Breaking Through Cycles: Insights from Three Global Reports on the Real Breakthroughs in the AIoT Industry
Today, AI is at a crucial inflection point of deep integration with the physical world. To better understand industry trends and clarify the boundary between hype and reality, three newly released authoritative reports between July and August 2025 offer insights from different perspectives.
These three reports are as follows:
1. "Technology Trends Outlook 2025" — McKinsey Global Institute (MGI)
This report systematically outlines the top 13 cutting - edge technology trends that will impact enterprises and industries in 2025, covering three major areas: the AI revolution, computing power and connectivity, and engineering innovation. It emphasizes that AI has become an amplifier for all infrastructure and application scenarios. The integration of AI with the physical world, the Internet of Things (IoT), edge computing, and robotics is reshaping the global value - creation and industrial competition landscape.
2. "The State of AI 2025" — Bessemer Venture Partners (BVP)
From the perspective of a globally renowned venture capital firm, BVP conducts an in - depth analysis of the growth models of AI - native enterprises, the evolution of AI infrastructure, and how AI is reshaping enterprise - level software and vertical industries. The report focuses on the systematic innovation, commercialization paths, and implementation challenges brought by AI.
3. "The GenAI Divide, STATE OF AI IN BUSINESS 2025" — Massachusetts Institute of Technology (MIT)
Based on empirical research, this report reveals the ROI gap in the implementation of generative AI in enterprises. Although global enterprises are investing heavily in AI, 95% of them have not achieved significant commercial returns.
All three reports provide forward - looking analyses of the integration of AI and IoT, the industrial opportunities, commercial challenges, and technological evolution trends of AIoT.
They have reached a high level of consensus on issues such as AI being the infrastructure and industrial engine, scenario - focused and ROI - driven development, ecological collaboration, and trust systems. However, they also show distinct differences on practical issues such as whether to develop AI systems in - house or purchase them externally, pursue explosive growth or sustainable resilience, and focus on front - end experience or back - end automation.
What consensus have the three authoritative reports reached on the future of AIoT? Where do they have different views on key issues? This article will systematically sort out these cutting - edge insights, extract the latest paths for AIoT innovation and industrial implications, helping the industry to navigate through the hype and seize the next growth opportunity in the intelligent economy.
From Technological Frenzy to Industrial Consensus: The Main Thread of AIoT in Three Authoritative Reports
Consensus 1: The Deep Integration of AI and IoT is an Inevitable Trend
McKinsey points out in "Technology Trends Outlook 2025" that artificial intelligence has evolved from a single - purpose technological tool to an underlying operating system driving the digital and intelligent transformation of all industries (as shown in the figure above). AI is no longer just passively analyzing data but actively participating in scenarios such as process optimization, product innovation, energy management, robotics, and autonomous driving, becoming the intelligent brain of the physical world.
BVP's annual AI report emphasizes that truly impactful AI companies often break through by linking AI with the physical world, creating implementable business closed - loops and new service models through AIoT.
MIT's NANDA project further emphasizes that the combination of AI and IoT is not just about data collection and decision - making automation but enabling each physical node to have autonomy, collaboration, memory, and context - awareness capabilities.
All three reports clearly state in their core discussions that the deep integration of AI and IoT has become the definite main thread for global technological and industrial upgrading.
Against the backdrop of the current hype about the omnipotence of large models in the industry, both BVP and MIT offer more practical judgments.
BVP points out that although the breakthroughs in large models and AGI are highly imaginative, most of the most successful AIoT implementation cases are "small - scale, deep - integration" projects, focusing on specific industry pain points and deeply embedding into business processes, rather than blindly pursuing large - scale and all - encompassing general applications.
MIT verifies through numerous enterprise cases that only AI projects that are closely integrated with the physical world and business scenarios can create real value. An industry consensus is emerging: the industrial opportunities of AIoT start from a realistic approach of focusing on deep vertical scenarios and "fine - tuning small models" rather than simply talking about omnipotent large models.
Consensus 2: Scenario Focus and ROI - Driven Development are the Main Themes of AIoT Commercialization
Whether it is McKinsey's large - scale sample survey or BVP's investment analysis of AI - native enterprises, the conclusions are highly consistent. The commercialization of AIoT ultimately depends on the value creation in real scenarios and measurable business returns.
The BVP report repeatedly emphasizes that the development of AIoT enterprises can only achieve the leap from pilot projects to large - scale implementation by focusing on specific processes and business nodes that are "high - ROI, high - pain - point, and strongly - demanded."
MIT's NANDA project finds through empirical research on 350 enterprises that 95% of enterprises have not achieved commercial returns in the implementation of generative AI. The core problem is that they are detached from real processes and only stay at the stage of pseudo - intelligence or superficial integration. The report points out that activities such as data upload, model invocation, and report presentation often fail to bring substantial efficiency improvements and cost optimizations, but instead consume a large amount of budget and resources.
Only by deeply embedding AI capabilities into business links with clear returns, such as production, operation and maintenance, supply chain, and energy management, can the ROI closed - loop be truly achieved.
Consensus 3: Platform - Based and Ecological Collaboration is Better than Working Alone
Against the backdrop of the increasingly complex AIoT industrial chain and the accelerating technological evolution cycle, platform - based and ecological collaboration has become the main theme emphasized in all the reports.
The MIT report points out through enterprise cases that open cooperation with professional AI service providers and platform - type enterprises can significantly improve the success rate of projects.
BVP also emphasizes that AIoT enterprises do not need to reinvent the wheel in terms of basic algorithms and hardware but should embrace the industrial ecosystem, integrate resources, share capabilities, and improve efficiency through standard protocols, open interfaces, and multi - party collaboration.
McKinsey clearly states in its global technology trends analysis that the competitiveness of future enterprises will depend on their ability to find their position and value interface in the global distributed intelligent network.
The industry consensus is very clear: the success of AIoT depends on platform thinking and ecological co - construction. Only through cooperation and alliance can the collaborative innovation and large - scale implementation of the industry be truly stimulated.
Realistic Dilemmas in Path Selection: Different Answers from Three Reports
In the wave of the accelerating evolution of the AIoT industry, in addition to reaching many consensuses, the three authoritative reports also reveal some unavoidable structural conflicts and realistic challenges. Behind these differences are not only the differences in enterprises' own capabilities, industry characteristics, and development stages but also the trade - offs and choices commonly faced in the global technological transformation.
First, regarding the path choice between in - house development and external procurement, MIT's empirical research provides a clear data comparison.
The report shows that the commercialization success rate of enterprises' in - house developed AI systems is only 33%, while the success rate of projects that choose to cooperate with professional AI service providers or platform - type enterprises is as high as 67%. This huge difference indicates that most enterprises struggle to independently support the full - process closed - loop of AI systems in terms of algorithms, data, computing power, and operations.
MIT believes in the report that the in - house development model often falls into the dilemma of high investment and low output and may even become a technical trap of reinventing the wheel. However, some leading technology giants and industries with high requirements for compliance and data security, such as finance and healthcare, still prefer to develop core systems in - house to ensure data security, controllable capabilities, and differentiated competition. Although this strategy is reasonable, in the reality of scarce industry resources and rapid technological iteration, it often leads to longer project cycles, low ROI, and missed market opportunities.
The conflict between in - house development and external procurement is essentially a dynamic balance between industrial division of labor and innovation capabilities. Enterprises need to make more rational trade - offs based on their own resource endowments and business requirements.
Second, regarding the choice between pursuing explosive growth and building sustainable resilience, BVP's analysis is highly representative.
The report distinguishes between "supernovas" — AIoT enterprises that achieve a surge in users and valuation in a short period — and "stars" — enterprises that have long - term in - depth operations in niche markets, high customer stickiness, and a stable profit structure.
Reality shows that in the early stage of the AIoT industry, "supernovas" with explosive growth may emerge due to technological breakthroughs and market opportunities. However, after the bubble bursts and the market returns to rationality, only "star - type" enterprises with continuous innovation capabilities, a deep understanding of industry needs, and the ability to weather economic cycles can truly stand firm.
BVP's investment cases show that simply pursuing scale and speed or blindly betting on niche markets will bring hidden dangers in customer retention, profitability, and ecological construction. Truly competitive AIoT enterprises must find a dynamic balance between explosive growth and in - depth operation: they should be brave in innovation, quick to iterate and seize opportunities, and patient in building barriers and deeply exploring customer value.
The debate between explosive growth and resilience is essentially a race between short - term opportunities and long - term value. The industrial development requires the intersection of these two curves to achieve sustainable growth.
Finally, regarding whether to focus on front - end experience or back - end intelligence, the research data in the three reports also reveal an obvious gap.
Currently, many enterprises are concentrating their AI investments on front - end processes such as sales, marketing, and customer interaction, hoping to drive user growth and brand upgrading through intelligent interfaces. However, the research by BVP and MIT points out that the projects that truly achieve significant ROI for enterprises mostly come from invisible value - chain processes such as back - end intelligence, process optimization, and operation and maintenance management.
For example, MIT's research finds that more than half of the generative AI budget is invested in sales and marketing, while back - end intelligence contributes the highest return on investment.
McKinsey's trend analysis also shows that AIoT can only achieve fundamental breakthroughs in efficiency improvement and cost optimization by deeply embedding into enterprises' core operations, supply chain, and asset management.
The conflict between front - end experience and back - end intelligence is not just a choice of resource allocation but represents different understandings of the essence of AIoT commercialization. In the next stage of industrial upgrading, the strategic focus must shift to the invisible value chain, using AI capabilities to drive real process reengineering and organizational transformation.
Key Drivers of Deep Transformation: Autonomy, Collaboration, and Trust Reshape AIoT
AIoT has been given high expectations and imagination in the industrial community. From smart cities and intelligent manufacturing to autonomous driving and digital energy, almost every field related to the intelligent reconstruction of the physical world is labeled with AIoT. However, the gap between the ideal and the reality is extremely obvious.
Both McKinsey's global research and MIT's empirical data reveal that most AIoT projects still stay at the stage of superficial intelligence — data collection, device networking, and basic automation — but are disconnected from enterprises' core business processes and difficult to form a complete value closed - loop.
Many seemingly impressive smart hardware and AI scenarios are actually just "smart vases" that do not really solve the core pain points of production efficiency, cost optimization, or business growth. Only AIoT projects that are deeply integrated with real business scenarios and embedded in core operation processes can bring continuous and measurable commercial returns.
Furthermore, the key to the real breakthrough of AIoT lies in the construction of an autonomous action system and an autonomous economic entity.
Traditional IoT mostly plays a passive role in data perception and upload. In the future, AIoT nodes must have the capabilities of autonomous decision - making, memory, context understanding, and collaborative learning. This means that each device and node can make independent decisions and collaborate with other nodes based on real - time data, its own experience, and the external environment.
Only when the AIoT system evolves from a "data carrier" to an "autonomous intelligent agent network" can the entire industry give birth to a real distributed physical intelligent economic entity, achieving autonomy, collaboration, and self - evolution.
To achieve this transformation, continuous breakthroughs in algorithms and hardware are needed, and the industry also needs to promote platform - based openness, standard protocols, and multi - party collaboration. The establishment of an ecological cooperation and trust system is an insurmountable threshold for the large - scale implementation of the AIoT industry. As shown in the MIT report and BVP cases, by building an open ecosystem, standard protocols, data governance, and compliance transparency to form a cross - industry and cross - platform collaborative network, it is easier to gain industrial trust and achieve large - scale expansion.
Finally, the value anchor of AIoT is shifting from the traditional "data upload to the cloud platform" model to a model centered on "back - end intelligence + scenario ROI + distributed collaboration."
Rather than chasing superficial intelligence in front - end experience, the strategic focus should be placed on back - end process optimization, intelligent device management, and in - depth exploration of high - ROI scenarios. Only when AIoT becomes an invisible engine for improving enterprise operation efficiency, reducing costs, and creating new business models can its industrial - level value be truly unleashed.
Conclusion
The real transformation of AIoT is not just about making each physical device smarter but enabling each node to have memory, context understanding, autonomous decision - making, and collaborative evolution capabilities, thus becoming a new type of intelligent economic entity.
The next decade of AIoT has begun. Only by going beyond superficial intelligence, deeply rooting in value - creating scenarios, and building a distributed intelligent collaborative ecosystem can real growth dividends be obtained in the new wave of intelligence. Now is the starting point for action.
References:
1. "Technology Trends Outlook 2025", Source: McKinsey
2. "The State of AI 2025", Source: Bessemer Venture Partners
3. "The GenAI Divide, STATE OF AI IN BUSINESS 2025", Source: MIT