Nobody Blocked the Nvidia-Arm Deal. That's Exactly Why It Died.

On September 13, 2020, Nvidia agreed to buy Arm from SoftBank for $40 billion - $21.5 billion in Nvidia stock, $12 billion in cash, an earn-out and an employee equity pot stacked on top.¹ It would have been the largest semiconductor deal ever signed. Sixteen months later it was dead, and Nvidia mailed SoftBank a check for a deal it never got to make. No judge enjoined it. No commission voted it down. The most consequential merger block of the decade arrived without anyone ever pulling the trigger.

The story everyone tells is simple: the FTC blocked the Nvidia-Arm deal. It's the wrong story. The FTC never blocked anything. It filed a suit and set a trial date - and the trial never happened, because Nvidia and SoftBank killed the deal themselves first.²³ The interesting question isn't why one regulator said no. It's why a company that fights antitrust suits for a living chose to surrender before its day in court.

What actually happened on February 7, 2022: a surrender written in the language of defeat, claimed as a victory by a commission that never had to win

On February 7, 2022, Nvidia and SoftBank jointly announced they were terminating the agreement, blaming 'significant regulatory challenges preventing the consummation of the transaction, despite good faith efforts by the parties.'² Read that sentence carefully: it is the language of a company giving up, not one that lost. The FTC had filed its administrative complaint just over two months earlier, on December 2, 2021, in a unanimous 4-0 vote - and had set its in-house trial to begin on August 9, 2022.³ Nvidia walked six months before it would have had a chance to argue. The day after the termination, the FTC issued a statement noting this was 'the first abandonment of a litigated vertical merger in many years,' careful to frame the outcome as Nvidia abandoning the deal rather than the Commission enjoining it.⁴ The FTC, in other words, took credit for a win it never had to earn in court.

The counterfactual: would a single regulator have killed it?: alone, the FTC was a fight Nvidia could win — which is exactly why the FTC was never the real problem

Imagine the deal faced only the FTC. Nvidia is one of the best-funded litigants on earth, and the FTC's case rested on a vertical theory - the claim that owning Arm would let Nvidia choke off rivals in three specific niches: high-level ADAS processors for cars, DPU SmartNICs, and Arm-based CPUs for cloud computing.³ Vertical cases are notoriously hard for regulators to win; the government had recently lost prominent ones. With a single venue, Nvidia's rational move is to fight, win, and close. The reason it didn't fight is that the FTC was never the only gate. It was one of five.

Stack them up and the picture changes entirely. The UK's Secretary of State had referred the deal to the CMA for a Phase 2 in-depth investigation on November 16, 2021 - and not only on competition grounds, but on national-security grounds, the harder kind to bargain away.⁵ The European Commission had opened its own in-depth probe in October 2021, worried that folding Arm into a chip company would compromise the 'neutrality' that makes Arm's designs licensable by everyone, Nvidia's rivals included.⁶ And China's SAMR had only accepted formal notification in January 2022 - meaning the largest single market in the deal's logic hadn't even started reviewing it when the clock ran out.⁸ Winning the FTC trial would have bought Nvidia nothing if London, Brussels, and Beijing each held a separate veto.

Why Arm's neutrality was the thing nobody could fix: you can promise firewalls, but you can't divest the suspicion that an owner will quietly favor itself

Arm's value is that it sells to everyone and competes with no one. Its instruction-set architecture sits inside chips from Apple, Qualcomm, Amazon, and yes, Nvidia's rivals - a Switzerland that licenses its designs across an industry that otherwise fights to the death. Put a chip company in charge of Switzerland and every licensee starts wondering whether the next design quietly favors the owner. That single worry is what united the regulators: the EU framed it as a threat to Arm's neutrality, the FTC as control over technology 'rival firms rely on.'³⁶ It was also structurally unfixable. Nvidia could promise firewalls and behavioral commitments, but the conflict of interest was the deal - you can't divest the suspicion that an owner will favor itself. A merger whose central objection can't be remedied is a merger that has no path through a determined regulator, let alone four.

The fair objection: maybe Nvidia just got the better deal anyway: a genuine consolation prize doesn't rescue the bet it was consoling you for losing

Here's the honest counter: perhaps walking away was cheap, even shrewd. Nvidia kept a 20-year license to Arm's technology, so it lost ownership but kept access.² And the termination charge of $1.353 billion, while real, was a rounding error against the AI-driven valuation surge that followed.⁷ You could argue Nvidia got the technology it needed without the integration headache or the regulatory leash. There's truth in that - but it doesn't rescue the original decision. Nvidia signed a deal it believed it needed to own Arm to win, structured nearly two-thirds of the price as its own stock, and spent eighteen months in five regulatory trenches before conceding the structure was impossible. The consolation prize was genuine; the strategic bet was still lost. SoftBank kept the $1.25 billion Nvidia prepaid at signing as a deposit, and recognized it as profit.² You don't pay a billion dollars for a deal you'd have been just as happy to skip.

The lasting lesson isn't 'antitrust is back.' It's that in a world where one company's chips touch every market, every market gets a vote - and a deal big enough to matter everywhere becomes a deal that has to be approved everywhere. Nvidia didn't lose a court case. It lost a coin flip it had to win five times in a row. The most expensive merger that never happened proved a quiet new rule of global strategy: the larger the asset, the more places it can be stopped, and stopping it requires no one to actually say no - only for no one, anywhere, to be able to say yes.