What is the hype about Deepseek?
DeepSeek, a Hangzhou-based Chinese AI company, was founded by a quantitative hedge fund company High-Flyer in 2023. The company focuses on AI research and has been building open-source large language models (LLMs) optimized for limited compute resources (combination of NVIDIA H800 and Huawei Ascend 910C chips). The top proprietary models from leading AI companies – Open AI, Google, Anthropic – are following the “bigger the better” playbook, relying on massive energy and compute infrastructure (GPU chips and data centers). With the small difference that Meta is partially open source, these models are trained on large data sets in a closed environment. DeepSeek’s release of efficient and high-performance models, comparable to the US top AI models, not only rattled the established players but also precipitated geopolitical contestation between China and the US. This contest is a newer iteration of the ongoing race to gain supremacy in the global AI landscape.
Redefining AI Innovation:
While DeepSeek is not reinventing the wheel and is broadly follows the same agenda, it has relied more heavily on reinforcement learning (RL) and mixture-of-experts (MoE) methods and refined chain-of-thought (CoT) reasoning very effectively. It is widely reported that DeepSeek has delivered its R1 reasoning model at a fraction of the cost of leading models – approximately $5.5 million, compared to the hundreds of millions dollars spent on leading proprietary models by the companies like Open AI and Anthropic. Deepseek has alleviated concerns that Gen AI development requires massive investments in compute infrastructure and energy resources. This could force the US AI industry to rethink its investment strategy and shift from the scaling at-all-costs-framework towards efficient and innovative resource optimization.
One might argue that the methods such as MoE and RL methods were developed in academic research decades ago; and some, like transformer models and CoT reasoning, have been introduced and used by leading tech firms in the US and Europe. Nevertheless, DeepSeek has effectively combined these techniques, as published in their research paper. It remains to be seen whether Chinese firms and academia can truly come up with the next set of game-changing techniques, products and approaches in AI development.
Figure 1. (Source: Huggingface.co )
This pivotal moment presents an opportunity to redefine public sector innovation by investing in open-source AI capabilities built on sovereign, adaptable models like DeepSeek. This strategic move would not only enhance digital sovereignty but also unlock unprecedented value creation, positioning the GCC as a global AI leader.
Why Open-Source AI? The Case for Digital Sovereignty
Relying on proprietary AI systems from foreign vendors introduces risks, limited transparency, dependency on external updates, and vulnerabilities in data governance. By contrast, open-source AI frameworks like DeepSeek empower nations to control their technological destiny.
- Data Security & Compliance: Open-source models enable governments to audit algorithms, ensuring alignment with local regulations (e.g., UAE’s Data Law and Saudi’s NDMO standards). This mitigates the risks of third-party data exploitation.
- Customization: Open-source models like DeepSeek’s architecture allows customization for Arabic language nuances, cultural contexts, and sector-specific needs (e.g., healthcare, education, smart cities), fostering citizen-centric services.
- Cost Efficiency: Eliminating licensing fees and reinvesting savings into local AI talent pipelines creates a self-sustaining ecosystem.
- Sustainability: Supports sustainable AI development by leveraging energy-efficient models and enabling reusability of AI resources.
For nations prioritizing economic resilience in their long-term vision, open-source AI is a cornerstone of strategic autonomy.
Value Creation Through Local Ecosystems
Building open-source AI is not just a technical endeavour—it’s an economic catalyst.
- Talent Development: Collaborate with universities to train a new generation of Arab AI engineers, reducing dependence on expatriate expertise.
- Public-Private Partnerships: Support startups in developing solutions on sovereign AI platforms, fostering innovation in sectors such as fintech and logistics.
- Global Leadership: Export UAE- and Saudi-developed AI tools to MENA and Global South markets, positioning the GCC as a hub for ethical, culturally aware AI.
Strategic Advantages of DeepSeek
DeepSeek’s architecture aligns with GCC ambitions:
- Efficiency: Optimized for high performance with lower computational costs, making it ideal for large-scale public sector adoption.
- Transparency: Open-source code allows scrutiny, fostering public trust in AI-driven governance.
- Interoperability: Integrates with existing digital infrastructure (e.g., UAE’s “Gov 4.0” platforms), maximizing ROI.
The Path Forward
To capitalize on this opportunity, the UAE and Saudi Arabia should:
- Launch National AI Labs: Build in-house expertise to customize and maintain open-source models.
- Establish Governance Frameworks: Define standards for ethical AI use, data privacy, and cross-agency collaboration.
- Foster Regional Alliances: Facilitate knowledge sharing of best practices and tools across GCC nations to build a unified AI ecosystem.
The Missing Components:
Deepseek provides only open model weights for their R1 model and does not disclose details on the data that the model was trained, the training code and even scaling laws. These components are important for other players to develop similar models for their local socioeconomic contexts in the open-source environment.
How Can Frost & Sullivan help in navigating the complex terrain of rapidly evolving Generative AI landscape:
Our expertise in AI and policy innovation can help governments and public sector entities to identify key opportunities and develop short term and long-term strategic roadmaps by navigating the complex terrain of the GenAI landscape.
- Shaping sovereign AI industrial policy for the governments across the region
- Developing an open-source AI development strategy to foster local open AI innovation ecosystem
- AI skills and competency gap assessment for sovereign AI
- Developing government’s strategy for AI agents with most plausible use cases for different functions and government departments
- Identifying fit-to-purpose low-cost open-source use case prioritization by developing LLMs and SLMs for local contexts (in the areas of healthcare, education, culture promotion, research etc.)
- Differentiating AI snake oil from real value deriving solutions – AI induced hype, risks and threat perceptions and awareness
Conclusion
In a world where AI dominance is synonymous with geopolitical influence, open-source capabilities are the region’s gateway to digital self-reliance, AI sovereignty and global leadership. By investing in open-source AI solutions like Deepseek, governments and other public sector entities can future-proof their agencies, foster innovation, and leverage sovereignty into a comparative advantage. The players in GCC can collaborate with other open-source AI players like Deepseek, Alibaba’s Qwen, Meta and Mistral, to advance their open-source AI capabilities and orchestrate innovation ecosystem with greater efficiency.
Co-authored by: Saurabh Verma, Vice President, ICT Growth Advisory, Frost & Sullivan and Sreekanth Mukku, Independent Consultant and AI Policy Expert
