Boom or Bust? What DeepSeek Means for AI's Future

January 2025
IoT & Emerging Technology

Last week, Chinese AI company, DeepSeek, launched its open-source AI model for advanced reasoning, problem-solving and decision-making.

Compared to the current global leader in AI, OpenAI, there are notable differences in its development and capabilities. For example, Open AI’s solutions are built using proprietary models requiring more computational power. As a result, considerably less investment was needed to create DeepSeek-R1, compared to OpenAI’s models. OpenAI reportedly spent $5 billion alone in 2024, whilst the reported cost to develop DeepSeek-R1 was less than $6 million; this represents 0.1% of OpenAI entire spend in the year.

Why the AI Bubble Hasn't Burst…

The term ‘bubble’ refers to the market hype and overvaluation of companies developing AI solutions, which results in a subsequent drop in a company's value. While Nvidia’s share price dropped 17% - equivalent to a $600 billion loss in a single day, after DeepSeek’s announcement, it’s essential to understand that stock prices can be highly reactive. Despite being a leader in AI, Nvidia generates revenue from multiple sectors, and the long-term impact of any potential reduction in AI investment remains uncertain.

While we believe AI development companies are overvalued, the metaphor of a bubble bursting no longer applies to most technology markets, even those that can be considered emerging, such as AI. It is also important to remember that AI is an enabling technology that can enhance other services through automation and not a technology that often provides direct revenue.

Most companies offering AI-based solutions will be users of AI themselves, not AI developers. The number of companies developing AI is relatively low, given the specialist knowledge required, with many users. From Juniper Research’s extensive analysis of various AI use cases, we understand that there is already sustained demand for AI-based services. Indeed, the ability of AI to learn and automate various tasks leads to the technology having a vast total addressable market (TAM) of users, and we expect sustained growth of this over the next few years.

 
Is There Even a Bubble to Burst?

Since the dot-com bubble burst in the late 1990s, technology in general has become increasingly pervasive, with large technology vendors such as Microsoft, AWS and IBM developing AI alongside other core areas of their portfolio. Whilst there are similarities between the AI bubble of today and the dot-com bubble of the 1990s, such as rapid increases in company valuations and notable innovative technologies, there is a remarkable difference that set it apart: the dot-com bubble was far more speculative, with investment often coming with inflated expectations for immediate growth.

As a technology, AI has been in development for several years, but has only found commercially viable use cases in the last few years. Over the previous 24 months, these have risen from automation through robotic process automation (RPA), content creation through generative AI, and substantial degrees of self-management through agentic AI.

Juniper Research believes that technology is so well-established that many governmental and supranational bodies are rushing to implement regulations around the use of AI; given the rapid uptake of the technology amongst various industries and the potential impact that it can have after years of digital transformation. This is a technology whose rate of innovation is outpacing the rate at which legislation can be implemented and, to an extent, outpacing the understanding of its limits.

This position is similar to the development of facial recognition technology and blockchain technology during their initial steps to mass market adoption, with these technologies still being widely used today - albeit with slightly more regulation on their use. So, the question is not when or if the AI bubble will burst, but if there is a ‘bubble’ to burst at this point. The answer is that AI technology, wide as its use cases and implementation are, is already a technology so well-entrenched in various industries that users of digital products and enterprises should be expecting decreases in the cost of AI.

The shock to the market is not the decrease in the cost of AI claimed by DeepSeek, but the reduction scale in such a short time. This ‘shock’ will be temporary, and we expect Nvidia’s share price to eventually recover, but the long-lasting impact on the market is the potential for cheaper AI-based solutions in the future. With this decrease in cost, we can expect the market to become segmented as AI further specialises and different degrees of quality from AI solutions emerge.


As VP of Telecoms Market Research at Juniper Research, Sam produces high-quality research on telecommunications technologies and the future of digital content. His recent reports include Future Leaders 100: Telecoms 2025, Direct-to-Cell, and Operator Revenue Strategies Sam has been interviewed by leading media outlets, including the BBC and Wall Street Journal, and is a regular contributor to messaging conferences and telecommunications industry events.

Latest research, whitepapers & press releases