8 Min. Read Time \u00b7 As of: 23.04.2026<\/p>\n
On April 22, 2026, Merck and Google Cloud announced a multi-year partnership worth up to one billion US dollars at Cloud Next in Las Vegas. The core piece is an agentic AI platform that simultaneously enters research, production, commerce, and administration. For German mid-sized companies in the pharmaceutical and chemical environment, this is less of a headline and more of a blueprint: it shows how Agentic AI is built not as a pilot, but as infrastructure. Those who read carefully now recognize the levers for their own house.<\/strong><\/p>\n What is Agentic AI?<\/strong> Agentic AI describes AI systems that independently plan multi-step tasks, make decisions, and control tools or data APIs to achieve a goal. Unlike classic chatbots, agents do not just react to prompts, but orchestrate workflows, execute intermediate steps, and document results with an audit trail. For companies, this means a step away from individual assistants towards a work level that supports specialist processes.<\/p>\n The key data points are clear. US-based Merck (operating as MSD in Europe) and Google Cloud have entered into a multi-year collaboration with an investment volume of up to one billion US dollars. The official announcement from the Merck Newsroom<\/a> mentions four areas of application: research and development, manufacturing, commercial, and corporate functions. The target architecture is based on Google Cloud’s agentic platform around Gemini Enterprise, expanded by joint engineering teams from both companies.<\/p>\n Thomas Kurian, CEO of Google Cloud, positions the deal within the Cloud Next 2026 as the first end-to-end integration of agentic AI in a global pharma value chain. This is more rhetoric, of course. But the structure behind it is sober: data connection between research and production, agents as the executing layer over enterprise systems, Gemini as the governance layer. This exact layering is the relevant lesson for mid-market architects.<\/p>\n In parallel, Google Cloud has announced a partner program worth 750 million US dollars. It addresses system integrators, software manufacturers, and data service providers that set up Agentic AI implementations for mid-market customers. For DACH companies currently discussing Agentic AI projects with service providers like Accenture, Deloitte, or T-Systems, this program represents the refinancing backdrop for consultant offers in the coming months.<\/p>\n The initial reflex of many medium-sized business CEOs when faced with such news is: that’s too big for us. And yes, the absolute investment sum has little to do with the reality of a family-owned pharmaceutical company in Baden-W\u00fcrttemberg. However, the architecture behind it is highly relevant. Three principles can be derived from the announcement that are relevant for any company with over 100 employees that wants to seriously engage with AI.<\/p>\n Firstly: Agentic AI is treated as a platform decision, not as a feature rollout. Merck is not building a copilot for marketing and a predictive maintenance case in production. Merck is building an agent layer that reaches both functional areas, with a shared data and authorization model. For medium-sized businesses, this means: architecture decision before use case selection. What identity, data, and compliance level should support the addition of a second and third agent?<\/p>\n Secondly: industry expertise is the key differentiator. Google Cloud provides models and infrastructure, while Merck brings regulated workflows and scientific data. Value creation only emerges from the interplay between the two. DACH medium-sized businesses have exactly this type of data and process expertise in their domains, but rarely showcase it. Those who fail to properly inventory their in-house domain data today are giving away the only raw material that distinguishes Agentic AI from generic chatbots.<\/p>\n Thirdly: personnel and process take center stage, not the model. According to the announcement, Google Cloud sends engineers to Merck to collaborate with Merck teams. This is not a standard managed service, but a co-creation gesture. For medium-sized businesses, this is a reminder that Agentic AI will not succeed without joint teams between technology and business units. Those who set up agents as a pure IT project will end up with a proof of concept without adoption after three months.<\/p>\n For life sciences companies in the German-speaking region, a clear picture emerges. The deal takes place in a regulatory landscape where the EU AI Act has been in effect since April 6<\/a> and the EU FMD revision is reorganizing pharma serialization<\/a>. Both topics require data governance, which an agentic AI platform inherently needs. Those who think about both in parallel save on double investments.<\/p>\n The use case sequence in a medium-sized pharma or specialty chemicals group should remain realistic. In the first step, agents in quality management, specifically in deviation handling and CAPA documentation, are worthwhile. The data situation is good, and the benefit effect is high. The regulatory framework already requires clean audit trails, which agentic AI provides as a byproduct. In the second step, sales and marketing automation can be connected, especially in HCP engagement and content compliance. In the third step, more complex cases come into play: production-related agents at packaging lines, scheduling optimization, and predictive quality applications.<\/p>\n Those who want to copy the Merck pattern do not copy the volume but the structure. Specifically, this means: building a data product team that curates scientific data, process data, and commercial data. Establishing an agentic AI platform lead in the IT leadership, who not only accompanies projects but also makes architectural decisions. Setting up a works council agreement that allows co-creation between the specialist department and IT without getting stuck on each individual agent.<\/p>\n For a medium-sized pharmaceutical or chemical company with 500 to 3,000 employees, the Merck pattern can be broken down into a manageable roadmap. The milestones are deliberately measured in months, not weeks, because governance issues require time.<\/p>\n The reaction of Merck’s competitors is predictable. Pfizer, Roche, Novartis, Bayer, and Boehringer Ingelheim will announce comparable packages within the next nine to twelve months, some with Microsoft, some with Anthropic plus AWS, and some with mixed models. The DACH pharma landscape is being rearranged at a speed that presents a clear question to medium-sized businesses: own platform or connected platform? Those working within a joint venture with a large corporation should clarify governance issues now, not in two years.<\/p>\n There’s a second signal that’s being overlooked in the reporting. The deal is a billion-dollar bet on Gemini. For European medium-sized businesses, this implicitly means that Google’s focus on Enterprise Agentic AI will intensify in the coming months. Those who choose a platform now will receive disproportionately high engineering support in the first 18 months because Google Cloud is looking for references for Gemini Enterprise. This window will close as soon as the first deal stories are running. Pragmatic companies will take advantage of this.<\/p>\n Another issue that needs to be on the table is how a medium-sized business rethinks its contract regime with such a provider. Framework contracts that still target classic software license logics do not handle Agentic AI workloads well. Consumption-based models, audit rights, and liability clauses for autonomous decisions need to be included. Those who only involve the purchasing department on the third use case will pay the lesson twice.<\/p>\n The sober essence: The Merck-Google announcement is not a recommendation to buy anything. It’s an occasion to schedule three meetings. First round: strategy and IT leadership to discuss whether an Agentic AI architecture decision is pending for 2027. Second round: compliance and data protection to discuss whether data governance supports the EU AI Act and simultaneous Agentic AI use. Third round: HR department and works council to set the framework for co-creation models. Those who have structured these three conversations by the end of June 2026 will be prepared for the next wave of competitor announcements.<\/p>\n One last point that’s missing from the reporting: The announcement comes at a time when German family businesses are redefining their IT strategy cycles. Many companies have completed a complete overhaul of their SAP landscapes between 2023 and 2025. The available capacities in IT and specialist departments are now supposed to flow into a new area. Those who understand Agentic AI as the next strategic layer and not as an appendix to the ERP project will approach planning with different consequences. The Merck example shows what the layer can look like when it’s thought of as a platform. Now is the moment when medium-sized IT leaders bring an honest interim status to their supervisory boards: What do we know about our own data, which use cases carry, where do we have competence, and where do we need partners? The answers can be unpleasant. Unpleasant is better than late.<\/p>\n The official announcement speaks of a multi-year partnership with an investment volume of up to $1 billion. The formulation “up to” is contractually usual. A large part is spread over license and engineering costs over a period of three to five years, with annual releases depending on achieved milestones.<\/p>\n<\/details>\n Indirectly, yes. The program is aimed at system integrators and ISVs that implement Agentic AI projects. For medium-sized businesses, this means that the usual consulting offers are subsidized by partner funds. If you’re currently requesting offers, you should ask about co-funding mechanisms, not just daily rates.<\/p>\n<\/details>\n A big one. If you use Agentic AI in regulated processes like pharmaceutical production, you quickly fall into high-risk application cases under the EU AI Act. The consequences concern documentation requirements, human-in-the-loop requirements, and transparency towards users. If you build your Agentic architecture with a view to these obligations, you save on later retrofitting.<\/p>\n<\/details>\n Sort of. For the first major Agentic AI application case, a clear lead platform makes sense, because integration, identity, and data layers would otherwise be built twice. For second and third applications, other platforms can be docked. It’s essential to have an exit strategy in the contract framework, so the lead decision doesn’t become a de facto lock-in.<\/p>\n<\/details>\n The American Merck & Co. is known as MSD in Europe and has published the press release in parallel in the MSD newsroom. The German Merck KGaA in Darmstadt is an independent company without involvement in this deal. If you talk about Merck in Germany, you need to communicate the difference clearly.<\/p>\n<\/details>\n For a medium-sized pharmaceutical or chemical company with 500 to 3,000 employees, three to five roles are sufficient as a core team: an Agentic AI platform lead, two data product managers, and a compliance officer. Additionally, rotating department representatives are brought in for each use case. The model grows with application cases, not linearly with company size.<\/p>\n<\/details>\n EU AI Act Takes Effect on April 6, 2026: What Mid-Market Tech Teams Need to Clarify by August<\/a><\/p>\n Next-Level Pharma Serialization: EU FMD Revision for IT and Supply Chain<\/a><\/p>\n AI Agents in ERP: Autonomous Processes with SAP, Microsoft, and Oracle<\/a><\/p>\n<\/div>\nThe Essentials at a Glance<\/h2>\n
\n
What exactly was announced<\/h2>\n
Why the Deal Pattern is Relevant for Medium-Sized Businesses<\/h2>\n
What can be specifically derived for DACH pharma and chemicals<\/h2>\n
Opportunities for DACH mid-market<\/h3>\n
\n
Risks that must not be underestimated<\/h3>\n
\n
A realistic rollout path in six milestones<\/h2>\n
What Executive Management and IT Leadership Should Have on the Table Now<\/h2>\n
Frequently Asked Questions<\/h2>\n
Is the $1 billion immediately due or is the volume spread out?<\/strong><\/summary>\n
Can medium-sized businesses benefit from the $750 million partner program?<\/strong><\/summary>\n
How does the EU AI Act affect Agentic AI projects in medium-sized businesses?<\/strong><\/summary>\n
Do I have to choose between Google Cloud, Azure, and AWS now?<\/strong><\/summary>\n
Is the Merck announcement in Europe the same as in the US?<\/strong><\/summary>\n
How big does my own data team need to be to operate Agentic AI properly?<\/strong><\/summary>\n
Editor’s Reading Recommendations<\/h2>\n