Artificial Intelligence Trends and Insights - 2025

The AI landscape is increasingly competitive, with rising competition, open-source momentum, and the ascent of China posing significant monetization threats^[1]. Despite the potential for AI to 'do your work for you,' reminiscent of the early days of email and web search, the path to monetization is complex^[1]. The intense competition and innovation, accessible compute, and global adoption of AI-infused technology create both opportunities and challenges^[1]. The race to build the most capable general-purpose models may lead to commoditization and diminishing returns, as output quality converges across different providers^[1].

AI model compute costs are high and rising, while inference costs per token are falling, leading to performance convergence and increased developer usage^[1]. The cost of training frontier AI models has seen ~2,400x growth over eight years^[1]. As inference becomes cheaper and more efficient, competitive pressure among LLM providers increases, focusing on latency, uptime, and cost-per-token^[1]. This shift benefits users and developers with lower unit costs but raises questions about monetization and profits for model providers^[1]. The AI developer ecosystem is expanding, exemplified by NVIDIA's growth to 6 million developers^[1]. Computing-related patents in the USA have exploded, particularly post-ChatGPT launch, indicating heightened innovation^[1].

AI performance has surpassed human levels of accuracy and realism in many areas^[1]. In 2024, AI systems exceeded human performance on the MMLU benchmark test^[1]. Conversations with AI are becoming increasingly realistic, with a significant percentage of testers mistaking AI responses for human-generated content^[1]. AI is also achieving increasingly realistic image generation, as demonstrated by advancements in models like Midjourney^[1]. Furthermore, AI is enabling realistic audio translation and generation, with companies like Spotify beginning to accept audiobooks AI-translated into 29 languages^[1].

AI adoption is rising across various industries and sectors, including technology, enterprise, education, government, and research^[1]. Tech incumbents are prioritizing AI, with CEOs emphasizing AI's transformative potential in areas like coding, search, shopping, and healthcare^[1]. Traditional enterprises are also increasing their focus on AI, targeting growth and revenue rather than just cost reduction^[1]. Global CMOs are increasingly using or testing AI tools for marketing activities^[1]. In the education and government sectors, there's a growing trend of announcing AI integrations, such as Arizona State University’s ‘AI Acceleration’ and the creation of ChatGPT tailored for USA federal agencies^[1].

The development of AI presents both significant benefits and risks^[1]. The potential for AI to free humanity from repetitive work, increase production, accelerate scientific research, and provide cures for diseases is immense^[1]. However, there are also risks associated with the misuse of AI, including lethal autonomous weapons, surveillance, biased decision-making, and cybersecurity threats^[1]. Balancing these benefits and risks requires careful consideration and thoughtful leadership^[1].

CapEx spend among big technology companies has been on the rise for years, driven by increased data use and storage, and this trend has accelerated with the rise of AI^[1]. Big Six tech companies in the USA have seen a +63% Y/Y increase in CapEx^[1]. AI model training dataset sizes are growing exponentially, further driving the need for increased CapEx^[1]. This investment is benefiting companies like NVIDIA, with data center revenue as a percentage of global data center CapEx increasing^[1]. Data centers are key beneficiaries of AI CapEx spend, with construction value and capacity seeing significant growth^[1]. However, data centers are also electricity guzzlers, necessitating a focus on energy efficiency and sustainable practices^[1].

Open-source AI is experiencing a resurgence, offering lower costs and greater accessibility for developers and enterprises^[1]. China is emerging as a leader in the open-source race, with several large-scale models released^[1]. While closed models dominate consumer market share, open-source models are preferred by startups, researchers, and independent developers^[1]. China's advancements in AI are part of a broader effort to shift from low-cost manufacturing to high-value technology, with implications for national security and geopolitical power^[1]. The competition between the USA and China in AI is intensifying, requiring strategic responses to promote innovation and maintain a competitive edge^[1].

AI momentum is extending into the physical world, with intelligence embedded in vehicles, machines, and defense systems^[1]. This includes the rise of self-driving fleets, AI-driven mining exploration, agricultural modernization through AI-powered weeding, and intelligent grazing systems^[1]. Technologies like Starlink are expanding global internet access, enabling new users to come online with AI-native experiences^[1]. As AI continues to evolve, it is expected to fundamentally reshape how work gets done, how capital is deployed, and how leadership is defined across companies and countries^[1].