DeepSeek’s impact on the global IT industry has been reflected in the desire of market participants to increase their spending on computing infrastructure, according to Bloomberg Intelligence analysts. By 2032, tech giants will collectively spend more than $525 billion annually on developing their computing infrastructure.
Image source: NVIDIA
This year, hyperscalers like Microsoft, Amazon, and Meta✴ Platforms are expected to spend $371 billion on these goals, Bloomberg Intelligence reports. This is 44% more than the previous year. According to analysts, this amount will grow to $525 billion by 2032. After DeepSeek language models appear on the market, the expansion of these expenses will be faster, according to experts. In addition, the priority in the investment sphere will shift towards reasoning artificial intelligence. This is also facilitated by the activity of OpenAI.
To some extent, this forecast is intended to assuage the concerns of investors in the capital of companies that supply components for artificial intelligence systems. Some of them assumed that DeepSeek’s ability to offer efficient language models with fewer resource requirements would reduce the cost of purchasing specialized equipment.
The spread of reasoning AI will also cause a redistribution of financial flows. Models of this generation will provide developers with more opportunities to monetize software. In terms of costs, the bulk of them will also be transferred from the training period to the operation period. Accordingly, costs for training models will grow at a slower rate than before the advent of DeepSeek solutions. However, this trend will be offset by increased costs for reasoning artificial intelligence.
This year, developers’ expenses on training models will amount to more than 40% of their total costs, but by 2032 their share will decrease to 14%, according to analysts at Bloomberg Intelligence. At the same time, reasoning AI will have drawn up to half of all investments by that time. Google, which has its own chips, will be able to switch to working in new conditions fairly quickly, while companies like Microsoft and Meta✴, which depend on Nvidia’s solutions, will have a much harder time doing so.