Cloud & AIoT: Who is the Mastermind Behind the Scenes?

[baoshi.rao]

The transition of cloud computing to IoT presents significant challenges, but the trend is inevitable.

You may have noticed that in recent times, the frequency with which cloud computing, AI, and the Internet of Things (IoT) are mentioned together has reached unprecedented levels. Some might dismiss this as a marketing gimmick or industry jargon, where vendors feel compelled to mention these three to stay relevant.

However, there are no coincidences in such alignments.

The reason cloud computing vendors are increasingly embracing IoT lies in underlying logic and industry trends. The key question is: How can cloud vendors successfully sell AI technologies and capabilities to businesses, factories, or government agencies—essentially creating what is termed 'industrial AI' or the industrial internet market?

Exploring this question reveals that without networked, AI-capable edge devices, the vision of cloud and AI in industries cannot materialize. This reality contradicts the earlier belief that cloud computing would virtualize the hardware-heavy IT industry, with all computing happening remotely.

Instead, as AI opens new market possibilities, edge hardware has become an inseparable part of cloud computing. Moreover, as AIoT integrates into specific scenarios, cloud computing gains new opportunities to function as an operating and coordination system.

Today, a major variable in the cloud computing market stems from the ambition and challenges of becoming the mastermind behind AIoT.

To better understand the relationship between cloud, AI, and IoT, we must revisit the origins of the enterprise market and examine the booming industrial internet and industrial intelligence of recent years.

In reality, the enterprise IT market has been evolving for two decades. If the goal were merely selling apps and websites to industrial clients, the business model would remain unchanged.

Cloud vendors, representing the industrial internet players, aim to sell intelligent technologies and platforms that directly impact production workflows. But here’s the catch: While consumer internet relied on PCs and smartphones, what will drive the industrial internet?

For AI to be applied in production, it cannot depend on office computers or phones but requires production, monitoring, and service devices with AI capabilities. In other words, businesses need a new wave of hardware tailored for AI-driven production.

Thus, AIoT is both the ideal and the inevitable path for the industrial internet. Significant progress has been made in this field over the past two years.

From a hardware perspective, the development of AIoT in the industrial internet can be divided into three stages.

In this process, cloud computing gains a massive opportunity: becoming the command center for AIoT.

The reason for outlining the industrial demand for AIoT is to address a critical question when examining the cloud + AIoT market shift: Can enterprises adopt AI by simply purchasing hardware? Can this business model exist independently of the internet?

Connecting to public clouds introduces security risks, which are often unacceptable for core production systems. However, as discussed earlier, the push for cloud-based AI isn’t solely due to the dominance of internet and cloud companies.

From a broader trend perspective, public clouds occupy an unavoidable position in enterprise AIoT adoption. Several key capabilities in industrial AI require public cloud support, and bypassing them would entail significant costs and operational challenges.

Here are a few examples of what cloud computing brings to AIoT:

1. AI Training and Inference: For most enterprises, local processing of heavy AI tasks incurs substantial computational costs. The ideal workflow involves cloud-based training, edge deployment, and iterative cloud retraining with new data. This allows businesses to update AI models as needed via the public cloud.

2. Data Management: AIoT systems rely on a cycle of data learning, storage, and application, demanding real-time data access and storage. High synchronization requirements make local-only processing impractical. A hybrid approach—local preprocessing with cloud-based deep processing and storage—is more viable.

3. Toolchain Synchronization: Deep AI adoption in enterprises requires seamless compatibility across numerous AI hardware and software components. Maintaining synchronization with evolving toolchains to sustain competitive advantage often necessitates cloud integration for efficiency.

4. Edge Computing: While enterprises cannot rely solely on cloud or edge, edge computing strikes a balance between efficiency and cost. The required edge infrastructure often falls within cloud vendors' service offerings, benefiting from rapid advancements in public cloud technologies.

As previously noted, enterprises adopting AI typically seek solutions rather than standalone APIs or hardware, especially non-cloud-native and large organizations. But who orchestrates these solutions?

Cloud vendors are well-positioned to lead this charge. Offering integrated AIoT services aligns with the industry’s efficiency demands.

These points illustrate how cloud computing integrates with AIoT hardware systems. Securing these roles positions public clouds to reap substantial rewards in the era of industrial intelligence. However, this transformation faces hurdles due to the current chaos in the industrial internet, delaying the full realization of 'cloud + AI + IoT.'

While industrial AI sounds promising—with reports even heralding it as the fourth industrial revolution—real-world adoption remains sluggish. Factory owners, retail chains, and municipal services are not rushing to embrace AI or upgrade to AIoT en masse.

The disconnect stems partly from immature technical solutions and a fragmented AIoT market. The lack of standardization and vendor-specific solutions create confusion and compatibility issues.

From an enterprise perspective, reluctance to adopt public cloud-based industrial AI often stems from:

1. Overhyped Claims: Vendors tout AI’s benefits, but businesses face steep costs and uncertainties, especially in industries lacking AI precedents. Concerns over practicality and cloud security further deter adoption, reducing AIoT to a vendor showcase.

2. Market Fragmentation: Every vendor promotes 'cloud + AIoT,' but their solutions vary widely in approach, metrics, and terminology. This confusion makes it hard for users to compare options, with even basic digital service providers jumping on the AI/IoT bandwagon.

3. Premature Adoption: ...

The integration of AIoT solutions with industries remains a significant challenge today. Most so-called AIoT solutions still primarily provide machine vision solutions centered around smart cameras. These solutions are applicable to nearly all industries but mostly serve as supplementary enhancements. Truly industry-specific AIoT devices and technologies that deeply integrate with sensors, production lines, and operating systems remain largely untapped.

Overall, the industrial application of AIoT systems presents an excellent opportunity for cloud computing providers and serves as a core solution for cloud + AI to unlock vast industrial markets. Particularly for China, with its massive industrial structure and pressing demands for quality improvement and efficiency, guiding cloud + AIoT to drive industrial revolution is far from impossible.

However, in this transformation driven by cloud computing, promising prospects coexist with real-world challenges.

During the野蛮生长 (wild growth) phase, the混乱 (chaos) of各自为战 (fragmented efforts) has become the most obvious绊马索 (stumbling block) in this long-tail industry that requires platformization and standardization.

Of course, challenges also意味着 (signify) opportunities. To untangle the复杂局面 (complex situation) and propel cloud + AIoT into a rapid development红利期 (dividend period), three key areas deserve attention today:

1. Standardization: The lack of standards has long been a critical issue in the IoT industry. As demand forces the industry toward standardization and platformization, this problem may finally find a solution. Currently, IoT protocols and standardization efforts led by academic organizations, specific technologies, and operating systems are underway. While the难度 (difficulty) is high, there is hope—especially for AIoT standards统一 (unification) at the operating system and development system levels, which may be relatively more feasible.

2. Hardware Innovation: The weakest link in the cloud + AI + IoT组合 (combination) lies in IoT hardware innovation. There is still a significant gap between China's industrial chain and global领先水准 (leading standards). Addressing the core需求 (demands) of AI-driven industries by填补 (filling) the空白 (gap) in industrial-grade IoT devices presents an opportunity with both industrial value and profit potential. This is also a关键环节 (key step) to突破 (break through) the current AIoT困境 (dilemma).

3. Developer Ecosystem: For AIoT to penetrate industries, the focus must be on solutions tailored to industry-specific needs and peculiarities. This often requires developers who can bridge the gap between major technology providers, hardware manufacturers, and industry users, facilitating沟通 (communication) and供需协调 (supply-demand coordination). High-quality developers can trigger a鲇鱼效应 (catfish effect) in the chaotic cloud + AIoT industry, forcing the supply chain to协同 (collaborate) for特定市场 (specific markets). Thus, nurturing and empowering developers is a critical task for major players.

In summary, the path of cloud computing pivoting toward IoT支点 (fulcrum) amid智能变局 (intelligent transformation) has gained某种共识 (some consensus) in today's industry. However, compared to external expectations or the ambitious goals set by industry insiders during the early stages of development, the actual progress falls far short.

Breakthroughs may happen overnight or beyond the horizon. But there is no doubt that the cloud computing industry is rushing toward the transformative出口 (exit) brought by IoT.

The终点 (destination) of this revolution is highly anticipated.