Six-Months After DeepSeek Shock, Will US Remain Dominant?

Changes brought about by open-source AI model

RSS

July 04, 2025

  • Miho Tanabe

Summary

◆Since the “DeepSeek shock” in early 2025, the development of generative AI models has accelerated, and the performance, lightweight design, and cost reduction of open-source AI models (models whose technical information is publicly available) have progressed. Several models capable of high inference capabilities even with limited computing resources have emerged, and the technical superiority previously held by closed-source AI models (models whose technical information is not publicly available) is rapidly eroding.

◆Improvements in the performance of lightweight models have made local deployment and commercial use more realistic, expanding the options available to corporations. In response, closed-source AI model corporations are also lowering prices and introducing lightweight models, intensifying competition. US corporations maintain their advantage in terms of performance and cloud connectivity, but price, flexibility, and transparency will become key competitive factors in the future, and traditional advantages alone will no longer be enough to ensure continued success.

◆In addition, demand for GPUs, which are high-performance computing resources, continues to grow. With the spread of lightweight models, there are increasing cases where corporations are deploying AI within their own environments or on edge devices with on-device data processing capabilities, and demand for GPUs is spreading in various areas. Furthermore, with the evolution of generative AI, the capabilities required of GPUs are diversifying. Demand for GPUs is expanding not only in terms of quantity but also in terms of applications and performance, and it is unlikely to decline in the future.

◆In the future, with the expected widespread adoption of generative AI by corporations, countries will need to urgently establish usage policies and legal frameworks. Furthermore, the spread of “Embodied AI” such as physical AI will also gain attention. Generative AI is entering a new phase as a technology that underpins society and industry, and it is entirely possible that emerging corporations will challenge the existing technological hegemony.

Daiwa Institute of Research Ltd. reserves all copyrights of this content.
Copyright permission of Daiwa Institute of Research Ltd. is required in case of any reprint, translation, adaptation or abridgment under the copyright law. It is illegal to reprint, translate, adapt, or abridge this material without the permission of Daiwa Institute of Research Ltd., and to quote this material represents a failure to abide by this act. Legal action may be taken for any copyright infringements. The organization name and title of the author described above are as of today.