# Eric Schmidt’s Vision: The New Reality of AI & Drones.

*ExpovistaTV* · https://www.youtube.com/watch?v=i97WyPOqJE4

## Summary

This recording features Eric Schmidt speaking at a satellite event of the Munich Security Forum, arguing that cheap drones and AI-driven autonomy are transforming warfare, as demonstrated by Ukraine. He urges governments to test spending a fraction of budgets on drones and unmanned systems, explains why abundance, autonomy and AI dominance will define future conflict, and warns that Europe is falling behind the US and China. He also discusses open-source models, the agentic revolution, energy needs for AI, and deterrence through unmodelable algorithms.

## Key points

- Eric Schmidt argues that cheap drones are decisive in modern warfare: a 2 kg drone costing around $5,000 can destroy a US tank worth $5 to 30 million.
- He advises governments to spend 10% of budgets on new technologies like unmanned systems and run trials against the traditional 90%.
- Schmidt identifies abundance, autonomy and AI dominance as the pillars of future warfare, driven by consumer product lines and contract manufacturing rather than traditional procurement.
- He says every successful future war plan will have AI as its strategic and targeting base, and algorithmic warfare will exist as a category within two or three years.
- AI-driven coding gains will produce an enormous number of zero-day cyber attacks, so military AI systems and models must be hardened and protected.
- The US is investing roughly $2 trillion in data centers with an estimated 90 gigawatts of additional power needed, while China spreads influence through open-source models.
- Schmidt calls it a tragedy that Europe lacks scale, data centers and financing, and urges Europe to build its own open-source models at a cost of 10 to 20 billion.
- Because European energy costs are high, AI training will likely happen elsewhere, which Schmidt sees as a national security issue for Europe.

## Notes

### Cheap drones and modern warfare
- At the Ukrainian pavilion at CES, a kamikaze drone was shown that flies one way up to 100 km with a 4 kilo payload, costing less than $2,000 in Ukraine thanks to tax exemptions.
- Eric Schmidt points out that a 2 kg drone costing around $5,000 can take out a US tank worth $5 to 30 million, and tanks are hard to defend.

### Schmidt's procurement advice
- Take 10% of defense budgets and put it on new technology, spending 90% on traditional systems, and run a trial comparing them.
- For every tank bought, build an unmanned ground vehicle; for every airplane, build a thousand drones.
- Embed tech people in every layer of ministries to avoid missing a once-in-a-hundred-year transition, analogous to the invention of the tank 100 years ago.

### Three pillars of future war
- Abundance: drive prices down by a factor of 10 using consumer product lines and contract manufacturers rather than military procurement.
- Autonomy: modern vision systems allow automatic target detection; systems Schmidt watched in Ukraine had zero collateral damage.
- AI dominance: strategy and algorithmic warfare using these tools is the future; every successful war plan will have AI as its strategic and targeting base.

### Cyber and geopolitics
- AI coding gains (Claude Code, Copilot) mean an enormous number of zero-day cyber attacks; military AI systems and models need hardening against adversarial attacks.
- The US is deploying about $2 trillion of capital into data centers, needing an estimated 90 additional gigawatts; a nuclear plant is around 1.5 gigawatt.
- China releases open-source weights, so much of the world will use Chinese models; Europe lacks scale, data centers and financing, which Schmidt calls a tragedy.

### Europe's role
- Europe should build its own open-source models (costing 10 to 20 billion) and organize a group of 10 European computer scientists aligned with the British.
- Training should be done in a European NATO-type country with European values and privacy laws.
- Helsing in Germany is cited as a good national champion; Europe should lead the agentic revolution.

### Deterrence and timelines
- Algorithms cannot be counted like nuclear weapons, and opponents with different strategies are hard to model, making algorithmic warfare self-deterring.
- Schmidt expects algorithmic warfare as a category within two or three years.
- The "San Francisco consensus" expects major AI breakthroughs in 2 to 3 years; Schmidt doubts 2 years but expects huge impact over 5 to 10 years.

## Transcript

[00:05] Every day for the last 3 years, we have seen dramatic images from the war in Ukraine. Periodically, we also see how Ukraine is defending itself by developing new tactics, but also new tools. Drones, in particular, are frequently pointed out as the protagonist of the new warfare. We had a live demonstration of this technology at the Ukrainian pavilion during CES. The drones on display were very cheap to produce in mass quantity but also very deadly. >> This is also military drone and we are right now looking know how to use it in a in a civil in the civil sphere. >> It's actually a kamicazi drone. >> Oh wow. So the idea of this drone was to develop super super cheap and effective comma because you need like enormous

[00:50] amount of them you know on the front line because it flies only one way >> right >> and it's really really cheap and it can fly up to 100 km with 4 kilo of payload >> how cheap >> in Ukraine less than 200 $2,000 uh in Ukraine abroad it would be more expensive obviously because you know we have some like our country like help um helps companies like us in we're not paying taxes and so on. >> Everybody will be a bit more expensive but still not 20,000 like we see like for example >> in fact Eric Schmid is suggesting that every government should consider investing in alternative modern weapons like drones alongside the traditional war machines. Take a listen. [music] is take 10% and put it on new stuff.

[01:44] Spend 90% on all the stuff that you think you should buy and run a trial. Literally test whether that 10% is better than the 90%. And for example, for every tank you buy, build a [clears throat] unmanned ground vehicle. For every airplane you buy, build a thousand drones or whatever, whatever the numbers are appropriate and run the test. uh you should be skeptical in this in your organization. If you don't have tech people like these people embedded in every layer of in in this case the chancellory and the mods and so forth of the countries you're going to miss out on a once in aundundredyear transition. This transition is analogous to the invention of the tank 100 years ago. And by the way, tanks are not a good place

[02:34] to be in right now. 2 kg drone costs what $5,000 takes out 5 million to30 million in the case of US tanks. Boom. Just like that. And they're very hard to defend against. >> Eric Schmidt was a guest during one of the satellite events of the Munich Security Forum. And he has no doubt that what we are seeing in Ukraine has many elements of the warfare of the future wherever this may happen. >> [music] >> Um, I spent lots of time in Ukraine trying to understand the future of war and I think I I understand what's going to happen. Um, and I'm not talking about Ukraine specifically. I'm talking about combat. Um, all of us are familiar with tanks, boats, and air force jets of one kind or another. All of you command some

[03:18] combination of them. In every case, there's going to be an automation equivalent of that. Instead of buying tanks, you'll be buying unmanounded ground vehicles. And I'll describe why in a sec. Instead of buying big ships, you'll be buying the equivalent of drone ships automated in various ways. And in terms of t and in terms of air force, you'll have drones. And the drones will be attack drones, but the drones will also be able to defend themselves. This is the future of war. And of course the existing systems be around but the quicker you can make this transition the smarter and more sophisticated your national security approaches will be. The elements that are composed of this are first abundance. I come from a

[04:03] background of abundance. You all spend a time of scarcity and ex exquisite expensive things. Uh in my view that's wrong. Sorry to be obnoxious but I'm an American. um it's much better to say how do we solve this problem with abundance with inexpensive systems that work well in swarms and that technology now works. So the first thing is drive the prices down. And when I say down, I mean down by a factor of 10, not 10%. And you do that by using product lines of one kind or another. You use um consumer product lines, you use contract manufacturers. You don't use the traditional military procurement. The second thing that you do is um think of it as autom automation, right? Autonomy. Um, so for example,

[04:52] with modern vision systems, you can have automatic target detection, and so far the systems that I've watched in Ukraine have had zero collateral damage. These things are accurate enough that they're more accurate than humans, which is important in in com in combat and in national security. And the third one is AI dominance. And AI dominance can be interpreted as many things. It can be interpreted as vision systems. What I'm talking about is in strategy and algorithmic warfare using all of these tools is the future. It's how war will be done. And obviously having spent a fair amount of time watching this real war, these are to be avoided, right? So I'm not in any way endorsing this. But the the the core point I'm trying to say

[05:36] is that this is happening very very quickly in Ukraine and all of you will either help or be dragged into it or will fight it because you don't like it. But the fact of the matter is this is the future and I'm convinced of it. Wars are won and lost through a whole bunch of players. There's no single reason why you win or lose. But every successful war plan will fundamentally have AI as its strategic base and as its targeting base. The only question is how much scale do you have in order to do this? By the way, you have to have lots of training data, right? So the key with AI based things is that you actually have to the system has to learn. Um and thank goodness there aren't that many wars, but things like how do you identify

[06:24] targets? What does the enemy do? How do they respond? All of these are playing out right now between Russia and Ukraine. In AI development, the fastest gains right now are occurring in coding and you see this with Cloud Code and Augment and Microsoft Copilot and so forth. Um, my category of jobs, the job that I did 50 years ago is going away. It's rather disturbing. Um, but the fact of the matter is that's where all the initial gains are right now for a whole bunch of reasons. If you can write code, you can also write cyber attacks. So, one of the things that's going to happen is you're gonna have an enormous number of zeroday cyber attacks coming out of this because of this innovation cycle. So, having

[07:08] said that, AI and AI servers that are used in the military are legitimate military targets and therefore need to be hardened and protected. They need to be hardened physically. They need to have reliable power. But also the models uh there are things called adversarial attacks where you can go in and take the model which is from your perspective is just a bunch of numbers and change them enough that the AI does the wrong thing. So protecting them cyber security cyber attacks are all incredibly important. I think this audience knows this is a growth business. Unfortunately when I talk about AI globally I always talk about what the US is doing and what China is doing. And the simple summary is the US, thank God that the American

[07:52] financial system will give these people billions and billions of dollars with no credible strategy for for making money at it. Um, Germans don't do that. Europeans don't do that. Um, but the fact of the matter is this enormous wave. It's $2 trillion of capital going into the data centers in America and around the world for those companies. Um the estimate for power needed in the United States is 90 gawatts of additional power and as reference if you're not familiar with this the nuclear power plant is around 1.5 gawatt gives you a sense of the scale involved and this is these are the plans in China you have the same thing except that the Chinese are using a technique called open source they're releasing all of the

[08:34] weights and models um which is important strategically because it means the majority of the world the third world if you will the other countries that we don't typically visit are going to be using Chinese models not American models whereas Europe Europe has a couple of leading champions Mstral is one I'm an investor for example but Europe does not have the scale does not have the data centers does not have the financing to do it this is a tragedy Europe should organize in my view and I've said this and I'll say it again your own open-source models and they should be done thank you Um, it makes no sense for you to to allow your intelligence systems and I mean human intelligence, not military intelligence, to be

[09:19] controlled by the US or China, right? Europe has its own special thing going and I happen to like Europe and I want you guys to be successful and I don't want you to be overrun by the Chinese or the US. You're going to have to do this and furthermore it's going to cost you 10 to20 billion. That's what they cost. Why open source? >> Um, open source is the quickest way to catch up and having hired tens of thousands of people in Europe, the engineers in Europe are fantastic, right? So, you need a way of organizing them. You need a way to move quickly and open source is the easiest way to do it. My specific proposal is that Brussels or equivalent organize a group of 10 computer scientists who are obviously European

[10:06] and align with the British because they're kind of near you um and figure it out. It can be done fairly quickly. >> The importance of open-source models is often mentioned in this type of meetings also from the leading authorities in this field. Take a listen. Let's face it, the single most important thing that has happened last year, if you were to ask me the one single most important event last year and how it has activated AI researchers here in this region, it's actually Llama 2. It's an open- source model. All were possible because of some of these open source languages. And so I think that democratizing activating every region, activating every country to join the AI advance is probably one of the most important

[10:54] thing. >> For more content about the impact of the Yama model and the importance of opensource models in general, you can watch the video at the link above. But let's carry on with the future of warfare. What I'm going to say is that in a national security perspective, first the money is showing up which is great and that money needs to be applied to the right areas specifically the algorithms lowcost a factors technically drones drone industry you have a number of national champions here you have a in Germany you have a company in Helsing called Helsing that's quite good um there are a number of such examples and we need a lot more of them the um I most interested in not the problem of volume

[11:41] which everyone is focused on because I think the volume stuff will ultimately be sorted out but rather the intelligence layer and I'll give you an example um the way AI is evolving is we start with um the AI that you all know which is sort of chat GPT language models things like that and then we move to strategy and planning which is where we are now with strategy and planning you can begin to do steps Think of it as workflow. And when you can do that, this is what militaries do. This is what businesses do. This is what universities do. Europe should lead in the agentic revolution where you build agents to solve the problems that your citizens, your military, your universities have. This is the year of the agentic

[12:24] revolution. And that stuff is happening very fast. I'm talking about every month. You saw last week all the publicity about a fake world of agents that were beginning to learn how to interact with each other. And everyone was surprised that they didn't immediately try to assassinate each other. Well, that's presumably in game three of the simulation. But the important point is that the agent model is going to drive huge revenue in America. It should drive huge revenue in Europe. Along with that, start thinking about military uses. There's all the supply chain stuff that's it. And you all spend so much money in my view wasted it on these complicated systems just to have agents that figure this

[13:04] stuff out and move it around. It's possible now using a technique called MCP which is used throughout the industry model context protocol. You basically take all your databases and you start asking it questions and it figures it out. I mean these are huge breakthroughs architecturally and Europe is not leading in these areas. They will make you much more effective in whatever your objective is military, business, profit, entertainment, education. The the key next step after that is what agents do you build for war. So most people when they talk about AI talk about it in the context of what in America is called project maven, which is essentially vision and the project maven thing was very controversial.

[13:48] Google started it with the Pentagon. Google canceled it. Um, I was on the Pentagon side, which is a mess. Um, and then eventually it's been taken over largely by a palunteer. That's not the kind of AI I'm talking about. It's an enabler for reasoning. And every general I've met talks about how they want to have battlefield awareness and they want to be able to see everything. Well, here's what I want. When I sit in a command center and I watch the commanders who are very very committed to doing the right thing, I want the computer to give them a plan, a battle plan that we can prove will win in a series of iterations. There's a technique called reinforcement learning which was invented um some years ago. It

[14:33] was heavily used by Deep Mind winning the go game if you remember that about 10 years ago. And reinforcement learning allows you to sort of prove what the optimal paths are. So here's an example. This would never happen, but you and I are opponents. You have you have algorithms and I have algorithms. So now you and you and I are going to have a um and we're we're mad at each other. What is the deterrence that keeps you and me from fighting each other? Unlike traditional weapons, you can't count the algorithms. If your algorithms are better than mine, and I suspect they would be, I better be very careful before I do something because you get a vote and you get a vote to to win. I spent a lot of time

[15:19] trying to figure out what this fight looks like. I'm sure within two or three years, algorithmic warfare as a category will exist. What I don't believe is that the algorithms will be the same. And I'll give you an example. In one of the more recent battles in Ukraine, and I won't go into the details, um it was pretty clear that the Russians were focused on land and the Ukrainians were focused on the people on the other side, right? These are different military strategies. It's not possible to model strategies where you have different goals, right? It's too hard. So the deterrence is if you and I have different strategies, I don't know what you're really going after and therefore I can't model what

[16:09] you're going to do. You become harder to predict with less prediction capability. I'm less likely to start a war. So deterrence, >> it is a form of deterrence. And and the reason I'm hammering on this is that almost everybody learned from the nuclear stuff. Sorry to as an aside. So Henry Ken and I were best best friends. We wrote two books together and uh and of course I miss him every day and he spent a lot of time here. He was by the way here at the original MSSE way back when when I think none of us were here. Um and but anyway, Henry would would tell all these stories about how he did deterrence and how the deterrence in the 1950s was erected and it was fundamentally because you could count

[16:53] things. I'm talking about a world where you don't know. You can't model what the algorithms are going to do. Algorithmic warfare should be extremely self-deterring because of that issue. >> Some of these early capabilities were already on display at the Ukrainian booth during CES. >> Run uh you know the sophisticated things like like uh AI models. Then you just fly over your fields and the system automatically makes a photos detect what what are you going you know to detect and then just collects all the statistics calculates all the statistics that you want and it depicts here on your on your screen. >> There's wishful thinking and then there's reality. The wishful thinking is that there's going to be a new algorithm

[17:40] that replaces the transformer architecture that Google invented in 2017 that uses less hardware. That is a great idea. It's it's not obvious how and when that will occur. If you look at let's use Google as an example. Google is a highly efficient answering machine. The computational costs of doing the same thing with any large language model are many orders of magnitude more expensive which is why these things take so much time and the value is there. And the part of the reason is that they do something called test time planning where they are actually learning as they're going. They literally you you you feed it what it knows and it knows something and then you feed it back in again. It's called it's essentially a

[18:24] recursive process. The important thing about all of this is it uses an enormous amount of energy. Um as I mentioned before in the United States the um it's pretty clear that every real estate developer in the United States has become obsessed with making more money from data center builds than from building building on farmland. And the US has lots of extra land. So you're going to see lots and lots of that. And this particular company um is working in an area called the Perian Basin which has a great deal of natural gas. In looking at it for Europe, Europe has uh France has excess nuclear capacity as far as I can tell and the the only problem is the prices in France for energy are high. I was recently

[19:08] contacted by somebody who's figured out a way to build a gigawatt in Bulgaria which I thought was interesting. My guess for Europeans because the energy costs are so high is that people are here. The training will be done somewhere else. This is a a national security issue for Europe as a strategic issue. If I were doing things for the military in Europe, I'd want my training to be done in a European NATO type country. European values, European laws, European privacy laws. And that's hence why I keep going on about Europe needs a strategy. You need some form of open source model and you need to do the training in Europe. >> So you're building this facility where natural gas is abundant. So you're in

[19:53] the right place for data centers. Middle East would seem to be a logical place to put some of them, but not the ones for military use. >> Yeah, I think I think that for the militaries will ultimately figure out that controlling that value chain is really important. You want to have pretty much vertical control over your stuff. >> Now before this event, we followed the World Economic Forum in Davos and it was evident that even among the leaders in AI, there was no consensus about the timeline [music] for AI to become a truly revolutionary. Eric Schmidt confirms the current situation. The real narrative is that America because of money, capital scale, craziness, people like me touting a future in two years where everything

[20:41] will be different. The AI boom in the US is like a religion, right? And I call it the San Francisco consensus. And um the people in San Francisco that I know and I spent lots of time with them all believe within 2 or 3 years something very interesting is going to happen. And one of my friends said h and he goes then we don't know what will happen. And by the way this is the way the pitch goes. When these systems these systems are smart and they can do planning and so forth, there's a point at which the the system can begin to build itself. It can take off on its own. The technical term is called rein uh re essentially recursive self-improvement recursively getting smarter and smarter. Now, if you think about it, if the

[21:33] computer evolves faster than we do, we're in trouble. And then everyone goes, "Oh my god, we're in trouble." Now I personally don't think it'll be within 2 years but the the but I do believe that over a 5 or 10 year period the rise of this foreign intelligence and how it gets hard harnessed is going to be huge. My view about Europe since we're in Europe is Europe needs to be a player there and I don't see Europe doing it right now because of the lack of hardware and companies and so forth and uh Europe is in big danger to be left out of this and China for example does understand this and the Chinese are working very very hard using a different approach. They do more computation at the edge. The computers are not as

[22:19] powerful because the US won't allow them to get those chips. Um, and that's there's a complicated discussion about all of that. I should mention that ASML, which is a European company, is incredibly impressive and incredibly important. If Europe could just create more ASMLs, everyone would be good, right? It's it's a true monopoly on a technology that is essential for the AI revolution. The these transitions take decades, right? Did anyone here arrive in a driverless car here in Munich? This is the car capital of the world, isn't it? Germany, right? By the way, in America, most people don't arrive in self-driving cars, too. And they were invented in the basically the late '9s. So, what is that 30 years, right? It

[23:05] just takes a long time. The industry uses the phrase technology overhang to refer to the fact that the rate of diffusion or adoption of the techn tech the stuff we're talking about is slow. It's slow for many reasons. Sometimes it doesn't work. Sometimes you don't have the right people on the other side. But at least in the case of national security, you have an obligation to protect your countries. You have to, in my view, take the jump, take the risk, and at least run the test. Right? To me, it's an it's a moral imperative. It's a national imperative. And is exemplified in the war. um with respect to to the rate at which this stuff is going to happen, a lot of it depends on how easy it will be for normal people to use it,

[23:49] right? So, for example, Google and I obviously have a pro Google bias. Um many of its Gemini answers are really better than the search results, right? And Gemini is really good. It's currently the best performing model, right? So, it's according to the rankings. So, good for them. Uh so the the key thing to understand here is you have this brutal competition for innovation which is being funded by trillions of dollars. We all benefit from that competition. Let's see how far it goes. Specific things that are going to happen are agents will be used by businesses to improve customer service in there's evidence that customer service jobs for example are declining because you can automate those. A

[24:30] general rule about automation is that highly um highly repetitive jobs and highly dangerous jobs are automated out first. And I think that's what you'll see. [music]