According to external sources, the AI hardware boom is expanding from GPU to a wider infrastructure. As computing demand continues to rise, electricity, heat dispersion, memory, network equipment and data centre capacity are becoming a more visible part of the AI industry chain.

In Weida remains the center of this round of AI capital spending. The article mentions that income from data centre operations amounted to $75.2 billion in the first quarter of the financial year, an increase of 92 per cent over the previous year. For its part, the International Energy Agency predicts that global data centre power will rise from about 485 TWh in 2025 to about 950 TWh in 2030, indicating that the expansion of computing power is increasingly constrained by infrastructure carrying capacity.

Power may be limited first.

According to the article, simply increasing the number of AI servers does not solve the problem of the availability of computing power. If the data centre does not have enough power, it will be difficult to release the new equipment. As a result, power supply equipment, refrigeration systems and on-site power generation programmes are considered to be the direct beneficiaries of AI expansion.

For example, Vertiv provides power supply and heat management equipment for high-density data centres; Bloom Energy provides on-site fuel cell power generation programmes, which are increasing in areas with limited grid capacity.

Memory price escalation system cost

In addition to electricity, the increase in memory prices was seen as another signal. The article mentions that as memory costs rise, the price of the system rises. This means that the infrastructure bottlenecks are not fixed, may shift between links and further affect the cost of server construction.

From this perspective, AI is moving from a single chip company to a more complete infrastructure chain. Those who are able to relieve the pressure on electricity, heat dispersion, memory and room capacity are more likely to gain market attention in the next phase.

It's not just about speed.

According to the article, when comparing AI infrastructure companies, it is important not only to see the growth of income but also to combine the valuation levels. One of the simplified methods given was to divide the expected growth rate by the market rate, and to see how much investors paid for each percentage point increase.

  • 40% growth ÷ 40 times gain, divided by 1.0
  • 25% growth ÷ 20 times gain, 1.25
  • Higher scores, lower valuing costs for unit growth

It was also mentioned that AI ' s related income growth, free cash flows, customer concentration, capital expenditure demand, and whether companies are at the infrastructure bottlenecks are also the focus of observation.

Overall, external sources believe that Weidar will remain at the centre of the AI computing system, but the next phase is no longer limited to GPU manufacturers. With the expansion of the data centre, electricity, heat dispersion, memory, network and room capacity are becoming more important components of the AI investment narrative.