Published on Jan 07, 2025
For those not around for the decades-long string of events that led to Big Tech becoming the target of antitrust enforcers, you’re in luck. Companies’ current fights for dominance in artificial intelligence should give you an idea of what you missed in the development of Internet search, advertising technology, mobile phones and e-commerce.
Much like that previous tech era, companies now are scrambling to scoop up or forge partnerships with AI-related businesses that will make them leading market players.
Some key participants in those previous battles are the same ones to watch in growing AI markets. Big Tech companies such as Google (GOOG), Microsoft (MSFT) and Amazon (AMZN) all have noteworthy operations in the technology; Nvidia (NVDA), the No. 1 maker of semiconductors powering the AI revolution, is also drawing regulators’ attention.
Enforcers’ efforts to determine whether proposed mergers would advance or hinder AI’s development will reverberate throughout the dealmaking world. The most important AI transactions will raise key competition concepts: killer acquisitions; interoperability; potential and future competition; and exclusive dealing.
So far, regulatory reviews of AI-related deals haven’t translated into much action. Last month before the holiday break, the European Commission unconditionally approved Nvidia’s proposed acquisition of Run:ai, which makes software that optimizes data crunching on Nvidia’s graphics processing units (GPUs).
The commission found that Nvidia lacked the market power, technical ability and incentive to make it harder for Run:ai’s rivals to design software compatible with its GPUs. The regulator appears to have largely based its conclusion on the software maker’s competitors, which said they would provide plenty of options to Run:ai’s so-called GPU orchestration products.
But don’t assume this decision signals how the EC and its counterparts around the world will handle future AI deals. As the technology spreads beyond the tech sector to myriad industries, so, too, is regulators’ attention. Enforcers already are warning how algorithms can help rivals share sensitive information, fix prices and otherwise collude. Any deals that would give a company an incentive and ability to engage in such conduct will likely raise red flags in government reviews.
AI-related mergers could provide yet another reminder to enforcers—if one is needed—that technology markets, driven by network effects and first mover advantage, can incline toward consolidation.
The decision not to open an in-depth, phase 2 investigation of the Run:ai transaction was anticlimactic after Italy’s competition regulator asked the EC to review the deal even though it didn’t meet the commission’s revenue thresholds.
Initially, the commission had reservations about Nvidia owning the software maker. When accepting the Italian enforcer’s referral in October, the EC said the transaction “threatens to significantly affect competition in the markets where NVIDIA and Run:ai are active.”
Yet its phase 1 investigation found otherwise, and the commission concluded that “the availability and widespread use of [other] tools” would ensure Nvidia’s GPUs will continue to work with Run:ai’s competitors.
One industry participant mostly agreed with the commission’s assessment.
“There are many options in the market,” said Haseeb Budhani, co-founder and CEO of Rafay Systems, which makes software that helps companies deploy Run:ai and similar products on GPUs.
Budhani argued in theory that Nvidia could make it more difficult for Run:ai rivals’ software to run on its chips by tweaking its CUDA software to work exclusively with Run:ai. But the tech executive said he doesn’t see why Nvidia would do so. CUDA is the software that programmers use to help design products to run on Nvidia GPUs.
Nvidia has more incentive to encourage software makers, including Run:ai competitors, to design their products to work on its chips and CUDA libraries so the company could sell more hardware, he said.
“It’s better to have 10,000 software companies in their universe,” said Budhani. His company participates in an Nvidia program that helps AI startups, but Rafay also supports GPUs for Nvidia’s chief AI chip rival, AMD.
When asked what he would do if he were an Nvidia executive, Budhani said he would “focus on the ecosystem” and “invest heavily in ensuring that the ecosystem flourishes. The broadest ecosystem will ultimately win.”
Nvidia has shown a willingness to do this by making Run:ai software open source, Budhani said. The company is doing this “to help the community build better AI, faster,” according to a December 30 Run:ai blog post announcing that Nvidia had completed the acquisition.
In recent cases, the EC has considered whether building an “ecosystem” of products and services with mergers unfairly strengthens a buyer’s hold on market or prevents emerging competitors.
Although the agency’s full decision hasn’t yet been published, it appears the EC didn’t analyze the deal under the ecosystem legal argument and instead undertook a more standard review of non-horizontal issues that the merger raises.
The antimonopoly group Open Markets Institute has said the EC will “rue” its decision to skip an in-depth probe in the case.
“We view this transaction as a means for Nvidia to reinforce its dominant position in the supply of advanced GPUs by integrating Run:ai software to build an additional barrier around its chip empire,” the policy group said in a statement. “The ensuing bottleneck poses a serious risk for the entire AI sector – weakening the supply chain and undermining Europe’s AI companies to remain competitive in global markets.”
Some in the AI industry want enforcers to carefully monitor Nvidia’s M&A activity to ensure that it doesn’t use its hardware dominance to favor some software products over others.
Nvidia didn’t respond to a request for comment.
Even after the deal’s closing, enforcers are still focused on Nvidia. DOJ, the EC and France’s Autorité de la concurrence are inquiring into whether Nvidia’s conduct regarding its GPUs is anticompetitive. And the State Administration for Market Regulation (SAMR) in China has opened an investigation into whether Nvidia violated the conditions of the regulator’s approval of the company’s 2020 acquisition of Israeli chipmaker Mellanox.
In its Run:ai decision, the commission acknowledged that Nvidia “likely holds a dominant position” in making GPUs that power corporate data centers.
The FTC for its part is focusing on Microsoft’s conduct in the sector. The U.S. agency issued a civil investigative demand, similar to a subpoena, to the company last month, about its AI, cybersecurity and cloud businesses, according to a source familiar with the matter.
As competition agencies burrow into these investigations, antimonopolists and other Big Tech opponents are watching warily. These groups have said Big Tech companies previously delayed necessary enforcement actions in search, adtech, social media and ecommerce. The companies did so by misleading government investigators about how they would use deals to piece together dominant businesses or snuff out nascent competitors, according to these critics.
They point to the FTC’s greenlighting of Google’s proposed takeover of DoubleClick in 2007 and AdMob in 2010 as examples of how the search giant created a dominant business in plain sight. Those acquisitions helped the search giant become the market-leading adtech company, a dominance Google abused, DOJ said in its antitrust suit against the business.
The FTC came under fire for allowing Facebook (META) to buy two growing rivals of the social media company—Instagram in 2012 and WhatsApp in 2014. The deals are central to the FTC’s suit against the company alleging Facebook abused its dominance in social media.
U.S. enforcers appear determined to avoid repeating this hands-off approach with AI, issuing a joint statement in July affirming their intention to monitor the technology’s development closely.
The agencies are facing pushback. The sector’s defenders have said government action would thwart U.S. companies, particularly Nvidia, and give foreign rivals an advantage.
This argument could resonate with some regulators, who consider leadership in AI of national security importance. The Biden administration has restricted the sale of Nvidia chips to China based on national security concerns, a stance that President-elect Donald Trump is expected to adopt in his second term.
A complicating factor in how European agencies will proceed with AI deals is the regulators’ view that they must accommodate the creation of national champions—domestic companies that are big enough to compete with Chinese and U.S. rivals. This consideration could lead to a more permissive atmosphere in Europe for mergers in the AI sector.
The UK’s Competition and Markets Authority initially positioned itself as a leading enforcer in tackling antitrust issues raised by AI. But the rhetoric of its senior officials is shifting to align with the Labour government’s pro-business agenda.
“We are fully supportive of the important role of the largest firms here – they have driven fantastic progress in this space and are a key source of investment in the UK,” CMA Chief Executive Sarah Cardell wrote in a November blog post.
To defend AI-related mergers, tech sector advocates have dusted off an argument they used to delay past enforcement efforts: Governments should avoid trying to pick winners and losers in developing markets they don’t understand.
Enforcers have been at pains to develop the knowledge and expertise to grapple with tech deals, including those involving AI. The coming months will reveal whether the enforcers have decided to aggressively act now against anticompetitive deals in these markets – or wait to do so years from now.
M&A Matrix
The remainder of this article is currently locked and only available to subscribers. To read more, please contact us or request a trial.
