It is hard to understand what is going on right now without going back in time, to where these men’s paths crossed. Well, “crossed” should really be read as a crossroads where everyone goes through Musk. Because, let’s face it, he is the one at the centre of all the dealing right now. To the point that he could well be the one to tip Silicon Valley over. Or tech. Or the United States. And that is not just a pretty anaphora.
The tensions between Elon Musk and Masayoshi Son, which have been stirring Silicon Valley for a long while, reportedly go back to December 2016. Some even speak of hatred from the former for the latter. And of pragmatic indifference the other way round.
SoftBank 🇯🇵 then puts $1bn into OneWeb 🇬🇧, promising to create 3,000 American jobs, a pledge made directly to President-elect Trump. For Musk, for whom Starlink 🇺🇸 is still just a project that would launch in 2018, it is a declaration of war.
As so often with SoftBank 🇯🇵, the investment ends badly. OneWeb 🇬🇧 goes bankrupt in March 2020, and it is Bharti Global 🇮🇳, Eutelsat 🇫🇷 and the British government that recapitalise the company, leaving SoftBank 🇯🇵 with less than 12%. The company would go on to become Eutelsat OneWeb 🇫🇷 at the end of 2023.
But in the meantime, the story got more complicated. In April 2017, Son and Musk meet to discuss a possible investment in Tesla 🇺🇸, going as far as discussing taking it private.
The deal falls through. But when, a year later, Musk wants to raise $500M, Son comes knocking again (the Son also rises, hehe). A godsend for the early investors, who like the idea of fresh cash and a possible synergy with SoftBank Telecom 🇯🇵.
Musk flatly refuses, stung by the Tesla 🇺🇸 non-deal, but above all by SoftBank 🇯🇵’s investment in a competitor. All the more so because Musk is wary of SoftBank’s spray & pray strategy: investing in every company in a sector, as it did in ride-hailing (Uber 🇺🇸, Didi 🇨🇳, Grab 🇸🇬), to consolidate positions and be sure of pocketing the market, whoever the big winner turns out to be.
In fact, SoftBank 🇯🇵 later became a major investor in the electric and autonomous vehicle industry, funding many of Tesla 🇺🇸’s competitors. Notably with a huge investment in Rimac 🇭🇷, the Croatian electric supercar maker that openly aims to compete with Tesla 🇺🇸 at the top end of the market.
In 2018, Masa himself meets Jens Peter Clausen, then head of Tesla 🇺🇸’s Gigafactory, to put him in charge of Zymergen 🇺🇸, a biotech that had just raised $400M from the Vision Fund 🇯🇵. Enough to make Musk furious all over again.
These two had everything going for them to get along. Two brilliant minds. Two visionaries obsessed with the future. Two men convinced that artificial intelligence would be the next revolution. And that it had to be in the right hands.
They reportedly first met in the 2010s. Altman is a rising figure in Silicon Valley thanks to Y Combinator. Musk is already a legend. A partner at the accelerator takes Altman to visit SpaceX 🇺🇸, and he comes out fascinated. He would describe that moment on his blog3 later:
What stuck with him, he wrote, was the look on Musk’s face: absolute certainty when he talked about sending big rockets to Mars.
The admiration is mutual: Musk sees in Altman a sharp mind, able to challenge the tech giants with a radically different approach.
A dialogue begins. In 2014, they start talking about artificial intelligence and its dangers. Musk regularly uses the word robocalypse at the time. Stopping progress is impossible, but they can at least make sure it does not fall into the wrong hands. In May 2015, Altman sends an email4 to Musk: 5
“Been thinking a lot about whether it’s possible to stop humanity from developing AI. I think the answer is almost definitely not. If it’s going to happen anyway, it seems like it would be good for someone other than Google to do it first.”
The idea takes shape: create an AI lab able to rival DeepMind 🇬🇧, while making sure the technology stays accessible to all. OpenAI is unveiled with great fanfare in December 2015. Altman talks of a non-profit meant to save the world from a dystopian future and highlights his relationship of trust with Musk. Musk pushes the project to display massive ambition: rather than $100M, he suggests a $1bn commitment. He even promises to cover whatever is missing, a promise that would turn into just $45M out of his own pocket.
From the start, tensions can be sensed behind the scenes. Musk is in a hurry. He wants an OpenAI capable of toppling Google, with fast execution and colossal resources. But the organisation moves at its own pace, following Altman’s vision: a methodical, collective approach, less aggressive than that of the Tesla boss.
Then comes the first real point of friction, in 2016. OpenAI negotiates a partnership with Microsoft offering privileged access to its cloud infrastructure. Altman defends the deal as a way of making up for the technological gap and the lack of funding. Musk, for his part, explodes by email:
The deal, he wrote, made him feel nauseous.
He refuses to let OpenAI become too dependent on Microsoft 🇺🇸. Altman, ever pragmatic, adjusts the offer: no marketing commitment, no obligation to promote Azure. Musk ends up accepting, not without one last Muskism:
Not looking like Microsoft’s marketing puppet, he wrote, would be worth far more than $50M.
But the real issue lies elsewhere. From the beginning, OpenAI has been caught between two opposing forces:
To attract the best researchers, Altman sets up high salaries and unprecedented compensation, mixing startup perks with academic prestige. Up against Google and DeepMind, the battle for talent is fierce. Which is what Altman explains to Musk in 2015:
DeepMind, Altman warned, was about to try to poach everyone at OpenAI the next day to kill it, and he asked Musk whether he was fine with proactively raising salaries by 100-200k.
Musk approves.
Without hesitation.
But he has a problem: if he has never been able to work with a board, or co-lead a project, it is because he is hot-headed, moody, and needs to control everything. Immediately.
In 2017, tensions peak as Musk starts to feel OpenAI slipping away from him. He begins to doubt that OpenAI can survive as a non-profit. In his view, the ambition of matching Google DeepMind would require billions in computing power. The organisation, for its part, realises it will have to raise colossal sums to reach its goals.
The summer of 2017 marks a turning point: Musk and Altman discuss turning OpenAI into a for-profit entity to attract investors. But Musk does not just want to raise money. He demands total control.
After quietly sounding out investors around him, he proposes a billion-dollar fundraise to cement his dominance. But Altman and the other founders smell a trap. 6
The goal of OpenAI, they told him in substance, was to avoid an AGI dictatorship; they feared DeepMind might create one, but creating a structure in which he could become dictator would be no better.
In September 2017, Musk pushes to create a new structure and take the helm. He demands more than 50% of the shares, the CEO job and absolute control over the project’s direction. His aim was to value his stake at more than $80bn in order to… go to Mars. At the same time, he registers an entity called Open Artificial Intelligence Technologies, Inc., a public benefit corporation meant to become OpenAI’s new vehicle.
Text message from Greg Brockman to Shivon Zilis, a Musk confidante who acted as the link between him and OpenAI
OpenAI’s board panics: Musk wants to turn the organisation into an arm of Tesla, when it was supposed to remain an independent alternative to Google. Talks turn sour, Altman and his allies refuse to give in. But they know OpenAI will soon run out of money. The estimated burn rate is around $300-500M a year before even having a viable model. And more than a billion after that.
On 20 September 2017, after weeks of negotiations over the board structure and his position as CEO, Musk once again threatens to leave and issues an ultimatum:
He had had enough, he wrote, it was the final straw: either they went off and did something on their own, or they carried on with OpenAI as a non-profit. He would no longer fund OpenAI until they had made a firm commitment to stay, or he would just be a fool essentially providing free funding for a startup.
“Discussions are over”, he concludes.
In January 2018, he goes even further: either OpenAI merges with Tesla, or he withdraws his support. Musk is now convinced that OpenAI is hopelessly behind Google.
The board says no. Musk slams the door, cuts off his funding, turns away from OpenAI and focuses on his other companies, keeping AI in the back of his mind.
But the OpenAI story obviously does not end there. In 2019, to make up for the loss of Musk’s funding, Altman makes a strategic choice: creating a capped-profit subsidiary, which immediately draws in Microsoft and other investors. Altman informs Musk of the Microsoft deal by email on 17 December. “Sounds good”, he replies laconically. But the day after Christmas, he comes back on the attack.
And when, in March 2019, OpenAI publicly announces its change of structure following Microsoft’s arrival, Musk is visibly irritated.
And this is only the beginning.
When GPT-4 comes out, he publicly accuses OpenAI of using Twitter/X to train its model, without backing up his claims, even though quite a few experts agree with him. Then he files a first lawsuit in March 2024, claiming that OpenAI had betrayed its founding principles. He withdraws it. But files a new one last August, claiming he was duped when he co-founded OpenAI. Which hit back by suing him for harassment and defamation.
And this is no doubt only the beginning.