AI will change the world like no technology before it. Why am I so sure of this? The technology imitates the cognitive abilities of us humans - and at incredibly low marginal costs. As I explained in the last Blog article As I have shown, this is the pattern of every major technological change in history. Only at the beginning are jobs at risk, then we flood the world with waves of innovation. Each time, these have created more jobs than ever before.  

My favourite example of this mechanism? The aqueducts in ancient Rome. Of course, some water carriers lost their jobs at first. But when the marginal cost of fresh water was actually zero, ancient Rome was transformed. A sanitary infrastructure emerged that was unrivalled: sewers like the Cloaca Maxima, Shared toilets and up to 1000 baths just for fun.  

The new technology not only freed people from the burden of carrying water, it also improved the sanitary situation in the megalopolis. Completely new forms of water utilisation also emerged: street cleaning, household water supply, powering machines, new manufacturing processes, better irrigation in agriculture and even the use of water as an energy source. Simulated naval battles for entertainment. 

Source: https://www.wissenschaft.de/geschichte-archaeologie/verbleites-trinkwasser-im-alten-rom/

Our entire AI ecosystem must grow!  

The investment costs for the water infrastructure were huge, but every additional litre of water was practically free. The resulting improvements in sanitation, prosperity and quality of life are what the ECO Association emphasises in its latest study calls "spillover effects". Of course, the Ecommerce Association is not talking about sewers, it is talking about AI infrastructure as the backbone of future economic growth in Germany. ("Spillover effects of data centres: the backbone of the AI revolution in Germany.“)  

So the lessons from the history of ancient Rome and the research of the ECO Association are clear: we need to build AI data centres in Germany. The idea sounds simple, but proves to be a challenge in detail. In order to build data centres and generate measurable spillover effects, Germany must invest in its entire AI ecosystem. Such a system does not just consist of a few specialised engineers building server halls and equipping them with GPUs. It is about the entire value chain from physical resources (such as land, energy and water) to technology (such as GPUs, networks and cooling) to better regulation, capital and ultimately investment in people and the further development of our culture. 

From AI infrastructure to spillover effects for all  

Together with the expert group "AI infrastructure" from the AI Park, we have created an overview of the components that are part of a successful AI ecosystem:  

• People & Culture: Openness to new technologies, optimistic risk/opportunity assessment, dealing with ambiguity, culture of experimentation, speed of adoption. 

• Regulation: Sufficient clarity of regulation, scope and incentives for innovation, acceptable risk for investors, coherence of federal levels, coherence of regulatory objectives. 

• Technology: Access to relevant data, AI-using applications, access to basic models, availability of chips, availability of data centres, reliable networks. 

• Physical resources: Space for data centres, access to sufficient energy, access to cooling resources, access to the data network, access to the heating network.  
  
• Capital: Venture capital for the start-up and growth phase of start-ups, public and private investment in AI infrastructure, incentives for research and development. 

1: People remain at the centre - even with AI  

Even when we talk about artificial intelligence, people and their interaction with each other remain at the centre of the AI ecosystem. Specifically, we need the following types of people in particular: 

• AI experts: Researchers, data scientists, data engineers, software engineers, experts in chips, hardware, networks, security and more. 

• Ecosystem experts: Qualified experts to support the core AI experts in the areas of law, energy infrastructure, building management, education and more. 

• Leadership personalitiesManagers, investors and political decision-makers who enable and promote the development of the AI ecosystem. 

• Consumers: End consumers and companies that demand AI services, use them, create competition, provide feedback and guidance. 

However, the success of the AI ecosystem also depends on how these people interact with each other. When a new technology says "Hello World", different cultures react differently. Some immediately see the risks, others focus more on the opportunities. Some sit down and think, others get to work and experiment. Some focus on themselves and run, others think about their environment and what 'running' would mean for them.  

A successful ecosystem with an overall positive contribution has the following characteristics: 

• Openness and optimism: Openness towards new technologies and the associated changes. The conviction that society will benefit from new technologies and that it will be able to manage the risks. 

• Agility and speed: The willingness of decision-makers and engineers to experiment, learn from mistakes and then continue to work on progress. 
  
• Ethics and inclusion: Ability to include all relevant interest groups and their different views in the overall process. Ethically sound decisions and strong social institutions. 

What would be the greatest social lever to further develop Germany's AI ecosystem? We need an opportunity-orientated mindset.  

Throughout history, our country has always benefited from new technologies. Just as the industrial revolution has changed the skies over the Ruhr region temporarily darkenedAI will also have undesirable effects. The answer back then was: intelligent regulation, new technologies (filters) and ultimately further structural change with a better quality of life.  

We should not believe that the dark side of artificial intelligence will not reach us just because we wait and see. 

2: Making the potpourri of regulation manageable 

Even before the EU introduced its AI law, there were a number of regulations that affected the AI ecosystem in Germany. Data protection regulates which data may be used for which purpose. Companies from critical infrastructures such as banks or energy suppliers are subject to special regulations regarding their handling of IT. There are export controls for hardware and software, there are regulations on optimising the energy consumption of data centres, there is building law in general, there are municipal building regulations in particular and there are regulations on cyber security. 

All of this is done with good intentions. AI models should behave responsibly, AI infrastructures should be safe for everyone. Emissions should be low, heat should be reused and data centre buildings should not disturb their neighbours. But what happens if heat is to be used to heat the houses in the neighbourhood, but the residents don't want an industrial data centre nearby? Even worse: what happens if energy is only available in the north German countryside (because of the wind), but there are too few houses to heat there? 

The greatest regulatory lever for a flourishing AI ecosystem in Germany would be an intelligent, clear and coherent set of rules that is easy for investors to handle and offers them clear incentives.  

A positive example? When the BMWi decided to amend its regulations for solar installations on balconies ("Balcony power stations"), it triggered, unsurprisingly, a Unprecedented growth of this sector.   

3: The technology sector lacks secure access to chips and AI models 

Technology is a broad term. When we talk about a healthy AI ecosystem, it refers specifically to: 

• Data: High-quality, diverse and representative data sets, data governance and compliance frameworks. 

• Applications: Industry-specific AI solutions, universal AI tools and platforms, modelling and monitoring systems, continuous integration and deployment for AI. 

• AI models:  Pre-built large language models and other base models, APIs and interfaces for model access and fine-tuning, frameworks and libraries for machine learning, AutoML and tools for model optimisation. 

• Chips: Specialised AI hardware (e.g. GPUs, TPUs, NPUs), edge computing devices for on-device AI.  

• Data centres: Cloud computing infrastructures, HPC (High Performance Computing) facilities, specialised AI infrastructures. 

• Data networks: High-speed Internet connections, 5G and future network technologies (for Edge) 

• Actors in the ecosystem: Partnerships between science and industry, open source AI communities and initiatives, ecosystems for the shared use and monetisation of AI models and applications, collaborative platforms for AI development. 
  
• Cybersecurity: Robust security measures to protect AI systems and data, secure enclaves for sensitive AI workloads. 

Germany has a high level of expertise in many areas of technology. However, in two areas that are particularly important for the AI ecosystem, we cannot keep up with the world's leading companies: 

• Software Scale-Ups: There is no German company that can keep up with American providers of large language models such as OpenAI, Meta, Google and X. We lack the large software companies that integrate these models into their standard services and provide them with business models (e.g. Microsoft365 or GitHub Copilot). 

• Chips value chain: Some German companies are particularly strong in semiconductors for automotive and industrial applications (such as Bosch). Nevertheless, our contribution to the high-end chip value chain is rather small (e.g. Zeiss). 

Both levers would have a significant positive impact on our AI landscape. More companies with their own AI models would significantly increase the demand for AI infrastructure. A greater contribution to the global semiconductor value chain, on the other hand, would significantly improve Europe's negotiating position in the geopolitical poker for chips.   

4: AI has a lot to do with physical resources 

Many components of an AI ecosystem are virtual: algorithms, models, services, regulation, cybersecurity, to name but a few. However, the foundation for all of this is still physical:  

• Location: Space for a few but very large data centres, space for substations, distributed data centres for better latency and redundancy. 

• Facility management: The physical construction of the data centres, resources and knowledge for their maintenance. 

• energy resources: Access to sufficient and reliable energy supply, renewable energy sources for sustainability. 

• Cooling systems: Efficient cooling solutions for data centres, potential for liquid cooling technologies. 

• Network access: High-speed, low-latency data networks, robust internet backbone and international connectivity. 

• Heat utilisation: Access to the heating network for the reutilisation of waste heat, integration of district heating systems. 
  
• Security: Physical security of data centres, fire alarm and security technology through to intelligent building functions such as energy management, room control, bookings and access management. 

The biggest positive impact for the AI ecosystem in Germany would be greater availability of clean electricity. 

5: Money makes the world go round 

Another factor has a catalytic effect on almost all the other components mentioned: capital. Financial investment is required in almost all parts of the AI ecosystem: 

• Basic research: Supporting research in the fields of semiconductors, algorithms, power generation, energy grids and more. 

• Education: Training and further education in practically all sectors and consumers. 

• Venture capital: seed capital for innovations and series A/B/C financing to enable global growth. 

• Transformation Fund: Supporting critical but already existing industries with funding to reorganise their business models without major negative effects on employees. 
  
• Strategic subsidiesGeopolitically relevant industries are often heavily subsidised by other continents and countries (e.g. solar in China). In these cases, additional financial support is required to keep national players alive. 

The following levers would be particularly capital-intensive, but effective: 

• Energy supply: Creation of new, environmentally friendly yet reliable energy sources to supply the AI infrastructure. 

• Semiconductors: Strengthening the European contribution to the semiconductor value chain, particularly in the high-end segment. 
  
• Software start-ups: Promotion of software start-ups in Germany with a focus on globally competitive business models (with integrated AI). 

7 measures on the way to a flourishing AI ecosystem 

The ancient Romans built aqueducts, and the result was greater productivity and prosperity. How can we create precisely these spillover effects with AI, the technology of our time? I see five important screws that we should turn: 

1. Expansion of the energy supply: Increasing clean energy generation to meet the growing demand for AI infrastructure and data centres. 

2. More AI software scale-ups: Promoting an innovation ecosystem from which AI start-ups regularly grow into large, global billion-dollar companies.  

3. Opportunity-orientated thinking: Communicative and content-related focus on the opportunities of AI, rather than on possible risks that have not yet materialised.   

4. More semiconductor value creation: Invest in domestic semiconductor design and manufacturing capabilities to reduce dependence on foreign suppliers and support the development of AI hardware. 

5. Simplification of regulation: Deregulation and more coherence along the entire AI value chain.  

The chart above shows the 5 suggestions and reveals two further components of success: 

6. Capital: The scaling of software companies, the expansion of energy supply and a greater contribution to the semiconductor industry require enormous access to capital.  

7. Commitment and discipline: Expanding the energy supply, making a greater contribution to the semiconductor industry, changing our investment culture and simplifying regulations require passion and discipline. 

To summarise, Germany is at a crucial point in the development of its AI ecosystem. By focussing on energy infrastructure, software innovation, semiconductor production, provision of capital, deregulation, cultural change and discipline, we can become a leading global player in the field of artificial intelligence.