Siemens & Amazon: The Strategy Behind Industrial AI

Thank you Nina.

Yeah, what we usually say is that consumer AI makes the headlines but industrial AI makes the impact. What we mean by that is if you think about how we use large language models in industry, we can significantly add value to customers by accelerating innovations, by delivering products much more cheaply, and by bringing more intelligence into the service we provide. We believe that if you take industrial AI into the physical world, and there's a lot of talk about physical AI, this is exactly what's happening. So this is where large language models are being deployed in an industrial setting and thereby making industry more efficient.

Yeah. So you mean how has the relationship created value together with Amazon and Siemens? Let me give you two great examples. First, on the factory floor, we are using the industrial edge, and thanks to the hosting we have, we are now able to do this 80% faster than previously. So this is a great example of how a long relationship was the nucleus for accelerating and taking the latest technologies. Another one is using Amazon's large language model to increase our searches on our website significantly. We improved to three times, which is fantastic, so customers can find products much faster. So these are just two small examples, and there are many more to come as we work through this great partnership. And lastly, to answer your question specifically on customer zero: similar to Amazon, we are big believers in trying it out yourself first, being your biggest critic, because then you can improve before it hits the market and you have a reference. So culturally we are very well aligned on how we get new innovations together to the market.

Well, the biggest impact is everywhere where we have the most data, because AI helps us to do this in an automated way where we have an explosion of data. Areas like pharma life sciences, where we are just about to start to gather all the data and then make sense of it by applying algorithms. So AI is going to be a great help because sometimes we don't understand the mechanism, but we can observe what's going in and what's going out, and AI is really good at articulating that. So that's one. Then the other one is the 30 industries that we serve today as Siemens, in particular anywhere where there are manufacturing challenges, be it automotive, machinery. This is where, because of IoT and Industry 4.0, we have all this access to data, and we have the ability to get more intelligence into their operations.

Yeah, happy to. What we're seeing already now is that we have a lot of copilots coming, and we're really excited about these because copilots are lowering the entry threshold for engineers. When they use requirements tools, design tools, or lifecycle management tools, very often there is a training effect. By using copilots, we can guide new engineers and make them productive much faster. So we democratize the knowledge to get started. You're going to see this now; we have multiple copilots embedded along the entire value chain from requirements to lifecycle management to design to simulation and operations. You'll see it across the entire line, which is really cool because it accelerates the conversation. That's where we are today. Now what we think is that the true intelligence in industry comes from industrial data, and industrial data is very different from most AI, because GenAI is trained on language, words. In industry, you have requirements where words are used, but then you go into technical details, drawings, calculations, data from manufacturing, etc. We realize we can't just take existing models to improve the entire design process from cradle to grave. So we are starting to build an industrial foundation model specifically trained on industrial data, which you can't just download from the internet. This requires building alliances, bringing it by industry for industry, developing it for them, understanding the use cases, getting the requirements data, design data, production data, and building a generalizable model that will significantly accelerate – let's say cut in half – the time to develop new things. We think that's really exciting for industry because that's exactly what everybody needs: faster time to market, more flexibility, and more tailored products for customers. AI gives us much more flexibility than we used to have.

The difference really is that the copilot is for the engineer to understand how to use the tools. Copilots guide you, similar to having someone help you draft an email. That's what we do with copilots. What we're talking about with the foundation model is to capture the intent of the engineer when they design or calculate, and use that to accelerate the design process. For example, if you want to develop a new cup of water, you would say 'develop that for me, and it should be 1 or 2 ounces', and it would already define models to understand the size, shape, material, and whether it can be manufactured. With that, you significantly reduce time to market.

This is something we're working on right now, but it requires a lot of data to make it work. So we're going to see this happening in the key industries today that are under pressure to significantly accelerate, such as automotive, machinery, or defense, because defense is a key industry that has to ramp up very quickly. So everywhere there's time compression, that's where we'll see it first.

Maybe Marty, you want to start?

Absolutely, happy to. I'd say three things. Number one is trust. This has been built over many years, many teams have come to work together, and trust only develops over time. So it's a very trustful relationship, a beautiful platform to expand. The second is the cultural element. What I love is that Amazon is so customer-focused, which is great because at Siemens we also believe in customer zero and trying it out yourself first, having reference customers like PepsiCo to see how it works before scaling. That customer obsession is something we also learn from Amazon, which is fantastic. I should also mention the engineering culture; we have brilliant engineers and so does Amazon. We are always thinking in terms of first principles to solve problems. Marty was talking about LEO internet connectivity, and that is fantastic because it's exactly how we think – many engineering hurdles had to be overcome. Thirdly, it's very complementary in terms of the tech stack. Amazon provides very horizontal technologies; it wouldn't make sense for Siemens to build cloud infrastructure, and Amazon wouldn't build very specific industry models. Amazon looks at the horizontal level. So we combine the horizontal technologies from Amazon with the vertical specific applications for the 30 verticals that we serve as Siemens, in R&D, manufacturing, etc. This is where the win-win comes together.

Well, generally speaking, it's all about doing more with less. We are living in a world that is facing a lot of challenges when it comes to geopolitics or the conversation about global warming. So we have to think about how to provide better energy to customers in a reliable way with renewable energies coming into the system. How do we address the challenge of data centers requiring a lot of energy to run? Or in buildings, how do we address the climate and cooling aspects given that ambient temperatures are rising? The answer to this is industrial AI because we can improve the heating and ventilation within a building by 30%. We can do this in the white space of a data center. We can do this in grids where we can increase capacity without adding new copper in the ground, by using more intelligence based on the data that is there. So it's real world impact, making things much more sustainable and creating a better environment for everyone on this planet.

That's what I said. Consumer AI makes the headlines, but the true impact, not always seen, is actually happening in industry.

Absolutely. And let me add, it's fantastic to work with Amazon, to work with Marty and his team. We have dedicated resources and teams all coming together. Marty and I are both sharing the drive to move this forward. The two things right now that we're really excited about are to use a lot of insights from Amazon's logistics part, which we want to use going forward as we deliver our services to customers. So far, freight forwarding we have done quite successfully, but the next level will be real-time tracking of orders. So customers will always know – just as we as consumers know – we will also have that similar experience in industry. So there's a myriad of things we're going to do and we're going to see going forward.

Thank you.