A startling shift is underway in the artificial intelligence industry: computing expenses have now surpassed total human payroll costs. Nvidia recently confirmed that infrastructure spending for AI compute has jumped 69%, signaling a new era where hardware costs dominate budgets once reserved for researchers and engineers.
This trend challenges the assumption that AI's scaling is limited solely by algorithmic breakthroughs. Instead, the real bottleneck is becoming the cost of electricity and hardware. A new MIT study reveals that automation is currently economically viable in only 23% of visual tasks, highlighting that human labor remains the cheaper option for most jobs.
The financial implications are profound. As companies pour billions into developing models like GPT-5.5, the rising cost of compute is reshaping the landscape. Smaller players may find it increasingly difficult to compete, while giants like Nvidia benefit from surging demand for their GPUs.
This development underscores a pivotal moment: the era of cheap AI scaling may be ending, forcing the industry to rethink its economic foundations.