GuShiio.com 
A new round of AI boom will trigger fluctuations in the chip semiconductor industry.
On February 26, GuShiio.comAGI learned thatThe Chinese version of the “Technology Trends 2025” report recently released by Deloitte China shows that it is estimated that the global chip market will be US$576 billion in 2024.Among them, AI chip sales accounted for 11%, exceeding US$57 billion. It is estimated that by 2025, the value (scale) of the new generation of AI chips will exceed US$150 billion; by 2027, the global AI chip market will grow to US$400 billion, and conservative estimates will also reach US$110 billion.
The report also pointed out that as many as 70% of corporate organizations are actively exploring or applying large language models. As generative AI such as DeepSeek is applied to many enterprises, AI servers will account for nearly 60% of total server spending by hyperscale cloud service providers.Over the next decade, AI may focus entirely on performing tasks rather than just enhancing human capabilities.
In this regard, Meng Xiaofan, National Managing Partner of Deloitte China Consulting Business Services and Transformation, said: AI is reshaping all aspects of the enterprise at an unprecedented speed. From new generation infrastructure operations to decision support, AI has become the core of the enterprise’s future development. driving force. However, if companies want to fully tap the potential of AI, they must achieve comprehensive coordination in the fields of strategy, talent, architecture, and especially data. rdquo;
This is the 16 consecutive year that Deloitte has released such a focus around the world. Professional predictions on the future of technology in the next 18 to 24 months. Zhang Sen, partner of technology strategy and transformation services in Deloitte’s China technology and transformation department, told GuShiio.comAGI and others that with the widespread application of AI technology, whether it is in the future digital construction of enterprises or the application of AI, companies must begin to re-examine their own hardware strategy, infrastructure strategy, especially the increasing core value of AI to enterprises. ldquo; A large number of our enterprises are still in their infancy.& rdquo;
It is reported thatOver the past two years, ChatGPT has set off a new round of AI craze. Many chips used to train and reason new generation AI models and used in data centers cost tens of thousands of dollars. In particular, the demand for NVIDIA GPUs has exceeded production capacity. In today’s gold rush, companies that provide pickaxes and shovels are making huge profits.According to recent reports, due to the surge in demand for DeepSeek models, China companies are significantly increasing orders for Nvidia H20 GPUs.
Nvidia CEO Huang Renxun once pointed out that as the cloud GPU capacity has been basically saturated, the company will also launch a new generation of AI chips, which will significantly improve performance. Hyperscale cloud service providers are snapping up GPUs that have just left production lines and spending nearly US$1 trillion to build data center infrastructure to meet customers ‘needs for renting GPUs.
Today, AI chips are widely used in enterprise edges, computers, smartphones, data centers and other edge devices. Deloitte predicts that by 2025, AI PC device sales will account for half of all PCs. It is expected that by 2028, almost all PCs will be equipped with at least some AI chips, also known as neural processing units (NPUs), and these NPUs are equipped with equipment prices are expected to be 10%-15% higher, and some edge-side AI PC chips will have computing speeds of more than 40 TOPS per second (trillions of operations per second). However, some buyers are still cautious about these new PCs.
Deloitte research found that the new generation of AI chips based on data centers costs (may cost) US$30000; the new generation of AI PC (personal computer) chips may cost US$30; the AI chips on smartphones may cost US$3; and the cost of developing AI IoT chips should be around US$0.3.
In terms of capital markets, as of mid-December 2024, the total market value of the world’s top ten chip giants, including Nvidia, was US$6.5 trillion, an increase of 93% from US$3.4 trillion in mid-December 2023 and 235% higher than the same period in 2022; in terms of investment and financing,As of the end of 2024, the cumulative investment and financing volume of global AI chip startups has exceeded US$7.6 billion.These include RISC-V architecture AI chips, AI inference chips, photonic chips, NPU, AI chip manufacturing equipment and other sub-areas.
AMD CEO Lisa Su predicts that by 2028, the potential market size of global AI acceleration chips will reach US$500 billion.
“AI is breaking through the limitations of pure software form and moving towards a new era of hardware innovation.& rdquo; Meng Xiaofan said that the birth of a new generation of AI-specific chips allows AI models to be embedded in personal computers and edge devices to achieve localized and offline computing. This progress not only improves the user experience, but also lays a solid foundation for future computing infrastructure. Especially in fields such as medical equipment and robotics, edge computing devices integrated with AI have shown great application potential.
The “Technology Trends” report announced six major technology trends, including spatial computing leading the interactive revolution, customized small models becoming a new outlet for AI, hardware innovation leading a new era of AI, AI boosting the transition of IT departments, and the quantum era challenging encryption technology., AI subverts core systems.
It is worth mentioning that small models are becoming a new outlet for AI.Zhang Sen believes that in the past, companies tended to purchase ready-made large models, but this large-scale solution was often difficult to balance professionalism and flexibility. Today, as personalized needs grow, more and more companies are turning to tailor-made small models that can efficiently process specific tasks, generate multimodal content, and provide users with personalized virtual assistant services through multimodal collaboration.
“Enterprises have many customization needs, and under their own core competitiveness, enterprises need to do some integration (training) of large models and small models to achieve greater value. rdquo; Zhang Sen said.
Chris Bedi, chief customer officer of ServiceNow, said that AI Agents in specific fields or industries can change the landscape of human-computer interaction in enterprises.
However, in the long run, the AI chip industry faces both challenges and opportunities.
The Deloitte report shows that by 2025, AI chips will become increasingly important and are expected to account for 50%-75% of global chip sales. It is estimated that by 2030, the scale of the global chip semiconductor industry will reach US$1 trillion. However, whether it is the Chip Act or the terminal impact of the industrial chain, the development of the AI semiconductor industry will remain very fragile at least in the next year or two. Any downward trend in AI-related spending and shortage of parts and components may have an adverse impact on the wider global chip supply chain.
Liu Junlong, national managing partner of technology strategy and transformation services for Deloitte’s China technology and transformation department, emphasized that AI has gradually become a core part of corporate business. In the context of the vigorous development of local AI models, companies should keep in mind that the true value of technology lies in its deep integration with business scenarios. Deloitte will continue to help China corporate managers make corporate digital innovation and operations a strategic fulcrum and continue to build differentiated competitive advantages.
