Narrow Gate Holding said it is expanding investments across artificial intelligence (AI), robotics, semiconductors and energy, betting that the next phase of AI competition will be driven by computing infrastructure rather than software alone.
The U.S.-based investment firm said it views the four sectors as the core infrastructure underpinning future AI development, with demand increasingly shaped by access to computing power, electricity and advanced semiconductor technologies.
Narrow Gate said rising costs associated with developing frontier AI models are shifting competitive advantage toward companies capable of investing in long-term infrastructure. The firm cited industry trends showing that the cost of training leading AI models has increased from about $100 million to more than $1 billion.
It also pointed to data indicating that the United States accounted for roughly 75% of global AI venture capital deal value in 2025 and is home to nearly 60% of the world’s top-tier AI researchers, reinforcing its view that the country remains central to AI innovation.
The investment firm said growing demand for AI infrastructure is exposing what it described as a “terrestrial compute wall,” where the availability of land, electrical power and water is becoming a limiting factor for expanding data centre capacity.
According to Narrow Gate, these constraints are reshaping AI into an infrastructure-driven industry, where long-term investment in physical assets could prove more important than short-term gains in software performance.
The firm said it sees AI, robotics, semiconductors and energy converging into a single technology ecosystem, with chips and power infrastructure providing the foundation for AI applications and robotic systems.
As a result, Narrow Gate said it will continue to focus on later-stage companies developing the infrastructure required to support large-scale AI deployment, with investments centred primarily on the U.S. market while maintaining a global investment outlook.
