AsianFin -- Global Al chip sales in 2024 were estimated to account for 11% of the global chip sales of US$576 billion, according to Deloitte’s Tech Trends 2025 released on Tuesday.Growing from roughly $US50 billion today, the AI chip market is forecasted to reach up to US$400 billion by 2027, though a more conservative estimate is US$110 billion, the report said.Today, large language models (LLMs) have taken root, with up to 70% of organizations, by some estimates, actively exploring use cases of LLMs, as per the report.AMD CEO Lisa Su has moved her estimate for the total addressable market for AI accelerator chips up to $500 billion in 2028, a number which is larger than sales for the entire chip industry in 2023. AI servers will account for close to 60% of hyper scalers’ total server spending, according to Gartner. Some enterprises have found immediate business value in using LLMs, while others have remained wary about the accuracy and applicability of LLMs trained on external data. Over the next decade, AI’s role in business is expected to shift from merely augmenting human capabilities to executing tasks autonomously, said the Deloitte report.AI is currently reshaping every aspect of enterprises at an unprecedented pace. From next-generation infrastructure operations to decision-making support, AI has become the core driving force for the future development of businesses, said Meng Xiaofan, a managing partner for Consulting Business Services and Transformation at Deloitte China.However, for companies to fully unlock the potential of AI, they must achieve comprehensive alignment in strategy, talent, architecture, and especially in the domain of data, Meng added.Sam Zhang, a partner of Technology Strategy and Transformation Services at Deloitte China, emphasized that with the widespread use of AI, enterprises must reassess their hardware and infrastructure strategies to remain competitive.As AI becomes increasingly central to the value proposition of businesses, Zhang noted that many companies are still in the early stages of digital transformation.Nvidia CEO Jensen Huang has acknowledged that cloud GPU capacity is nearing saturation. The company is preparing to launch a new generation of AI chips with significantly improved performance. Hyperscale cloud service providers are investing heavily in infrastructure, with nearly $1 trillion being spent to build data centers capable of meeting the growing demand for GPU rentals.AI chips are being increasingly integrated into a wide range of devices, from enterprise edge devices and computers to smartphones and data centers. Deloitte forecasts that by 2025, AI-enabled PCs will account for half of all PC sales.By 2028, nearly all PCs are expected to be equipped with Neural Processing Units (NPUs), AI chips capable of performing complex computations. Devices with NPUs are expected to be priced 10%-15% higher than non-AI devices, with some edge AI chips reaching computational speeds of over 40 TOPS (trillion operations per second). However, some consumers remain cautious about adopting these new AI-powered devices.Deloitte’s research indicates that next-generation AI chips for data centers could cost up to $30,000, while AI chips for smartphones could be priced around $3. Meanwhile, AI chips for Internet of Things (IoT) devices may cost as little as $0.3.AI is breaking through the limitations of being purely software-based and stepping into a new era of hardware innovation, said Meng.The emergence of AI-specific chips has enabled AI models to be embedded into personal computers and edge devices, which allows for localized and offline computation. This advancement not only enhances the user experience but also lays the groundwork for future computing infrastructure, particularly in fields like medical devices and robotics, where AI-powered edge computing devices hold immense potential, Meng noted.Deloitte's latest report also highlights several key technology trends shaping the future. These include the rise of spatial computing, the emergence of customized small AI models, hardware innovations driving new AI capabilities, AI’s transformative impact on IT departments, and progress made in quantum challenging encryption technologies.Zhang Sen, a Deloitte expert, explained that while large models have traditionally been used by companies, these solutions often struggle to balance specialization and flexibility. As demand for personalized AI increases, more companies are turning to small models that can be tailored for specific tasks. These models, in combination with multimodal capabilities, are more efficient at handling specialized tasks, generating multimodal content, and offering personalized virtual assistant services.Companies have many customization needs, and given their core competencies, they need to integrate (train) large models with small models to unlock greater value, Zhang noted.Chris Bedi, Chief Customer Officer at ServiceNow, added that AI agents tailored to specific industries or domains could fundamentally transform human-machine interactions within enterprises.Looking ahead, the AI chip industry faces both challenges and opportunities. By 2025, Deloitte forecasts that global AI chips sales could account for 50%-75% of global chip sales revenue. However, the sector remains vulnerable to fluctuations in AI-related spending and potential component shortages, which could negatively impact the broader semiconductor supply chain.Liu Junlong, a managing partner at Deloitte China, emphasized that AI has become a core component of enterprise operations. As Chinese AI models continue to evolve, Liu stressed that the true value of AI lies in its integration with business scenarios.更多精彩内容，关注钛媒体微信号（ID：taimeiti），或者下载钛媒体App