Nvidia Could Have Paid Any Price for Arm. The Problem Was That It Was Nvidia.

On September 13, 2020, Nvidia agreed to buy a company that designs almost no chips, manufactures none, and yet sits inside nearly every phone on earth. The price was up to $40 billion in cash and stock for Arm Holdings.¹ Then Nvidia's stock kept climbing, and the implied value climbed with it - past $60 billion by early 2021 before anyone had signed a final clearance.⁷ Seventeen months later the deal was dead, and Nvidia paid SoftBank a breakup fee for the privilege of walking away.³ The usual lesson is that regulators got cold feet about a giant cash pile. The real story is the opposite: the money was never the problem. The buyer was.

The official story is that the FTC blocked the deal. It is tidier than the truth. The FTC did sue, in December 2021 - but the case never reached a verdict.² Britain's competition authority ran its own investigation and concluded the deal would harm competition. The EU opened an in-depth probe. Nvidia abandoned the deal before any court ruled, citing 'significant regulatory challenges.'³ No single regulator killed it. A globally shared conclusion did.

Arm sells the same thing to everyone, including its buyer's enemies: neutrality wasn't a feature of the company — it was the whole company

To understand why three jurisdictions reached the same answer, you have to understand what Arm actually sells. It does not make finished chips. It licenses the instruction-set and microarchitecture designs that other companies - Apple, Qualcomm, Samsung, Amazon, and Nvidia itself - build their own silicon around. Its entire commercial value rests on a single property: it licenses to all of them, on terms none of them fears. Arm is a Switzerland that happens to underpin the world's processors. That neutrality is not a nice-to-have. It is the asset.

Now look at the buyer. Nvidia is not a neutral party to the semiconductor industry - it is one of its fiercest competitors, racing Qualcomm, AMD, and the cloud giants for data-center and automotive silicon. Hand Arm to Nvidia and you ask every one of those rivals to keep buying the foundational technology for their chips from a company that profits when their chips lose. The FTC put the mechanism plainly: a combined firm would have the ability and the incentive to undermine rivals by restricting their access to Arm technology.² You cannot promise your way out of an incentive. You can only not create it.

Why behavioral promises couldn't save it: you can't firewall away an incentive that the deal itself creates

Nvidia's defense was that it would keep Arm open - same licenses, same terms, firewalls between the divisions. The trouble is that regulators reviewing a structural conflict treat behavioral promises as the weakest possible remedy, because they require permanent enforcement against a permanent incentive. The harm here wasn't a one-time price hike a regulator could monitor. It was the slow ability to shape which competitors got Arm's next generation, when, and on what terms - exactly where data-center and automotive battles are won. The CMA's final report found the deal would harm competition across multiple semiconductor markets, naming data center, automotive, and IoT.⁴ The EU opened its in-depth phase citing Arm's foundational role in the ecosystem.⁵ Three regulators, one diagnosis: the conflict lived in the structure, and you cannot remedy structure with a pledge.

Wasn't this just protectionism dressed up as antitrust?: what happened after the collapse answers the charge better than any regulator could

The fair objection is that this looks like regulators reflexively blocking a big American tech deal - and that licensees who didn't want a stronger Nvidia simply weaponized the agencies. There's a grain of truth in the second half: rivals and customers filed comments raising concerns, and the FTC cited that ecosystem opposition. But the formal action was taken by governments on a competition theory, not by the licensees themselves, and that theory was specific, not sweeping. Nobody serious argued Arm would hand Nvidia a 'monopoly on chip design' - Arm doesn't design finished chips. The documented worry was narrower and harder to dismiss: the incentive to degrade terms or withhold next-generation IP from direct competitors. The cleanest rebuttal to the protectionism charge is what happened next. SoftBank didn't keep Arm hostage; it took the company public on the Nasdaq in 2023 at roughly a $54.5 billion valuation - more than the original $40 billion headline, and well above the $32 billion SoftBank paid in 2016.⁸⁶ The market valued a neutral Arm above the price a conflicted owner had offered. The regulators weren't protecting Arm from being sold. They were protecting what made it worth selling.

Nvidia could have offered more money, more remedies, more firewalls - and it would not have mattered, because the thing it wanted to buy stops existing the moment Nvidia owns it. Arm's worth is that everyone trusts it not to take sides. A chip company that takes sides for a living cannot hold that. The deal didn't die because $40 billion - or $66 billion - was too much. It died because the only buyer Arm was actually worth that much to was the one buyer it could never become a part of.