The rapid spread of AI faces financial and infrastructure hurdles. According to a new Bain & Company report, sustaining AI infrastructure will require over $500 billion in annual data center investment by 2030. To support this investment, the industry is expected to generate $2 trillion in annual revenue. However, the report predicts that the industry will fall approximately $800 billion short of this target.
AI Requires Massive Investment
The underlying problem lies in the fact that computing demand is growing faster than the tools needed to meet it. Despite the slowdown in Moore’s Law, AI workloads continue to increase. This is forcing data centers to continually scale. As a result, it is estimated that the global AI computing demand could reach 200 GW by 2030.
Such a massive energy demand requires improvements to existing local power grids, extended electrical equipment procurement processes, and high-performance cooling systems. Furthermore, supply shortages are facing key silicon solutions such as HBM and CoWoS. According to the report, demand exceeds the industry’s capacity in every way except price.
If capital is in short supply, large data centers will gravitate toward systems that offer the highest return per watt and area. This will increase the importance of full-rack GPU platforms like the Nvidia GB200 NVL72 or AMD Instinct MI300X, while constraining the supply of low-volume configurations and high-end desktop solutions.
This impact is also felt on the PC side. If training costs remain high and data center inference is hampered by power constraints, the workload shifts more to edge devices.
This creates new opportunities for laptop and desktop manufacturers with NPUs in the 40-60 TOPS range. Inference at the edge is both faster and less capital intensive.
The report paints a picture of infrastructure investments taking years and being expensive. Meanwhile, AI models are doubling in size every six months. This supports concerns that high-performance silicon, memory, and cooling systems will remain both scarce and expensive over the next decade.
