The name DeepSeek has emerged seemingly from nowhere within the last week – and its newly launched AI tool appears to have blindsided many of the biggest names in tech and AI, wiping off an unprecedented US$589 billion from Nvidia’s valuation in a single day. What are the implications of the new launch for cloud infrastructure providers (hyperscalers), semiconductor players and wider global markets?
Hyperscalers and overcapacity risks
For hyperscalers, there is a potential risk of overcapacity in the short term due to the extensive data centre build-out. Many customers are still not fully utilising cloud services, which means the extra volume may not be needed immediately. This could lead to poor utilisation rates and impact return on capital employed (a measure of a company’s profitability and capital efficiency) in the short term. However, hyperscalers may cut back capex spending to counteract this.
Semiconductor equipment players
For companies like Nvidia and ASML, there could be some short-term demand fluctuations, but the long-term outlook is likely positive. Chip manufacturing companies tend to have a long-term focus and the improved performance that DeepSeek’s new architecture might confer is likely to be a positive over the long term. Despite the hit to its market value this week, Nvidia recently stated that it views the new models as an AI advancement, categorising DeepSeek’s innovation as an opportunity for increased demand for graphics processing units. Albeit, Nvidia might not be the most impartial judge of this.
Capex announcements and market reactions
Meta recently released high capital expenditure guidance, even after the DeepSeek model went live. Other companies have also made capital expenditure announcements that account for DeepSeek’s new model, indicating a commitment to higher spending even in light of DeepSeek’s innovation. Another response that might be likely is a push by US lawmakers for tighter chip controls on China, which could impact the market.
Lowering barriers to entry
One of the risks associated with DeepSeek is the lowering of barriers to entry in Large Language Model (LLM) creation. While this is beneficial for most companies, it could pose challenges for larger companies that previously benefited from the inaccessibility of LLM creation. This is likely more of a concern for companies like OpenAI rather than the hyperscalers.
Long-term outlook
Overall, the balance of probability leads up to believe DeepSeek may well be positive. Short-term demand might be impacted but the fall in the cost of large language model creation should make far more AI projects possible. Hyperscalers can protect themselves by reducing capital expenditure if needed but the bottom line is they now have more efficient data centres. We now focus on the slew of tech results due over the next fortnight to gauge how companies are reacting.
The value of securities and their income can fall as well as rise. Past performance should not be seen as an indication of future results. All views expressed are those of the author and should not be considered a recommendation or solicitation to buy or sell any products or securities.
