In Lewis Carroll's Through the Looking-Glass, the Red Queen famously remarks to Alice, "Now, here, you see, it takes all the running you can do, to keep in the same place." This observation has since transcended literature, finding relevance in science and technology. It aptly captures a profound truth: in a constantly evolving environment, entities must adapt and evolve just to maintain their current position.
Did anyone anticipate that the front-runners in the AI and large language model (LLM) space would need to move this quickly?
The Rise of DeepSeek
We write this on 27 January after a notably turbulent day in the markets.
What happened? DeepSeek, a Chinese artificial intelligence company, has released a reasoning model, DeepSeek-R1, which has attracted considerable attention. The release coincides with the period between President Trump’s inauguration and the Chinese New Year, prompting speculation about potential political motivations.
What is DeepSeek?
This model:
Performs exceptionally well: Comparable in performance to OpenAI's latest O1 model, standing up to some of the toughest reasoning benchmarks.
Offers significant cost advantages: DeepSeek's API pricing is about 20% of OpenAI's, making it highly competitive.
Is fully open source: Released under the MIT licence, enabling broad accessibility and collaboration.
Was developed on fewer resources: DeepSeek claims development costs of approximately $6 million and the use of 10,000 NVIDIA GPUs – a fraction of the investment seen from industry leaders like OpenAI.
Market Impact
The introduction of DeepSeek-R1 has triggered a dramatic market reaction, with nearly $2 trillion in market capitalisation seemingly evaporating from the "Magnificent Seven" stocks (Apple, Microsoft, Alphabet, Amazon, Meta, Nvidia, and Tesla) in one day.
Why did these, and many other AI-related stocks, fall so sharply?
Lower Training Costs: DeepSeek trained its model using significantly fewer resources compared to Western companies.
Competitive Edge: The launch demonstrates that Chinese firms can compete effectively despite US technology restrictions.
AI Monetisation Concerns: With lower usage fees, DeepSeek raises questions about profitability for established players.
Take a Deep Breath
Although AI-related stocks fell sharply, the S&P 500 Equal-Weight Index, a proxy for the average stock, was down less than 0.5%, with four sectors trading higher at the time of writing. Markets have been grappling with extreme levels of concentration, and DeepSeek has encouraged a healthy rotation away from Big Tech. If there is a silver lining, the changes underway could lead to a more balanced near-term market.
In the long run, model commoditisation and cheaper inference – which DeepSeek has also demonstrated – benefit Big Tech. For example, a scenario where Microsoft provides inference to customers at a fraction of the cost means Microsoft spends less on data centres and GPUs, or, just as likely, sees dramatically higher usage due to reduced costs. Apple, Meta, and Amazon are also likely winners, gaining access to high-quality open source models at much lower costs.
A Whale in Nvidia’s Moat?
Nvidia is primarily known for its GPUs that power gaming, professional visualisation, data centres, and automotive markets. The company has positioned itself as a dominant player in AI computing, providing essential hardware for training and deploying AI models. While Nvidia is best known for the chips used to "train" or build a new AI system, it has stated that it now generates just as much revenue from chips for "inference" – processing user requests using a finished model.
Although the stock fell yesterday, the future isn’t all doom and gloom for Nvidia. As training costs decline, AI demand could increase, particularly for inference rather than training, which may benefit Nvidia. On the flip side, increased competition could lead to a loss of market share over time.
While volatility may persist as markets digest DeepSeek’s release, these developments are ultimately positive for the long-term health of AI. Innovation and cost reductions are integral to enabling mass adoption in the future.
Embracing the Red Queen's Wisdom
Perhaps the Red Queen's remark isn’t just a cautionary tale about stagnation.
In this evolving environment, businesses must prioritise agility and strategic thinking. The AI race is intensifying, but success will depend on direction as much as speed. This may prove to be an endurance race rather than a sprint.
As Alice asks the Cheshire Cat, "Would you tell me, please, which way I ought to go from here?"
Keep an eye out for our Q1 workshop, AI and Tech Literacy for Future-Focused Leaders, where the IoD Next Gen Committee will provide attendees with a basic understanding of core AI concepts and terminology, alongside an overview of recent developments and their significance.