DeepSeek and Manus: Chinese Catalysts For The UK in The AI Arms Race?

The global AI landscape shifted dramatically with the release of DeepSeek R1 in January 2025 and the general agent Manus in early March, both of which were viral sensations.

With economic prosperity and national security at stake, it was a wake-up call for the west’s incumbents. But with the dust settling, might these Chinese startups actually help level the playing field for countries like the UK in the ongoing AI arms race?

 

DeepSeek: Open-Source AI at Fraction of The Cost

 

DeepSeek specialises in developing open-source large language models with remarkable cost-efficiency. Founded in 2023 by Liang Wenfeng and funded by High-Flyer (a Chinese hedge fund), DeepSeek R1 is a peer competitor with OpenAI’s ChatGPT and Anthropic’s Claude.

Its open-source approach distinguishes it from many competitors that maintain proprietary control over their models, despite their branding.

Draven McConville, the UK-based founder and AI investor, told us that DeepSeek could help give the UK’s AI startups a leg up.

“DeepSeek being open sourced could provide domestic researchers and businesses with a bleeding edge model they can freely access, modify and build upon without prohibitive costs.”

But what makes DeepSeek truly revolutionary is its cost structure. The company claims its V3 model was trained at a significantly lower cost and with less computing power compared to models like GPT-4 and Meta’s Llama 3.1. That’s great news for the UK, where large infrastructural projects typically take years, even decades to complete.

The LLMs of the incumbents have what the Japanese refer to as a hungry ghost’s appetite for energy.

“If models like R1 show us anything, it’s that the status quo can change and that previously accepted doctrines may need to be challenged”, says McConville. In other words, if new nuclear or SMR reactors are less of a roadblock, the UK can stay at the table.

DeepSeek achieves this cost-efficiency through innovative technical approaches, including the Mixture-of-Experts architecture. This design activates only the most relevant specialized sub-models for a given task, significantly reducing computational requirements. The models are also optimised to work efficiently on less advanced but more widely available hardware, such as Nvidia H800 GPUs. Necessity really is the mother of invention.

Manus Revelation: The Autonomous AI Agent

 

Developed by Chinese startup Monica and launched in March 2025, Manus is a general-purpose autonomous AI agent that goes beyond “traditional” chatbots. While chatbots primarily respond to prompts, Manus independently executes complex tasks across diverse domains without the need for constant human intervention.

The technology powering Manus involves a multi-layered architecture built upon Alibaba’s Qwen large language model with several fine-tuned models. It integrates with external tools like Browser Use, enabling the AI agent to navigate the web, interact with applications, and extract data like a human user. This integration-focused approach could offer the UK a faster route to enhance its AI capabilities by leveraging existing global advancements.

Its capabilities include resume screening, in-depth stock analysis with visualisations, personalised travel planning, and rapid website wireframing, design, and implementation. It’s early days for agentic AI, but the writing is on the wall: this degree of flexibility and reliability represents what might be the answer to the UK’s poor productivity (if adopted at scale).

Whether AI agents will replace white collar office jobs en masse or free up overworked managers to be more effective remains to be seen, but change is most certainly coming to the nation’s workforce.

 

Opportunities for the UK’s AI Ecosystem

 

The United Kingdom already holds a strong position in the global AI landscape, ranked as it is among the top nations worldwide in AI development and investment. Its ecosystem may not be as broad and deep as those in the US and China, but the UK does have some successes to point to such as DeepMind and the Alan Turing Institute.

DeepSeek’s cost-efficiency could particularly benefit the UK’s pursuit of sovereign AI capabilities. The ability to develop high-performing models at a fraction of the cost could enable UK firms and research institutions to build their own AI models tailored to national priorities, reducing reliance on American AI companies for critical public sector contracts.

And the autonomous capabilities of agentic AI could drive enhanced productivity across the UK economy. By automating complex tasks in areas like business operations, customer service, and data analysis, AI agents could free human capital for more strategic and creative endeavors.

Both technologies align with the UK’s national AI strategy of fostering an inclusive AI ecosystem and driving adoption across public and private sectors. They could potentially contribute to developing “national champions” in AI by empowering UK-based organisations to become leaders in specific domains.

 

Challenges and Considerations

 

Despite their potential benefits, adopting these technologies presents significant challenges for obvious reasons. Security and data privacy concerns are paramount given their origins in China. The UK government and businesses must carefully assess potential risks, especially for sensitive applications within government infrastructure.

Ethical considerations also require attention. Potential biases in training data and algorithms need identification and mitigation to ensure fair outcomes. The increasing autonomy of AI agents like Manus raises questions about employment impact and societal structures.

Successful adoption will depend on addressing existing challenges related to AI skills development, infrastructure investment, and public trust. Building transparency, addressing ethical concerns, and demonstrating safety will be crucial for effective integration into the UK’s economy.