Viewpoint: The AI reasoning market may exceed oil and become one of the world’s largest
Recently, Dylan Patel, founder of SemiAnalysis, predicted in a podcast interview that AI reasoning will become one of the world’s largest markets. Its size could exceed that of oil and account for several percentage points of global GDP.
Patel believes that with each model iteration upgrade, the rate at which the number and value of tasks that can be completed expands continues to outpace the growth of compute capacity. As a result, a long-term shortage of compute may persist.
Regarding compute-demand projections, Patel expects that by 2030, the combined compute demand of only two companies—OpenAI and Anthropic—will exceed 100 gigawatts.
As for the outlook for space data centers, he thinks their impact will remain negligible in the next 3 to 5 years, but by 2040, more than half of new compute capacity may enter space.
Patel pointed out that the key constraint lies in ground-level energy costs and the ability to build new data centers. Once the economics of deploying compute in space surpass those on the ground, the migration of compute to space will become an inevitable trend.
In the area of co-design between hardware and software, Patel said that the AI efficiency improvements over the past three years have not mainly come from hardware, but from optimizations at the model layer and across layers.
For example, DeepSeek carried out targeted optimization for Nvidia’s Hopper architecture, so its performance is excellent on that chip, but it performs worse on TPUs.
By contrast, Anthropic’s models are better suited to TPUs, while OpenAI’s models are more geared toward GPU architectures.
Patel also said that the so-called “Nvidia CUDA moat” is essentially not just CUDA itself, but the result of open-source model ecosystems being widely co-optimized around GPU use.
In addition, Jensen Huang has been strongly supporting emerging cloud-computing providers. The goal is to prevent the compute market from being dominated by a small number of giants, and to promote healthy industry development and innovation vitality.
#AI推理市场规模
Recently, Dylan Patel, founder of SemiAnalysis, predicted in a podcast interview that AI reasoning will become one of the world’s largest markets. Its size could exceed that of oil and account for several percentage points of global GDP.
Patel believes that with each model iteration upgrade, the rate at which the number and value of tasks that can be completed expands continues to outpace the growth of compute capacity. As a result, a long-term shortage of compute may persist.
Regarding compute-demand projections, Patel expects that by 2030, the combined compute demand of only two companies—OpenAI and Anthropic—will exceed 100 gigawatts.
As for the outlook for space data centers, he thinks their impact will remain negligible in the next 3 to 5 years, but by 2040, more than half of new compute capacity may enter space.
Patel pointed out that the key constraint lies in ground-level energy costs and the ability to build new data centers. Once the economics of deploying compute in space surpass those on the ground, the migration of compute to space will become an inevitable trend.
In the area of co-design between hardware and software, Patel said that the AI efficiency improvements over the past three years have not mainly come from hardware, but from optimizations at the model layer and across layers.
For example, DeepSeek carried out targeted optimization for Nvidia’s Hopper architecture, so its performance is excellent on that chip, but it performs worse on TPUs.
By contrast, Anthropic’s models are better suited to TPUs, while OpenAI’s models are more geared toward GPU architectures.
Patel also said that the so-called “Nvidia CUDA moat” is essentially not just CUDA itself, but the result of open-source model ecosystems being widely co-optimized around GPU use.
In addition, Jensen Huang has been strongly supporting emerging cloud-computing providers. The goal is to prevent the compute market from being dominated by a small number of giants, and to promote healthy industry development and innovation vitality.
#AI推理市场规模