox’s inter-connections. The result, according to Huang, is warehouses full of more effective and efficient servers that work in parallel.
Mellanox’s chips power high-speed networks connecting servers. The company’s switches, adapters and other interconnection hardware are crucial components in transferring information both within and between computer servers that make up corporate data centers and Internet-based cloud services. Together the two companies hope to produce what Huang labels “next-generation datacenter-scale computing solutions.”
About a quarter of Nvidia’s revenue comes from data centers, yet despite its best efforts, growth in this business has been fraught with difficulty. “In data centers, revenue… came in short of expectations,” the company announced in January, due to softness in demand from China and from the large cloud providers such as AWS, Microsoft Azure and Google Cloud. Nvidia thinks that Mellanox will expand its reach and create new sales channels.
But risky politics
Both Waldman and Huang appear certain about approval by China’s State Administration for Market Regulation: According to Mellanox’s report to the SEC, Nvidia has promised to pay Mellanox $350 million if the deal is struck down by regulators.
They have a strong economic argument: Nvidia is not buying a competitor, but taking over a company with a complementary product, which ought not antagonize the trust busters. But China’s approval is far from guaranteed: it could wield its antitrust power to strike back at the US in the trade war.
China’s Anti-Monopoly Law came into force 10 years ago, but its current enforcer was created only a year ago when three antitrust bodies were consolidated into SAMR. Its scant track record makes forecasting the decisions of the Chinese regulator nearly impossible, adding to the uncertainties surrounding the deal.
Politically too, the deal does not appear to land in the trade war line of fire. Government watchdogs in both the US and China are busy safeguarding their telecoms infrastructure and want to prevent the other side from using equipment to install a “backdoor” to computer systems or encrypted data that bypass security mechanisms. The US government accuses Huawei of using these mechanisms to collect intelligence on America and its allies. But the systems affected by the deal are used for computing inside data centers, not for communications across the Internet.
Beijing’s desire for an end to the conflict could make it lenient. President Xi Jinping said in December that he would consider approving the previously rejected Qualcomm bid for chip rival NXP were it presented to him again. Another indication of China’s peacemaking mood is its approval in February, after almost a year of deliberations, of US company KLA-Tencor’s $3.4 billion acquisition of Israeli electronics maker Orbotech. By then the deal had been approved by regulators in South Korea, Israel, the US, Taiwan and Japan, where the companies also operate.
But as long as the trade negotiating teams continue their deliberations, the Nvidia-Mellanox deal could fall prey to posturing as the Chinese side attempts to make a last-minute coup de force to the American counterparts.
Huang’s business rationale may prove to be wishful thinking that does not give enough weight to political considerations. China has made it a national priority to build out domestic chip manufacturing and the consolidation of foreign suppliers goes counter to its national interest. Government is one of the biggest buyers in areas such as energy and defense, which may skew the decision-making process against the deal.
This will not be the first time American companies fall prey to a simplistic view of China based on uninformed, emotional interpretations. After betting big on Chinese approval of the deal, Jensen and Waldman may wake up later this year to realize that they took on a lot more risk than they anticipated.