Must-read: Seven Insights into AI, IoT, and Edge Computing Worth Paying Attention to in 2025

With the deep integration of the Internet of Things (IoT) and Artificial Intelligence (AI), various industries such as industry, manufacturing, energy, and logistics are experiencing an unprecedented wave of intelligent transformation. From edge devices to cloud analytics, and from hardware deployment to data-driven decision-making, the IoT is reshaping the operational models and competitive landscapes of enterprises. However, along with technological opportunities come challenges: skill gaps, cybersecurity threats, supply chain uncertainties, and the impact of global trade policies are all testing whether enterprises can gain an edge in this digital transformation.

Recently, analysts and executives from IoT Analytics, Verizon Business, IDC, and Lantronix discussed with CRN some of the key insights they've observed in the IoT market this year. The author has compiled and translated the highlights:

Insight 1: IoT manufacturers are facing a significant AI skill gap

As technologies such as generative AI, edge intelligence, and large model inference are accelerating their penetration, the IoT industry is undergoing a structural reshaping of capabilities. The AI skill gap has become one of the core bottlenecks limiting the industry's leap forward. Sinha, the Chief Analyst for IoT components, connectivity, and security at IoT Analytics, said, "We've noticed a huge skill gap in the IoT market, especially in integrating AI technologies into IoT products and services."

Firstly, the imbalance between the talent structure and skill spectrum is becoming increasingly evident. For a long time, the core competencies of IoT manufacturers have been concentrated in traditional areas such as hardware design, embedded development, wireless communication protocols, and device management. The introduction of AI has forced enterprises to quickly fill the gaps in areas such as algorithm engineering, model training, computing power optimization, data governance, and MLOps. However, in reality, a large number of engineers lack systematic AI skills, and talents with cross-disciplinary capabilities are scarce and take a long time to train. This has led to a situation where "people are waiting for technology" when enterprises are promoting AI empowerment. Similar challenges also occurred during the "integration of OT and IT" - IT personnel were not familiar with the OT environment, and OT personnel did not understand the IT environment. If the two sides are required to collaborate without cross-domain training, a large number of complex problems will arise.

Secondly, the iteration speed of AI technology is much faster than the lifecycle of IoT products, resulting in a cycle mismatch. The lifecycle of traditional IoT devices often spans several years, while AI technology is updated on a quarterly or even monthly basis. During the process of product design, testing, and deployment, the technological baseline may have changed, making the AI solutions at risk of becoming obsolete before they are put into use. This cycle mismatch not only increases R & D costs but also forces IoT manufacturers to rethink their version planning, software-hardware separation, updatable architecture, and online model upgrade capabilities.

Thirdly, although external dependence can fill the gap in the short term, it is difficult to form long-term competitiveness. Many enterprises try to accelerate the implementation of AI projects by outsourcing or introducing third - party teams. However, third - parties must have an in - depth understanding of the enterprise's device logic, protocol stack, data characteristics, and business scenarios, which usually requires a lot of time and communication costs. In the case of NDAs, this process becomes even more complex. More importantly, relying on external forces makes it difficult to build sustainable internal capabilities, which will weaken the enterprise's initiative in the AI - driven competition in the long run. If an enterprise wants to introduce AI into the industry or within the company, it must enable its employees to master relevant skills instead of relying on short - term solutions from third - parties.

Therefore, building an "AI - ready" organizational capability for the IoT has become a key proposition that the industry must face.

Insight 2: Tariffs have truly changed corporate strategies and supply chain strategies

Tariffs have indeed changed the current business models of many enterprises. They have pushed up raw material costs, affected product pricing, and squeezed supplier profits. A market sentiment survey by IDC shows that 60% of enterprises believe that rising tariffs are threatening their profitability and the stability of their technology budgets.

Tariffs have led to delays in equipment procurement, supply chain disruptions, and forced enterprises to make strategic adjustments to ensure that there is no significant impact on customers, such as relocating manufacturing sites and diversifying supply chains.

Meanwhile, tariffs have also brought about some "innovation effects" in a sense. Carlos Gonzalez, the Research Manager of Industrial IoT and Smart Strategy at IDC, pointed out in the "IDC Global DataSphere IoT Device Installed Base and Data Generation Forecast" report he co - authored, "I can't say that tariffs are the only reason for the changes, but they have definitely had an impact. What we're seeing is not a flat demand for hardware but a downward trend. Enterprises currently don't plan to invest too much in hardware in the future, but the applications around data continue to grow strongly."

The current reality is that enterprises are gradually realizing that the hardware supply is already difficult and may remain tight in the future. Therefore, they must "do more with less hardware." This is why synthetic data is so crucial at present - it allows us to conduct more analysis based on existing information. A large amount of this data comes from unstructured environments, such as vision systems. By conducting deeper analysis on the data from existing cameras, more value can be mined from these unstructured sources.

Overall, tariffs continue to bring instability to the market. But even in an unstable situation, both customers and suppliers are well aware that some investments cannot be stopped: manufacturing upgrades, IoT network and system construction, and cybersecurity. Investments in these areas will not stop. Enterprises may look for various ways to offset the additional costs, and these measures will inevitably have an impact on pricing. Growth will still continue, but the way costs are passed on to the downstream will be different, and ultimately, it will still affect prices and supplier profits.

Insight 3: The increasing popularity of synthetic data in IoT applications

With the deep integration of the IoT and AI, data has become the core asset driving intelligent applications. However, when enterprises use data for analysis, modeling, and simulation, they often face multiple constraints such as intellectual property protection, sensitive information security, and privacy compliance. In this context, synthetic data is becoming a key tool for enterprises to solve this dilemma.

Synthetic data refers to artificial data designed to mimic real - world data. It is generated through statistical methods or using AI technologies (such as deep learning and generative AI). Although it is artificially generated, synthetic data retains the basic statistical characteristics of the original data it is based on. Therefore, synthetic datasets can supplement or even replace real datasets.

Synthetic data is a highly realistic replication of real data without involving original sensitive information. It can conduct multi - dimensional analysis and simulation while protecting intellectual property. Its main applications include:

Model training and algorithm development: Enterprises can use synthetic data to generate training sets and build AI models without directly accessing real production or customer data.

Cross - enterprise collaboration: Different manufacturers or partners can share data for joint analysis or system optimization without revealing core business data.

System simulation and scenario testing: Before deploying IoT devices, synthetic data can be used for simulation testing to verify the effectiveness of edge computing, AI inference, and network policies.

The driving factors for the development of synthetic data are:

Data security and privacy concerns: Enterprise users are highly sensitive to the security and confidentiality of their own data, which has become an important factor affecting IoT cloud applications, cybersecurity investments, and the implementation of AI projects.

Cross - system and cross - manufacturer analysis requirements: IoT devices are becoming increasingly diverse, and data is scattered across different systems. Synthetic data can enable cross - platform analysis without actually exchanging sensitive data.

Data volume and diversity requirements for AI applications: The accuracy of AI models depends on large - scale and diverse datasets, but real data is often limited or restricted. Synthetic data can make up for this shortcoming and accelerate the deployment and iteration of AI.

Overall, synthetic data is not only a tool to solve privacy and data security problems but also a key means for enterprises to maximize data value in the IoT and AI era. In the future, with the improvement of algorithm generation capabilities and simulation accuracy, synthetic data will play an increasingly core role in IoT device intelligence, cross - system interconnection, and end - to - end AI solutions.

Insight 4: IoT manufacturers (even competitors) are strengthening their interconnection capabilities

In the IoT field, interconnection between competitors is becoming an important growth direction. Customer demand is driving cooperation among manufacturers because the market cannot wait for each supplier to develop independent solutions.

In the past, IoT systems were mostly based on "single - manufacturer closed ecosystems," with each manufacturer building an independent system around its own protocols, devices, and platforms. However, as the deployment scale expands and the co - existence of cross - brand devices becomes the norm, customers are gradually no longer willing to accept "isolated systems." For example, a factory may use PLCs, robots, and sensors from multiple manufacturers simultaneously; in building scenarios, a large number of systems are built by different suppliers; and in home and consumer products, there are numerous brands and different protocol families... To achieve higher operational efficiency and lower system integration costs, customers are actively demanding data interconnection and system compatibility between different manufacturers, fundamentally changing the past competition model.

Meanwhile, as third - parties such as cloud providers, data platform companies, and AI service providers are becoming more involved in the IoT ecosystem, they propose that "as long as they can access all device data, they can help enterprises integrate data and achieve stronger cloud or AI capabilities." This has forced underlying hardware manufacturers to open interfaces, share data formats, and be compatible with standard protocols; otherwise, they will be excluded from the larger data ecosystem.

Therefore, whether through open standards or open protocols, this interconnection is indeed promoting cooperation among companies (including direct competitors). To achieve cross - manufacturer and cross - scenario collaboration, the industry is accelerating its move towards mature open standards. For example, OPC UA is currently the standard for open communication protocols between devices, and Matter is reshaping the interoperability ecosystem of consumer - grade devices, moving the smart home from platform fragmentation to unified interconnection.

Enterprises are beginning to realize that they should not only compete in hardware but also in ecological capabilities, service capabilities, and integration value - this will bring about deeper industry changes: from hardware differentiation to software, platform, and ecological collaboration differentiation.

Insight 5: The Industrial Internet of Things is driving the rapid development of hybrid AI models

With the continuous development of the Industrial Internet of Things (IIoT), industries are facing increasing pressure to move real - time intelligent capabilities to the edge. No single enterprise can handle this challenge alone. The future AI - driven IIoT will be centered around collaboration - hardware, software, and network suppliers will work together to build an integrated ecosystem to support hybrid AI models for edge intelligence.

Building a complete IIoT solution with embedded AI requires the joint contribution of the entire technological spectrum. For example, to implement an AI - driven drone or industrial robot, high - performance cameras and sensors, efficient processors, advanced video compression technologies, reliable network connections, and cloud platforms for orchestration and analysis are needed. In this environment, hybrid AI models have emerged, achieving a balance between speed, cost, and performance by sharing intelligence between edge devices and the cloud.

In industrial operations, every second can affect production efficiency and safety, so the immediate local decision - making ability of edge AI is particularly crucial. For example, robots can react immediately when they detect obstacles, compressors can predict potential failures, and drones can identify abnormal situations without waiting for cloud feedback. This not only improves response speed and device uptime but also provides guarantees in terms of data privacy and security. Meanwhile, the cloud is responsible for more complex analysis, large - scale data aggregation, and continuous AI model training. The two work together to form a hybrid architecture that can provide both real - time intelligence and support long - term insights and large - scale analysis.

The application of hybrid AI in the IIoT is accelerating, covering almost the entire industrial operation chain, from predictive maintenance and preventing equipment downtime to process optimization in manufacturing and energy systems, remote monitoring of pipelines, HVAC systems, and heavy equipment, as well as autonomous operations of drones and robots. To implement these applications, enterprises not only need efficient and secure computing hardware but also reliable network connection capabilities. Especially in remote or harsh environments, the stability of these infrastructures is directly related to business continuity and the level of intelligence.

As data gradually migrates from the center to the edge, it is estimated that about 70% of global data will reside at the edge in the next decade. According to Precedence Research, by 2034, the edge AI market will reach $143 billion, and the IIoT will be an important driver of the growth of the edge AI market.

Insight 6: Cybersecurity remains one of the biggest challenges for the IoT

As the number of IoT devices continues to grow, the potential attack surface is also expanding. Although surveys show that 98% of enterprises expect to obtain substantial benefits from IoT deployments within two years, and most enterprises expect to see returns in less than 12 months, 43% of enterprises still regard cybersecurity as the biggest challenge in IoT deployments.

Due to different deployment architectures, IoT devices may cover multiple locations, involve devices from different manufacturers with different security capabilities, and operate in environments with limited physical security. To address these diverse potential attack paths, IoT deployments often require a more complex security system than traditional IT environments.

Danny Johnson, the Vice President of IoT and Managed Connectivity Platform at Verizon Business, said, "To this end, enterprises are becoming more mature and intelligent: by implementing a zero - trust architecture, establishing dedicated networks with enhanced security to manage device connections, and using AI - driven threat detection technologies to identify and defend against evolving risks in advance. As technologies such as AI and the IoT continue to develop, the means and strategies for ensuring security also need to be continuously expanded and evolved to deal with new threats."

Insight 7: AI is revolutionizing the way IoT data is processed

AI is changing the way enterprises manage interconnected operations such as the IoT in ways that were unimaginable a few years ago. The latest report shows that more than four - fifths (84%) of enterprises consider AI to be a key technology for the IoT, and 70% of enterprises say that AI has accelerated their IoT deployments, and there are clear reasons behind this trend.

IoT sensors generate a huge amount of data, forming a large stream of unclassified information. This data must be processed and analyzed to be valuable. AI plays a role in this process by converting the collected large - scale data into actionable business insights quickly, efficiently, and with almost no additional manual intervention. In manufacturing, this means that AI can enable predictive maintenance, issue early warnings before equipment failures cause downtime, optimize the supply chain, and identify and correct inefficiencies in real - time. AI can also support automated decision - making on the shop floor, including event recognition, insight analysis, action planning and execution, and the generation of highly automated and almost real - time reports.

These changes are also driving the transformation of enterprises' IoT concepts. Enterprises that were previously cautious about the complexity of data management or had difficulty clarifying the return on investment are now accelerating their progress because AI can speed up the establishment of analysis frameworks and provide quantifiable results, providing a strong basis for investment decisions.

Source: 8 Big IoT Trends To Watch In 2025, According to Analysts And Executives, CRN. What is Synthetic Data? IBM

This article is from the WeChat official account "Internet of Things Think Tank" (ID: iot101), author: Sophia, published by 36Kr with authorization.