DeepSeek opens a new era, how will the Chinese data center industry chain change?

During the Spring Festival, DeepSeek's next-generation open-source model sparked heated discussions due to its astonishingly low cost and high performance.

Through innovative optimization, DeepSeek achieved cutting-edge AI model performance with approximately 2,000 H800 GPUs, with training costs far below industry norms.

Morgan Stanley's latest research suggests that this breakthrough could reshape the data center industry chain starting in 2026:

• GPU supply chain companies serving first-tier ultra-large-scale data centers may be the first to be impacted.
• As the barriers to AI deployment lower, suppliers targeting the enterprise market and ASIC solution providers may gain more development opportunities.

The impact on the industry chain will show differentiation

Morgan Stanley believes that, in the short term, the investment pace of large tech companies may not change immediately. For example, Meta recently announced a capital expenditure plan of USD 60-65 billion for 2025, a significant increase from USD 35-40 billion in 2024.

Market research by Morgan Stanley analysts also shows that there have been no changes in current GPU construction plans. If capital expenditure plans are adjusted, it may not happen until 2026 at the earliest.

DeepSeek's breakthrough will have varying degrees of impact on different segments of the industry chain:

1. GPU-related supply chain

• Companies in the supply chain related to NVIDIA, especially those serving first-tier ultra-large-scale data centers, may be the most affected.
• Suppliers serving second-tier and below cloud service providers and enterprise customers will be relatively less affected.

1. Server and component manufacturers

• If GPU prices drop, server assemblers may benefit as this will improve their profit structure.
• ASIC solution providers may gain new market opportunities.

Four potential market scenarios

Morgan Stanley believes that the market may see the following four scenarios:

1. Baseline scenario: GPU prices remain stable

• First-tier ultra-large-scale data centers will continue to build as planned.
• ASIC supply chain companies may perform better than GPU supply chains.
• Overall industry valuations may face pressure until the impact of DeepSeek becomes clearer.

1. Positive scenario: GPU prices drop but spending remains

• Data centers can procure more hardware equipment.
• Particularly beneficial for downstream companies like server assemblers.

1. Popularization scenario: Lower barriers to AI deployment

In the medium to long term:

• The decline in AI training and inference costs will drive more enterprise customers to build AI capabilities.
• Suppliers targeting the enterprise market will benefit significantly.
• Companies like Gigabyte, Lenovo, and ASUS are expected to benefit.

1. Disruptive scenario: Re-evaluation of computing demand

If DeepSeek's experience is promoted:

• Maximum computing power may no longer be the sole pursuit.
• ASIC solutions may gain more market recognition.
• Companies like WeiYing and YeCheng will welcome new opportunities