ARK Invest predicts that global AI infrastructure spending could reach $1.5 trillion by 2030, driven by decreasing training costs and increasing demand.
ARK Invest has forecasted that global investments in AI infrastructure could soar to approximately $1.5 trillion by 2030. This anticipated growth stems from decreasing costs associated with training neural networks combined with an escalating demand for computing power.
AI adoption is surpassing internet growth rates, prompting a significant scaling of infrastructure to meet demand. This is merely the beginning of a vast expansion as both consumers and businesses reveal their needs. @downingARK discusses the latest on AI infrastructure in a recent blog.
— ARK Invest (@ARKInvest) March 25, 2026
Rapid Cost Reductions, Rising Demand
Training costs for neural networks are reportedly plummeting by about 75% each year, while inference costs for high-performing models have decreased even more dramatically by around 95%. This trend reveals that as technologies become cheaper, the applications for AI expand significantly, making it economically viable for a broader range of tasks.
The speed at which AI is integrating into society is noteworthy, having reached a 20% penetration rate in only three years, compared to the six years it took for the internet to achieve similar levels. Corporate interest is also surging; for instance, requests via OpenRouter have surged 28 times since late 2024, while revenue for firms like Anthropic skyrocketed from $100 million in 2023 to $14 billion by early 2026.
Growing Infrastructure Investments
Since the debut of ChatGPT, the demand for advanced computing has escalated sharply, reflected in Nvidia's revenue growth from $27 billion in 2022 to an expected $350 billion in 2026. Investment in server systems has seen a steep rise, increasing from 5% growth per annum prior to 2022 to a remarkable 30% in the last three years. Nvidia’s GPU-based solutions and specialized chips now dominate 86% of the server market.
Private investment in AI infrastructure exceeded $200 billion in 2025, with major allocations to foundational model developers. Meanwhile, significant deals such as Meta's $30 billion agreement with Blue Owl highlight the trend toward alternative funding avenues.
Intensifying Competition in Chip Manufacturing
The increasing demand for AI technologies has sparked fierce competition among chip manufacturers. While AMD has made strides in reducing costs for smaller model inferences, Nvidia continues to excel with heavy models due to its advanced architecture. Major tech companies like Google and Amazon are developing their own semiconductor solutions to cut costs and improve efficiency.
Forecasts indicate that by 2030, annual AI infrastructure investments could triple from current levels, with specialized ASICs projected to constitute a third of the total computing power market. Experts assert that the infrastructure improvements currently underway are foundational for a significant generational shift in technology.
The ongoing development of AI infrastructure is not merely a fleeting trend; it represents a paradigm shift that will require immense computational resources to support the deployment of AI across various sectors.
