Vinod Khosla, an early investor of OpenAI and the founder of Khosla Ventures, says slowing AI development could hand China a strategic advantage. He joined Bloomberg Open Interest to argue that America should accelerate AI and energy infrastructure while protecting workers displaced by automation, not their individual jobs. He also says Silicon Valley has failed to clearly communicate AI’s potential benefits in healthcare, education and economic growth.
See omnystudio.com/listener (https://omnystudio.com/listener) for privacy information.
Bloomberg Talks curates top interviews from around Bl...
What Vinod Khosla said
Written from the verified transcript and checked against it. Every figure links to the moment it was said.
Vinod Khosla said the bigger risk than AI itself is AI in the hands of nation-state adversaries, and he disagreed with slowing AI development for safety, arguing the focus should be on competing with China. He said permitting delays slow model deployment but not development, and predicted fusion from Commonwealth Fusion and geothermal from Mazama will help by 2028, with a short-term power problem solved by pricing. Khosla argued for protecting people impacted by job loss rather than protecting jobs, supported leveling capital gains and ordinary income tax rates, and proposed a $400 billion annual deficit-neutral program to support displaced workers. He said AI leaders signaling safety concerns sent the wrong message, and he made his first quantum computing investment this year, expecting Shor's algorithm to be solved before 2030. He also discussed cheaper climate technologies and his purchase of the Seahawks as an AI hedge.
Key takeaways
Khosla said AI in adversary hands is more dangerous than AI itself, and the US must focus on competing with China.
He proposed a $400 billion annual deficit-neutral program to support displaced workers.
He made his first quantum computing investment, predicting Shor's algorithm before 2030.
He said fusion and geothermal will solve power constraints by 2028, with short-term pricing fixes.
He called AI leaders' safety slowdown messaging a mistake, arguing it distracts from the China race.
Numbers and commitments
Figure
What it refers to
Type
At
$400 billion
annual program to support displaced workers, deficit neutral
4:18Can you think about sort of how the political argument could be made to the populace that this is an appropriate thing?
6:15How fearful are you that we could undergo a structural shift that impacts the very project of capitalism itself?
7:26What else is going on sort of with the guys you talk to at your level?
8:46Do you see any political leaders who you think are willing to take up the mantle?
9:20If it was a mistake to have these AI leaders from Dario to Sam to Elon say, 'Okay, maybe there is a safety issue and we need to slow down.' Was that the wrong message to be sending to the public?
10:18Where are you on quantum computing and should we expect developments there anytime soon?
12:19Why buy the Seahawks among all the many things you're already doing?
We have a choice in front of us. AI can either be more like a new renaissance of creativity and discovery or more like a new industrial revolution of upheaval and disarray. As AI becomes more capable, people must remain at the center of AI decision-making. Alignment is not an abstract research question. It is the work of ensuring that these systems reliably remain under human control, reflect human values, and help people guide the next stages of development.
OpenAI's Sam Altman speaking at the United Nations yesterday. This morning, President Trump posting that AI will be a big topic of conversation with China and that both countries want to quote leave it exactly where it is. Let's bring in an early investor of OpenAI, Vinod Khosla, founder of Khosla Ventures, a veteran in all things tech and investing. Vinod, great to have you in. And look, we have so much to discuss today, but if we could just start with that because I have to say you've been really prescient on this. As early as 2023, you were basically saying the US can't slow down because we have this concern with what's happening in China. Your man Sam Altman has said he's open to the idea of maybe slowing things down for the name of safety. I wonder what conversations you've had with him and how that is developing at this point.
Well, without going into specific conversations, I might diverge a little. I think much more dangerous than AI is AI in the hands of our nation-state adversaries. So we should be very clear that's way more dangerous when there's bad intent behind it than if it's on its own.
So can I jump in on that because this seems to be where the problem is going to be going to 2030. So we're off on a good start with sort of the model front but the power problem here. Now, China put up about five times as much power as we did in 2025. They have this 60 gigawatt, you know, dam coming in, which is just a massive impressive project. Are we going to win sort of on the models and maybe even the governance, but lose sort of in the power story eventually?
Well, permitting and other delays are clearly slowing us down, but I do think it doesn't hurt the rate of model development. It does hurt the rate of deployment of models for inferencing. I'm less concerned about power in the early 30s and beyond than I am today because the power demand was growing very, very slowly. In fact, slightly less than GDP the last 20 years and it surprised everybody. It shouldn't have been a surprise. We invested in Commonwealth Fusion in 2018, the same year we invested in OpenAI. That was 2018. But it takes 10 years to develop a technology and I'm pretty certain fusion will come out very well and we have new sources, super hot geothermal like Mazama will come out very, very quickly, my bet is by 2028. And so there are alternatives. We have new investment in a company called Subcritical you'll hear more about soon. So there are alternatives coming but power plants take a while to build and they respond slowly. So, there is a short-term problem and that'll be solved with pricing.
Does that short-term problem to Mike's point though make us all the more vulnerable for our adversaries to speed up ahead of us? If what's growing in the United States, at least for now, is this NIMBYism and this push back to data centers getting constructed?
Well, this is one of the places I agree with President Trump on is we need to push ahead for development of AI and power. I think in the end it will help the country tremendously but there is a short-term mismatch.
I think some states can take the lead here and not be subject to as much NIMBYism. They have less population density. I do think the data center providers should subsidize rates where they are. So that's very easy to do because they have room in the economics to do that.
So I'm a big fan of this fourth turning concept. So every 60, 80 years you get sort of a big pivot in the politics. Now you have been talking about sort of taxing capital and that's important with AI taking maybe more jobs disrupting industry. Maybe there's a transition period where we do create more jobs but then we have to go through this kind of disruptive period. Can you think about sort of how the political argument could be made to the populace that this is an appropriate thing? We just had tax cuts that seem very popular on one side of the aisle and then we have maybe more extreme views on the other that won't be adopted by the center.
Well, there's two entirely different questions. One is deployment and I just talked about many sources of power. Commonwealth for fusion, real fusion we're doing, Mazama for bio-geothermal energy, a new distributed source of power coming online. You'll hear soon about, I can't talk about yet.
