Dan Roberts, co-founder of IREN, said the biggest obstacle facing artificial intelligence today is not the supply of advanced chips but the infrastructure needed to deploy them. He stressed the need for integrated power, data centers, GPUs, and enterprise software solutions to work together.
The infrastructure bottleneck
Roberts argued that while semiconductor shortages grab headlines, the real constraint is the physical and digital backbone that supports AI workloads. Without reliable electricity, cooling, and software orchestration, even the most powerful processors can't deliver results.
Why integration matters
Rather than focusing solely on chip design or manufacturing, Roberts called for coordinated investment across the entire stack. Power generation must pair with data center construction, which must pair with GPU availability and enterprise software. A gap in any one link stalls progress.
Roberts' remarks come as AI companies scramble to secure not just hardware but the complex ecosystem required to run it. His point suggests that the race to scale AI may hinge less on the next chip and more on building the grid, the buildings, and the code to tie it all together.
