I'm thinking we could get started on the news of the day. This AI executive order that President Trump just signed has cybersecurity components therein. What does that mean for IBM? How does it impact the way that you work with the government?
Well, first, I think this kind of hits the Goldilocks spot right now where you want to have light regulation, not overweight and not creating a big bureaucracy that can slow things down. I do believe that this is a point at which innovation has to carry on. That said, put some light guardrails in place that lets the government come and intercede when they believe and they have some evidence that something may be too dangerous. So that's that. The part that really excites us is the whole part about the clearing house and about how do you do open source patching. We can all obsess about these models finding things. Oh my god, I got all these vulnerabilities in my infrastructure. By the way, models have been finding these things way before Mythos came out. So this is not new, but it has put sort of a pin in the problem. But I think now the impetus to how do you patch, how do you improve, how do you have the clearing houses, how do you leverage open source correctly. We are big fans of that part of the approach which got far less attention than the model part of the EO.
And that part directly ties to Project Lightwall, right? This $5 billion investment. Tell us about that.
It does. So, as you can imagine, we've been in conversations with many people, mostly our clients actually, who worry deeply about this and they've been sort of saying, look, there is no place. We use a lot of open source, but there is no commercial vendor behind it. If you go upstream, they'll fix the latest version, but not the version from three years before. What do we kind of do about it? So the same tool that we worry about in this case LLMs is actually a huge part of the answer where it lets you do a lot of the work of creating the patches, testing the patches, making sure they work in the client's environment. So that is why we decided to step up and say, look, we support about 50 to 60,000 open-source packages today. That's already the case and has been for years. And we said, could we increase that by another factor of four or five? And all of these technologies gave us the ability to do that. And at least from the first dozen clients that we approached, they were super excited about it. I've never seen clients sign on so fast. In most of these cases, it was one to two days and they said, 'Yep, we want to endorse this and join it.'
And in terms of the executive order, it's voluntary. It is a 30-day pre-deployment testing window only calendar days, so potentially even shorter than 30 days. Is that enough with all of the concerns around AI safety? Should Congress make this mandatory?
I believe that right now the approach being taken is the right one. First of all, yes, it's called voluntary, but it's really hard to say no to the guy who sits behind the Oval Office desk. So yes, it's voluntary, but like not quite voluntary. Two, federal acquisition budgets have a lot of teeth in them and they can turn down and say yes it's voluntary but if you don't follow it then nobody in the federal government can acquire anything from you. That as we have found out has a lot of teeth in it. So I think there's a lot of teeth that are possible once an EO has been put into place.
Got it. You know the Trump administration has been taking equity stakes in semiconductor companies, mineral companies and now quantum companies. There's the foundry that IBM is planning to build in Albany. And we seem to be witnessing a fundamental shift in the relationship between the government and the tech industry that could potentially lead to winners and losers and depending on who is in the Oval Office. Are you comfortable with this trend?
Quite. First of all, it's not a brand new trend. If you go back to what happened to Fannie Mae and Freddie Mac, I would say that was 20 years ago. If you look at different equity stakes that have been held over time that's quite there. The question becomes on is it an activist role that is being played or is the role more one-off. We'd like to get our money back if we're giving it to you. This is not meant to be a handout or outright grant. If that's the case, then there's an appropriate vehicle to get it back. Is it debt? Well, if it's a long range bet, having debt is very hard because you got to pay back interest and that keeps growing. If you make it as an equity, but you make it a passive stake. The government is a non-active member. They simply are saying, 'Look, if you succeed like crazy, I'd like to get my fair share of it.' Just like what happened during the 08 crisis. I think that's actually a good vehicle. Personally, I would have hesitated at taking debt. As a company, we can go to the public markets and raise debt. Government money is not that useful. Equity is different where it's not got to be paid back unless you succeed and then you can go figure out how do you make it succeed and then both sides have the right motivation to make the thing work out.
And where do you see this ending? I mean do you think that it would make sense for the government to own stakes in AI companies as well if they can afford to? Do you see that happening? I think they took a stake in Intel and Intel just came out with an AI chip.
So, you can see that playing out on the infrastructure side. But I would turn around and say I don't know. I think Intel is happy like crazy because when they took the stake it was under $30 a share and they're sitting at over a hundred today. So everybody benefited along the way because it is the endorsement, the willingness to lean in and say we'd like these companies to succeed is probably worth as much as the actual equity stake and the dollar infusion.
Yeah. What would you say to critics that this is no different from the state-backed technology models that we see in China?
It is a top-down industrial policy. So in some sense if you look at that, government has always put its finger on what can succeed and not. So if you go back to the beginnings of the defense industrial complex, there were big winners. If you choose who wins a particular aircraft yes you win on that aircraft but you're also getting technology that will flow over the next 10 years into the commercial sector. If you put your finger on who can succeed as being a supplier to NASA, you're giving them risk money to get something done which after that could become commercial. By the way, that was the whole semiconductor industry. People kind of forget that. I mean, the roots of Intel in the original Apollo program. So, there's lots of these things that have always been the case. I think it being more transparent is probably a good thing, not a bad thing.
The administration's rationale for this is that taxpayers should share in the upside when the government helps finance critical technologies. So what does accountability look like for a project like the Albany Foundry? What milestones should taxpayers expect IBM to hit and what happens if those deadlines are not met?
Yeah. So the milestones are being agreed with, in this case, commerce and in there within commerce the CHIPS team which sits in an intersection of both NIST and commerce and NIST is part of commerce. So it's not like something different. So the milestones you always work out is you don't just get all the money up front. The money is doled out according to the milestones and you have to stand up the independence and they will get the return when the entity in turn has an exit of some type. Till then there is no return for them and you have hard commitments. So take this one. We were building this for ourselves so far. It was only being used for IBM quantum efforts. As part of the agreement, we have a hard benchmark that we have to offer it to all other quantum companies and they get access to the same superior technology that we have been using at the bottom. So there is actually an ecosystem creation which I don't think would have happened unless the government and commerce has stepped in and it actually creates more competition not less. I know this gets sometimes lost because it's a complicated topic but that's kind of the reality of what has happened here. Same for Intel by the way. I mean them opening up their 18A and latest technologies for all comers to use is a big benefit for the country.
So the milestones and the accountability mechanisms they are all being worked out right now at the Department of Commerce. And is there anything you can share on what those might look like at this point?
Well you had agreements on what they were that was called a letter of intent when the announcement went out right that's all being sort of formalized into contracts and into formal documents as we speak. Milestones look like how do you go get what is your efforts to go get other customers to come use it? Okay. Milestones are what is the volume and the yields because it is a complicated technology that are going to be there because you need those in order to go get the other customers. So those are the nature of the milestones that are being discussed back and forth.
