In the Markets: Germany
6 min read
2024-09-23
Torsten Kraul
Partner & Co-Head of Digital Business Practice Group, Noerr

executive summary

Gallery

No items found.
Need this in PowerPoint?
Enter your email to request the file.
Thanks! We'll send the file to your email shortly.
Something went wrong while submitting the form. Get in touch with 20Minds via info@twentyminds.com

article

Sample

How has AI investment evolved in Germany?

Germany is home to a growing number of successful AI companies, including Aleph Alpha, Brighter AI, Nyonic and Celonis. With 508 AI startups and an average funding amount of US$14.8 million, Germany is of significant interest for domestic and international investors1.

Germany’s AI ecosystem is characterised by a collaborative approach between established companies, (often publicly funded) research institutions, and startups.

Germany’s AI investments focus on applying the technology across diverse sectors: human health and social work activities, manufacturing, and transportation, mobility & storage, being the most prominent among German AI startups.

German AI Startup Landscape by Sector 2023

Are there policies to help develop the German AI sector?

Germany’s National AI Strategy, launched in 2018, outlines a multi-pronged approach to strengthen research and development, promote AI adoption across sectors, and establish ethical guidelines. Backed by a €5 billion commitment by 2025, the strategy focuses on establishing AI competence centers, funding research projects, and supporting AI talent development.

The German government provides substantial funding for AI research and development through various programs and initiatives:

  • The German Research Foundation (DFG) funds numerous AI research projects across different disciplines.
  • The Federal Ministry of Education and Research (BMBF) actively supports AI competence centers and various research initiatives. The BMBF has committed to investing €1.6 billion in AI research until 2025.
  • The Federal Ministry for Economic Affairs and Climate Action (BMWK) promotes the use of AI in industry through funding programs and collaborations.

What is Germany’s approach to AI regulation?

The EU AI Act will apply directly in Germany. As the world’s first comprehensive law for regulating artificial intelligence, the AI Act aims to establish uniform requirements for the development and use of artificial intelligence. Most of its provisions will be applicable 24 months after entry into force, i.e., from August 2026.

What structures are used to invest in the German AI sector?

The investment structures and terms used in the German AI sector generally align with those found in other German technology investments.

What are the key approval requirements for investing into the German AI sector?

Recent AI Investments in Germany

Antitrust

Transactions may require a filing with the German Federal Cartel Office (FCO) if the parties meet either the turnover or the transaction value-based thresholds. The latter is likely to gain relevance for AI transactions as it only requires substantial operations of the target, but not necessarily significant turnover.

German merger control also captures transactions well below the threshold of acquisition of control. Most notably, the acquisition of 25% or more of the capital or voting rights is notifiable, and even an acquisition of less than 25% is covered where the acquirer gains material competitive influence.

On this basis, in Microsoft/OpenAI the FCO found that Microsoft’s investments are subject to merger control. However, the FCO concluded that OpenAI – in 2023 – did not (yet) conduct substantial activities in Germany.

In Renesas/Altium (2024), the FCO ultimately cleared the acquisition of a software designer for printed circuit boards by a semiconductor manufacturer. Interestingly – and likely a benchmark for future transactions involving the AI supply chain – the FCO scrutinised the transaction, but ultimately found no horizontal overlap or other effects on competition.

The FCO can also take action against companies that have a “paramount significance for competition across markets” (Section 19a ARC). This is a high threshold that so far has only been met by a handful of companies (e.g., Meta, Amazon, Alphabet). This said, in principle this can be a tool with significant impact on AI activity, such as access to relevant data to train LLMs.

Foreign Investment Review

The ever-increasing requirements of foreign-direct investment review regulation significantly affect transactions in companies dealing with AI. The German investment screening regime explicitly addresses certain acquisitions of companies developing or manufacturing items or software that employ AI, when the AI could be used for cyber-attacks, disinformation, or surveillance for domestic repression.

Acquisitions of more than 20% of the voting rights in such a target company must be notified to the Ministry for Economic Affairs and Climate Action (BMWK). BMWK may then initiate a formal investment review.

Furthermore, there are several AI-adjacent business activities within the broader information technology sector which also may trigger notification obligations and require BMWK’s clearance if the relevant acquisition thresholds are crossed. For example, acquisitions in companies dealing with (government) communications infrastructure and data processing may cause reporting obligations, as would acquisitions of companies doing business with IT security products or their components. Recent publicly known cases of government intervention illustrate how sensitive the IT sector is for Germany, particularly if investors from China are involved. In 2022, two Chinese (and one Taiwanese) acquisitions of semiconductor manufacturing companies failed due to government intervention. Furthermore, Chinese acquisitions in communications companies were prohibited in 2020 and 2023.

Export Control & Sanctions

The dynamic geopolitical world of export controls provides another regulatory risk facing AI transactions. Most AI-relevant export controls in Germany are regulated on the EU-level via the Dual-Use Regulation. There is no explicit export control list position for AI as such; however, several export control listings will affect goods handled by companies operating with AI. For example, there are positions for “neural network integrated circuits”, “neural computers” as well as corresponding software, and lithography equipment used to manufacture relevant computer chips. The export business of companies handling those items therefore faces corresponding political risks.

Recently, the Netherlands has – incoordination with the US and Japan – made use of the option to impose stricter, unilateral controls on semiconductor manufacturing equipment than those foreseen at EU-level. These restrictions are commonly understood to be particularly targeted at exports to China. Germany, on the other hand, has not yet unilaterally tightened AI-relevant export controls – at least not on a regulatory level. Nonetheless, there is scope for Germany to apply pressure to some bottlenecks in the semiconductor supply chain, as Germany is home to a few hidden champions in machine and plant engineering which occupy quasi-monopolies in the semiconductor manufacturing supply chain. Accordingly, political pressure could be applied by tightening relevant regulations or more informally by a (tacit) change in the practice of granting export licenses.

What else to consider when making AI investments in Germany?

Germany has a data protection authority in each of its 16 states, as well as a federal data protection and information commissioner, all of whom are active regulators. Notably, there are ongoing debates about the data protection compliance of GAI, in particular large language models (LLMs). As of today, there have been few court decisions and administrative proceedings.

During the lifecycle of LLMs, numerous data processing activities take place. For example, during training, developers compile, store, and use large data sets for training runs (including fine-tuning and corresponding testing). These almost inevitably contain large amounts of personal data. Similarly, LLMs regularly process personal data during use, for example within their output data.

All data processing requires a legal justification, such as consent of the data subject or the need to process data for purposes of legitimate interests. Developers and users of AI systems, as well as investors in relevant companies, will, therefore, need to assess whether a sufficient justification exists for each data processing activity. In addition, they must comply with a wide range of transparency and other obligations. Violations can result in fines of up to €20 million or 4% of the previous year’s global annual turnover.

Torsten Kraul is partner at Noerr and the co-head of the firm’s Digital Business practice group.

Sources

  1. AppliedAI Institute 2024. For comparison, average funding for AI startups in the UK, which generally has a deeper equity market than Germany, was reportedly only around USD 11 million in 2023 (Startups.co.uk 2024).
  2. Celonis, (2024), Retrieved from: https://www.celonis.com/press/crunchbase-celonis-raises-1billion/, Last accessed 21 July 2024
  3. Helsing AI, (2024), Retrieved from: https://helsing.ai/newsroom/helsing-further-strengthens-european-defence-capabilities-with-funding-round, Last accessed 21 July 2024
  4. TechCrunch, (2024), DeepL the AI Language Translation Startup Nabs 300M on a 2B Valuation to Focus on B2B Growth
  5. Financial Times, (2023), Retrieved from: https://www.ft.com/content/cf3a1cd8-64ef-4002-add6-3342ba9facff, Last accessed 21 July 2024