SPS 2025: Software-Defined Automation, Edge AI, and Data Center Growth Reshape Industrial Landscape
Based on over 60 interviews and 80 booth visits at SPS 2025, this analysis decodes the hidden logic behind the market’s uneven recovery. While attendance and exhibitor numbers rose, the real story lies in how software-defined automation, agentic and edge AI, and a new data center vertical are reshaping industrial strategy. From Audi’s 100% uptime with virtual PLCs to the surge in Chinese vendors, we unpack five deep insights that go beyond the headline trends.
---
The Uneven Recovery: Market Upturn with Hidden Fault Lines
At first glance, SPS 2025 painted a picture of cautious optimism. Attendance climbed 9% year-over-year, exhibitor numbers rose 5%, and the exhibition halls buzzed with renewed energy. Yet beneath the surface, the recovery is anything but uniform. The German Electrical and Electronic Manufacturers’ Association (ZVEI) warned of escalating global trade barriers and geopolitical uncertainty, casting a long shadow over traditional machine-building markets.
Siemens’ own fiscal 2025 full-year automation revenue fell 4% despite a strong Q4, reflecting sustained pressure in core factory automation segments. In contrast, Beckhoff reported 7–10% growth after a brutal 33% dive in 2024—a recovery driven by its adaptability in niche applications and growing traction in non-industrial verticals. This divergence reveals a crucial insight: the industrial automation market is no longer monolithic. Winners are those that can pivot swiftly to emerging demand pools.
[IMAGE: A split image: left side shows a crowded SPS 2025 exhibition floor with diverse flags, right side shows a factory floor with server racks being integrated.]
The most prominent new growth vertical is the data center segment. Industrial OEMs that once depended exclusively on automotive and machine-building customers are now reallocating engineering resources to serve hyperscalers and colocation providers. Data center automation—spanning cooling control, power distribution, and building management—has become a vital buffer against stagnation in traditional sectors. For example, Beckhoff’s TwinCAT platform is now deployed in cooling units for several major hyperscaler data centers, while Siemens reported a 30% order increase in data-center-related automation components in Q4 2025.
This structural shift is redefining supply chains. Traditional automation vendors now compete with IT-centric players like Schneider Electric and Vertiv, and the boundaries between factory floor and server room are blurring. The data center opportunity is not just about volume; it demands new competency in OT security, high-availability networking, and software-defined infrastructure.
Equally significant is the surge in Chinese vendor participation—up 41% compared to SPS 2024. Companies like Inovance, Estun, and Wecon are no longer mere component suppliers; they showcased complete motion control systems, servo drives, and even virtualized PLC solutions. Their presence signals a structural rebalancing of global competition. For European OEMs, this is both a threat and a push to accelerate innovation cycles, especially in software-defined automation where ecosystem lock-in becomes a differentiator.
---
Software-Defined Automation: From Vision to Production Reality
If SPS 2024 was the year software-defined automation (SDA) was introduced as a concept, SPS 2025 was the year it entered production. Siemens’ SDA architecture—built around Industrial Edge, vPLC (virtual PLC), and SIMATIC AX—is now a fully deployed platform strategy. The vPLC runtime runs on any x86 device, from Siemens’ own Industrial Edge devices to standard industrial PCs. The most compelling proof point came from Audi: the automaker has deployed vPLC at its Neckarsulm and Ingolstadt plants for six months, achieving 100% uptime in mission-critical body-in-white processes. This is not a lab experiment; it is production-grade virtualization that matches or exceeds traditional PLC reliability.
[IMAGE: A diagram showing a traditional PLC replaced by a software container running on an industrial edge device, with arrows indicating data flow to cloud and on-premise analytics.]
The hidden economic logic behind SDA is powerful: decoupling control software from hardware enables faster innovation cycles and IT-like engineering workflows. OEMs can amortize software investments across multiple hardware generations, roll out updates without physical intervention, and serve diverse verticals—including data centers—with the same code base. For machine builders, this reduces time-to-market for new features and simplifies global compliance with regional safety standards.
Beckhoff is also pushing aggressively in this direction. Its TwinCAT runtime now supports containerized deployment on standard x86 hardware, and the company demonstrated live hot-swapping of control applications at SPS 2025 without interrupting machine operation. However, the competitive race is shifting from pure technology to ecosystem lock-in and runtime compatibility. Siemens’ advantage lies in its mature Industrial Edge orchestration layer and the growing SIMATIC AX developer community, while Beckhoff bets on its openness and performance for ultra-low-latency applications.
Notably, software-defined automation also addresses one of the industry’s most pressing pain points: OT security. By centralizing control software management and isolating runtime environments, SDA reduces the attack surface. Several vendors highlighted integrated security features such as signed software containers, runtime attestation, and automated patch management—important steps toward making industrial DataOps both compliant and resilient.
The data center vertical is a natural early adopter of SDA. Hyperscalers often manage thousands of identical controllers for cooling, lighting, and power; virtualization allows them to treat these as software resources that can be orchestrated remotely. During SPS 2025, at least three major automation vendors announced dedicated software-defined control suites for data center infrastructure, signaling that this is no longer a side project but a core market.
---
AI at the Edge: Agentic Demos, Physical AI, and the Maturing Software Stack
Edge AI has been discussed for years, but SPS 2025 showed it reaching a maturity inflection point. The most striking demos were the first agentic AI applications—systems that do not just infer from sensor data but autonomously create and adjust workflows. One prominent example: a German robotics integrator showcased an agent that monitored a multi-station assembly line and, upon detecting a recurring jam, dynamically reordered downstream operations to avoid bottleneck downtime. This marks a fundamental shift from reactive analytics (e.g., “predict a failure”) to proactive decision-making (“reconfigure the process to prevent the failure”).
[IMAGE: A factory floor with robotic arms and cameras overlaid with transparent AI agent nodes (chat bubbles, decision trees) and edge computing cabinets.]
The software stack powering this shift is maturing. Standardized AI runtimes such as ONNX and TensorFlow Lite are now embedded in industrial edge devices from Siemens, Beckhoff, and Phoenix Contact. Orchestration layers that manage AI model updates, versioning, and resource allocation are becoming commodity components, lowering deployment friction. At SPS 2025, several booth demonstrations showed models being deployed from cloud to edge in under 30 seconds with zero downtime—a dramatic improvement over the days-long manual processes common two years ago.
Edge AI’s application scope is also expanding beyond vision-based quality inspection. Predictive maintenance now incorporates multi-modal data (vibration, thermal, acoustic, current) fused in real time. Quality optimization loops using reinforcement learning adjust process parameters on the fly. Adaptive control systems, such as those shown by Festo, learn ideal cycle times for different product variants without human programming. These capabilities are especially valuable in high-mix, low-volume production, where fixed automation struggles.
Physical AI—the convergence of AI with robotics and simulation—emerged as a forward-looking narrative at SPS 2025. Several vendors demonstrated digital twins that not only mirror production but also host AI agents that interact with physical robots. For example, a demo from Bosch Rexroth showed a digital twin of a welding cell where an AI agent tested thousands of toolpath variations in simulation before the actual robot executed the optimal one. This blurring of virtual and physical control hints at a future where production systems are co-managed by real-time digital twins and autonomous AI agents.
The implications for industrial DataOps are significant. To feed edge AI models, factories need robust data pipelines that span sensors, edge nodes, and cloud analytics. Several exhibitors announced new data management frameworks specifically designed for OT environments, including time-series databases optimized for high-frequency data and metadata tagging schemes that make industrial data discoverable and reusable. OT security remains a key enabler—without it, opening edge devices to continuous model updates and agent control introduces unacceptable risk.
---
Conclusion: Three Forces Converging
SPS 2025 confirmed that the industrial automation industry is navigating a period of asymmetric recovery. The data center vertical provides an unexpected growth engine, while Chinese vendors intensify competition and force innovation. Software-defined automation has crossed the threshold from pilot to production, with Audi’s vPLC deployment as a landmark case. Edge AI is evolving into agentic, autonomous systems that promise to redefine how factories adapt to change.
These three forces—data center automation, SDA, and edge AI—are not independent; they are converging. Software-defined control makes it easier to deploy AI agents. Edge AI optimizes data center cooling and factory production alike. Data center demand funds R&D that trickles down to traditional automation. For industry players, the message is clear: the future belongs to those who can integrate these trends into a coherent, software-driven, and vertically flexible strategy. The next SPS will likely show how quickly that integration accelerates.