Viewpoint: The AI inference market may surpass oil to become one of the world's largest markets

Recently, Dylan Patel, founder of SemiAnalysis, predicted in a podcast interview that AI inference will become one of the world's largest markets, potentially surpassing oil and accounting for several percentage points of global GDP.

Patel believes that with each iteration and upgrade of models, the number of tasks that can be completed and the expansion of value continue to outpace the growth in computing power, so a shortage of computing power may persist in the long term.

Regarding projections for computing power demand, Patel estimates that by 2030, the combined computing power demand of OpenAI and Anthropic alone will exceed 100 gigawatts.

As for the development prospects of space-based data centers, he believes their impact will still be negligible in the next 3 to 5 years, but by 2040, more than half of new computing power may be deployed in space.

Patel pointed out that the core constraints lie in ground-level energy costs and the ability to build power infrastructure. Once the economics of space deployment surpass those on the ground, the migration of computing power to space will become an inevitable trend.

In the field of hardware and software co-design, Patel stated that over the past three years, AI efficiency improvements have not primarily come from hardware, but from model-level and cross-layer collaborative optimization.

For example, DeepSeek has performed specialized optimizations for NVIDIA's Hopper architecture, resulting in excellent performance on that chip, but performing poorly on TPUs.

In contrast, Anthropic's models are better suited to TPUs, while OpenAI's models lean more toward GPU architecture.

Patel also said that the so-called NVIDIA CUDA moat is essentially not just about CUDA itself, but the result of the open-source model ecosystem generally being co-optimized around GPUs.

Additionally, Jensen Huang has been vigorously supporting emerging cloud computing providers, aiming to prevent the computing power market from being dominated by a few giants, and committed to promoting the healthy development and innovation vitality of the entire industry.

#AI Inference Market Size