Elon Musk: From Zero to Trillion-Dollar Hero? | BBC News

Hello. Welcome to AI Decoded. Two battles this week, thousands of miles apart, that could well shape the future of artificial intelligence. First, the courtroom drama in Oakland, California. Elon Musk accused Sam Altman of stealing a charity. The world's richest man said he had donated nearly $38 million to OpenAI on the basis it would be run as a non-profit, for the good of all humanity. Instead, Altman commercialised it and built it into a company worth a trillion dollars. The jury ruled against Musk, not because he was wrong, but because in their view he'd waited too long to say so. In the final days of that hearing, Musk abandoned the court and headed for Beijing to join President Trump at his high-stakes summit with President Xi. Because while the court case was important to who controls and steers the most powerful AI companies on Earth, the bigger battle, in China, was over who controls the chips that power AI itself. Daryl Flack is a government advisor on cyber security. He spent 25 years defending the UK's national interests, and he is the co-founder of Avela Security. Nina Shang has written several books on the AI industry in China, Red AI and the US-China tech war, and also with us our regular co-host Dr Stephanie Hare, author of Technology Is Not Neutral. Welcome to you all. Daryl, Sam Altman is now free to cement his commercial control of OpenAI, which appears headed towards one of the biggest IPOs that we've seen. The company is free to pursue a data centre expansion that could cost hundreds of billions of dollars. So on the face of it, this verdict in Oakland, California this week was hugely consequential. Absolutely. I think it was a win-win all round for Teev OpenAI. I think Musk was looking for 150 billion dollars in damages. He was looking to have Altman removed from the board, and he was looking to unwind all of that commercial structure around that, and he failed on all those counts. And that doesn't mean that OpenAI wasn't guilty, it just means by the statute of limitations, Musk left it too late to take it to court. So overall, if you're Sam Altman right now, you're sitting in a very good position and you're quite happy with that outcome. He says he will appeal, and maybe he does have a case, because we know that they did create this commercial arm which has become hugely profitable. But maybe in the eyes of the jury, Elon Musk, the richest man in the world, is the wrong person to make the argument. Well, I think when he also spoke to OpenAI around 2018, he was looking for Tesla to buy a stake and for it to be taken over by Tesla, which isn't a charity. So I think when you've got a sense of direction that you want to take what started out as a charity to be a commercial organisation, and then you get removed from that board or you leave that board, and it subsequently gets taken to Microsoft, when you go back many years later to cry foul, I think many on the jury would have seen that and thought actually it's just sour grapes. Nina, obviously the focus is on these very big personalities, Elon Musk and Sam Altman, which is interesting in and of itself to the rest of us, but this verdict is a very important win for Microsoft too, because they are closely entwined with OpenAI and that relationship is crucial to their AI development. Yes, this is clearly good for OpenAI and Microsoft because it removes a huge cloud of uncertainty hanging over an important partnership, but it's also not that bad for Elon Musk. Like you say, the world's richest man and SpaceX valuation continues to soar, so I guess we don't have to shed tears for both of them. Now that's a really interesting point. I think the industry writ large, Stephanie, sometimes tries to portray this united front when it comes to AI, to safety, to research, to inclusivity, but this case served as a very powerful reminder to all of us that AI giants are not charities. They're not there for the good of humanity. They are in it for the money and it is fierce competition. Yeah, I mean we saw this recently at the last AI Summit in India, where Sam Altman would not hold hands at the end when all the leaders held hands with Anthropic CEO Dario Amodei. They refused to hold hands. Yeah. They couldn't even do that. So they are fiercely competing for market share, to see who's going to have the best model, and even for who's a verb. So if you think about it, when you say, I'm going to just check it up on ChatGPT, what most people mean by that is, I'm just going to use generative AI, but ChatGPT is a verb, like I'm going to Google it, you know? So it's who's going to be the one that becomes the trusted partner for enterprise, but also in culture, in popular culture, when you think generative AI, who do you think? The question I'd love to ask with Nina joining us today is, in the United States, we often hear regulation will hinder innovation and then China will win. Is there a parallel conversation happening in China? Are they saying, if we don't stop regulating so much, the United States will just have this lead that we will never upset? What's the view there? I would say perhaps not so much in China in terms of having regulation would hinder technology's advances. I feel like the Chinese government definitely feel they are in the driver's seat and they have this God's view on China's tech development, as well as their citizens' well-being, and they have moved very quickly to issue laws banning youth from using addictive social media apps, but also, as you remember several years back, banning youth from playing internet games. So I think Beijing is more concerned in terms of AI competition as a national key strategic industry, and also in the perspective of national security. But regulation, they feel, is not something that hinders development, but rather could overall improve society's well-being. But let's be honest, Donald Trump would have loved this fight this week. He's all about making money. That's what drives him. That's the drug he takes, is let's have more of this competition that expands American competition. Except there are more than 60 loyal allies who've written to the White House this week within the MAGA movement, and Steve Bannon is one of them, saying that he wants more rigorous testing of the sort of pre-flight models of AI because they pose such a risk. So it seems to me that while Trump is driving the commercialization of the industry, the White House policy on this is at odds with the MAGA base. There's a deep, deep irony here, which is that the company that's probably the leading AI company in the United States on AI safety, which is anthropic, was itself declared a security risk, a supply chain risk, by Donald Trump. And that was the first company that was in the US classified systems. It's being used in Iran, etc. Still being used right now because you can't just pull that stuff out just because somebody wakes up one day and decides to declare you a supply chain risk. But what's also really interesting, and perhaps explains the change in position of Steve Bannon and other Trump advisers, is that Anthropic itself had to convene a group of other tech companies and say, we have created something called Mythos, which is so powerful and so dangerous that it could reveal all of the vulnerabilities that everybody's got around the world. We're talking banking, we're talking utilities, we're talking weapons, and we can surface those errors before people have a chance to patch them. So this is the nightmare scenario. So it was a private company, Anthropic, that convened this meeting, has been very safely, yeah, they were doing a regulator's job. You might have a commercial interest for doing that. Exactly, but they were doing a regulator's job, and the deep irony is they were doing it while themselves being declared a supply chain risk by the White House. But suddenly we're all talking about Claude and Mythos, and as I say, you fly a kite saying this is dangerous, this is far-reaching. It's a great marketing tool, yeah. But do you buy into some of that, the concerns? I think there's absolute concerns there. I think AI is going to change the landscape when it comes to identifying vulnerabilities and how you patch them. So absolutely, it is going to create a much faster time frame for organisations to be able to be aware and to enact some of the changes they're going to need to change. But I think doing what they did and inviting only the cool people in the tent, they chose the tech companies of the products that they wanted to have in their group. And whilst you could say that it's altruistic that they're trying to protect the world, equally they're trying to create a top tier of tech companies that have this information that can patch vulnerabilities fastest. So where does that leave anybody who's not at that top table? Anyone else who wants to... A critical infrastructure for a start. Absolutely. Yeah. If OpenAI was launched for the good of humanity, there's clearly a long way to go to convince the public that that's what AI companies are doing. Axios published a poll this week which tells us fundamentally people are turning against AI. Over 70% of Americans think it's moving too fast. Negative views have doubled in just three years. Only 18% of young people say they feel hopeful about it. If AI were a candidate for political office, says Axios, it would be losing in a landslide. And this court case ain't going to help. No. I mean, it is interesting that many of these AI companies, their CEOs have been writing these very long essays that they'd like to publish to great fanfare, saying we're going to wipe out most white-collar jobs within 18 months, two years tops, and then expect people to be happy about that news. And in the United States, there's another exacerbating factor, which is if you go to university there, you could very conceivably walk out with a six-figure debt. So you've got to get a job and start earning and paying off that debt straight away. So no, you're not going to enjoy AI, which yes, may have helped you with your homework or something that's convenient or a bit fun making memes, but if it's making it where you can't have a career path and you can't imagine ever being able to afford a home, have a family, like not being able to earn makes people feel very insecure so they make different decisions. We've seen that particularly with this younger generation, they already weren't feeling hopeful. And now we've added this. Can I show you all this? This is a clip that went viral this past week. It features one of the guest speakers at a graduation ceremony at the University of Central Florida. And she began by telling her students that AI will be the next industrial revolution. Just watch the reaction. The rise of artificial intelligence is the next industrial revolution. I struck a chord. So today we stand on this edge of another technological transformation, one that will be larger, faster, and more consequential than what came before. It will touch every profession, every classroom, every hospital, every laboratory, every person, and every relationship you have. I know what many of you are feeling about that. I can hear you. There is a fear. There is a fear in your generation that the future has already been written, that the machines are coming, that the jobs are evaporating, that the climate is breaking, that politics is fractured, and that you are inheriting a mess that you did not create. And I understand that fear. Does that surprise you that they all boo? Not really. Booing AI. I mean, it's amazing from a generation that has embraced technology. I think it's a clear case of Turkey's voting against Christmas. I think they can see what's coming. And if we look at employment and youth unemployment, as Stephanie's touched on in the UK, is very high. And what that's meaning is that when the cost of employing people is expensive, the cost of employing young people who are unskilled and untrained at the corporate and enterprise level, you're going to choose potentially AI to fill those junior roles, to do that research, to do that analysis. And the students, as you say, have been using that to help them in their homeworks, to help them in their dissertations. They know how valuable it can be. Once they get into that workplace, they're finding that there isn't any jobs for them, because it's being supplanted by AI, or the cost of employing people at that level is just too high. We have an audience question. Keep your emails coming and your questions, aidecoded at bbc.co.uk. This one's from Patrick in Minnesota. On a similar theme, he says, how do we reconcile AI optimism with the realities of current policies, which frequently prioritise economic growth over human enrichment? So I would love to take this first. Loving the beautiful state of Minnesota, so thank you very much for the question. We have to ask ourselves, I think, what does good AI look like? I don't think we would want to go back, just like we can't not use the internet anymore. We can't un-use AI. It's here. We want AI in areas where it's really appropriate, and I'm thinking specifically of places like in healthcare. It's helping doctors, it's helping nurses, and that helps patients. None of us would want that to go away. If anything, we want to go really hard on that. Also, because healthcare is a highly regulated sector, so if you're using AI, it has to be safe and accountable. Where we perhaps need to revisit this entire question is about the social contract. So if all of the money is going to very few companies, which don't represent most, frankly, of the economy, it's a very, very small thing. It's decoupled from the wider economy. That is not helpful for the wider society. If it's affecting young people, if it's causing AI psychosis, if it is leading to greater misinformation and disinformation, which interferes with our democracies, then we need to crack down on it. So the question is, are our existing laws sufficient and fit for purpose and just need to be reinforced, or do we need new legislation and regulation? It's worth saying, Nina, that that negative AI sentiment is a real liability for these AI labs because they need data centres to run their models. Those data centres need to be built somewhere in communities connected to power grids. And suddenly, because of the negativity, communities are starting to fight back, so planning applications are being blocked. And ultimately, this is a restriction, is it not, on compute power. So they have to win this PR battle with the public, otherwise they're doomed. Yeah, but I think a good contrast to think about, you know, NVIDIA's CEO, Jensen Huang, also did a commencement speech. And of course, he talked about AI a lot during his speech, and he wasn't booed. I think, you know, if you look at the positioning of his speech, he represented AI more as an opportunity, but also as a change in how future work will be different from what we got used to in the past. So he mentioned, you know, plumbers, electricians, those are new careers young people could consider. But, you know, he wasn't focused on the aspect of AI replacing entry-level jobs. So you're right, and I agree with Stephanie a lot on, you know, her points about regulation. The technology is indeed moving so fast beyond, you know, the society and regulators and governments can prepare for the very significant challenges and risks. So it's definitely not, you know, it's the same challenge for both China and the US. Even though China has implemented a lot of regulation, it faces the same fundamental challenge that there's no answer for. Well, compute power and data centers is precisely what brings us to our second story, because thousands of miles away from Oakland in Beijing, two countries have been fighting over exactly that. Trump and Xi met in Beijing alongside Jensen Huang, who Nina was just mentioning, the CEO of NVIDIA. And Trump has approved limited sales of NVIDIA's H200 chips to 10 Chinese firms. On Air Force One on the way home, Trump told reporters that he had discussed, and I quote, the guardrails for AI with President Xi. Sir, can I ask you two on technology? First, the computer chips, advanced computer chips, like the H200s that Jensen has, did that come up at all with Xi? It didn't come up, but as you know, Jensen was there, he's an amazing guy, NVIDIA. And he would be inclined, you know, they have a much higher level than the H200, but the H200 is good. China needs it. And so, yeah, it came up. They haven't bought it. We haven't bought it, no, because they chose not to. They want to try and develop their own. But it did come up, and I think something could happen on that. And on artificial intelligence? We have quite a bit higher level, but the H200s are good. It's a great chip. And then on artificial intelligence, what is it that you want to get done? Did you bring up? We did. We talked, you know, we're leading by a lot, but they're second and they're very strong. And we talked about possibly working together for guardrails. And what kind of guardrails? Standard guardrails that we talk about all the time. AI is fantastic. So many things can happen in terms of health, medicine, operations, everything, military, and so many things can happen. But it's also got some drawbacks. Nina, what's actually going on here? Well, what's going on is that Washington still lives under the delusion that it can cripple China's AI development simply by restricting access to advanced chips. But the moment China decided to build its own self-reliant AI tech stack, you know, including its own chips, software, clouds, infrastructure, and supply chains, that ship has already sailed and there's no turning back for China. And the fact that China looks likely to not buy the approved H200 chips anyway shows that Beijing is fully dedicated to a future not dependent on NVIDIA or the American AI ecosystem. Yeah, that's really important, isn't it? Because Trump approved the H200 chip sales in January to these Chinese firms, Alibaba, Tencent, ByteDance, JD.com, Lenovo, and yet nothing's been shipped. I mean, the Chinese said no thanks, and they're pushing all the effort towards their own domestic chip manufacturing. Yeah, for sure. I mean, you know, you can think about, actually, the US restrictions on Chinese technology really began escalating during Trump's first term around 2018. So Beijing has had years of warning that dependence on American technology could become a strategic vulnerability. So those restrictions absolutely hurt Chinese tech companies. You know, let's not forget, Huawei was once one of the world's dominant AI infrastructure companies, and its smartphone chips were globally competitive before US sanctions suddenly cut off access to advanced manufacturing. You know, so from Beijing's perspective, the lesson was very clear. Any critical dependence on foreign technology could eventually become a liability. So that's why China has spent years aggressively pursuing supply chain independence in chips, AI, infrastructure, and software ecosystem. So, you know, this chip has sailed, and there's definitely no going back. Can we talk about the chips? I mean, they are supplying the Americans with the H100 and the Blackwell series, which are the sort of lesser model. What does that actually mean in practice for China's ability to compete? Because the H200 is commonly perceived around the world as the best chip, which they don't have. So does America still have an edge in development? Well, I think America has the compute capacity from those chips, which enables it to train the large language models on the necessary amounts of compute power that it requires. But I think exactly as Nina's mentioned, what China is very good at, is they're very good at working with design constraints. And when there have been sort of constraints on exports, for example, they will work within what constraints they have. And as we saw with DeepSeq. DeepSeq was a really good example. When you don't have the compute capacity, then you find other ways of being efficient. You find optimizations, you find new algorithms. And so to be able to achieve a similar or comparative outcome with US counterparts on much less capable hardware is something that China has been incredibly good at over the years. And with their native capability that they're now building, that's only going to improve with time. So I think Trump has found himself in a difficult situation, really, where they've constrained access to chips, but it hasn't stopped China progressing. And now America wants the benefit of selling those chips in without giving them too much of a head start over the US. So they find themselves caught between two worlds. And presumably, Jensen Huang is on for the ride, because he could be selling chips to China and making a fortune. And he's not. It's win-win for him either way. And there's also something... How do you mean, win-win either way? Well, if he's able to sell fully and freely into China, he would be delighted. But if he isn't, and can't do it, then he gets to have the sort of prestige of saying, we have the leading chips. China doesn't get access to them. If you want to have top chips, top computing, you have to buy American. The real threat moment to his business will be when China is able to produce a chip that is as good or better. The Democrats... So just to add on to Stephanie's point, NVIDIA has a lot to lose, and America has a lot to lose if China indeed builds its own AI tech stack. The mistake many people make is that thinking about AI is only about who has the best single chip. It's not. Formula One isn't won by the best single engine. It's won by the entire system. Software, coordination, networking, and skill. So when you look at DeepSeek v4, which is already running on Huawei's Ascent platform and moving away from the NVIDIA's CUDA ecosystem already, that's a major milestone and symbolic moment. So first time a world-leading AI model running on not NVIDIA chips, not CUDA ecosystem. I think NVIDIA is actually facing potential, this breaking point of its very, very dominant ecosystem. CUDA has 85% market share in terms of AI ecosystem. So that is under threat right now. Well, that's all we have time for. Daniel Flack, Nina Chang, Dr. Stephanie Herr, thank you all very much indeed. If you have any questions or thoughts on the programme, then you can email us as I say at aidecoded at bbc.co.uk. On screen is the QR code that will take you to the back catalogue on YouTube. Every episode is there on the AIDecoded YouTube playlist, and it's also available on iPlayer. Do take a look at that. Thank you for watching. We'll see you next week.