Software giant SAP is quietly swapping some of the expensive artificial intelligence models inside its products for cheaper alternatives, a move its chief financial officer says could deliver more than €2 billion in efficiency gains over the next two years. The strategy, reported by Reuters, is a response to the rising cost of running AI features as more customers switch them on.
SAP, Europe's largest software company, is a behind-the-scenes powerhouse: its enterprise resource planning (ERP) software manages the back offices of many of the world's biggest businesses, handling everything from finance and HR to supply chains. The company now runs more than 70 AI models from providers including OpenAI, Anthropic, and Google, according to Reuters.
CFO Dominik Asam told Reuters that SAP has already pulled some of the most expensive models from certain features, such as its internal consulting assistant, and replaced them with lower-cost options when the results still met the required standard. The goal is not necessarily to cut SAP's total AI bill—as more customers turn these features on, overall usage and spending can rise—but to keep margins "healthy" as the company scales.
Why model choice matters for SAP's bottom line
SAP typically charges customers per completed task, not for the computing resources used behind the scenes. That means the revenue from an AI-powered workflow is relatively fixed, while the cost of running the model for each request—known as "inference" cost—can vary widely depending on which model is used. By routing more requests to cheaper models without hurting quality, SAP can keep a bigger spread between what it charges and what it pays to fulfill each task.
This is a classic margin-defense play. As AI usage grows, the cost of inference can balloon if a company relies solely on premium models. By diversifying its model mix, SAP aims to limit gross-margin pressure even as demand ramps up. For investors, this makes SAP's AI profitability increasingly sensitive to two things: its model mix and vendor costs, and whether compliance demands force it into a narrower, more expensive set of models for certain customers.
Compliance and local options add another layer
Cost is only half the constraint. Customers in government and heavily regulated industries are paying closer attention to where AI models come from and how data is handled. Many European clients are asking for local, compliant options, which could limit SAP's ability to use the cheapest models in those cases. That means the efficiency gains may not be uniform across all customers—some may require premium models for data-sovereignty or regulatory reasons.
This is a growing theme across the tech sector. As AI becomes embedded in enterprise software, vendors are having to balance cost, performance, and compliance. For SAP, the ability to offer a range of models—from premium to budget—could become a competitive advantage, allowing it to serve both cost-sensitive and compliance-heavy customers.
What it means for investors
For investors, SAP's €2 billion efficiency target is really about defending per-task margins. If the company can scale AI features without seeing its costs grow at the same pace, that should support profitability. But the strategy also highlights a broader shift in the AI industry: the era of unlimited spending on the most powerful models may be giving way to a more pragmatic approach, where companies pick the right tool for each job.
Investors should watch how SAP's model mix evolves and whether it can maintain quality while cutting costs. Also worth tracking is whether compliance demands force SAP into a more expensive set of models for certain customers, which could limit the efficiency gains. The company's ability to navigate these trade-offs will be a key factor in its AI-related profitability over the next couple of years.
For context, other companies are also grappling with AI costs. For instance, analysts have been adjusting targets on AI infrastructure software, reflecting the shifting economics of AI. And as Tesla expands its AI-driven features, the cost of running those systems is a growing consideration for that company as well.
SAP's move is a reminder that AI's financial impact isn't just about revenue—it's also about the cost side. For everyday investors, it's a signal that even the biggest tech players are looking for ways to make AI more efficient, which could have implications for the entire AI supply chain.
