DeepSeek V3 Update Leads a New Paradigm in AI: Algorithm Optimization and Computing Power Collaborative Development

DeepSeek recently released the V3 version update on Hugging Face——DeepSeek-V3-0324, with model parameters reaching 685 billion, showing significant improvements in code capabilities, UI design, and inference capabilities.

At the recently concluded 2025 GTC conference, an executive from a technology company highly praised the achievements of DeepSeek and emphasized that the market's previous belief that DeepSeek's efficient model would reduce the demand for chips was incorrect. He pointed out that future computing power demands will only increase, not decrease.

DeepSeek, as a representative product of algorithm breakthroughs, has sparked reflections on the role of Computing Power and Algorithm in industry development due to its relationship with chip supply.

The Symbiotic Evolution of Computing Power and Algorithm

In the field of AI, the enhancement of Computing Power provides a running foundation for more complex Algorithms, enabling models to handle larger amounts of data and learn more complex patterns; while the optimization of Algorithms can utilize Computing Power more efficiently, improving the usage efficiency of computational resources.

The symbiotic relationship between Computing Power and Algorithm is reshaping the AI industry landscape:

1. Technical route differentiation: Some companies pursue the construction of super-large Computing Power clusters, while others like DeepSeek focus on optimizing Algorithm efficiency, forming different technical schools.

2. Industry Chain Restructuring: A chip company has become a leader in AI Computing Power through its ecosystem, while cloud service providers reduce deployment thresholds through elastic Computing Power services.

3. Resource allocation adjustment: Enterprises seek a balance between hardware infrastructure investment and efficient algorithm research and development.

4. The Rise of Open Source Communities: Open source models such as DeepSeek and LLaMA enable the sharing of algorithm innovations and Computing Power optimization achievements, accelerating technological iteration and diffusion.

Technical Innovations of DeepSeek

The success of DeepSeek is closely tied to its technological innovations. Below is a brief explanation of its main innovations:

Model Architecture Optimization

DeepSeek adopts a combination architecture of Transformer + MOE (Mixture of Experts) and introduces a Multi-Head Latent Attention mechanism (MLA). This architecture functions like an efficient team, with the Transformer handling general tasks, MOE acting like an expert group addressing specific issues, and MLA allowing the model to flexibly focus on important details.

Training Method Innovation

DeepSeek has proposed the FP8 mixed precision training framework, which can dynamically select the appropriate computing power based on training needs, improving training speed and reducing memory usage while ensuring model accuracy.

Improvement in Inference Efficiency

DeepSeek introduces Multi-token Prediction (MTP) technology, which allows for the prediction of multiple Tokens at once, greatly accelerating inference speed and reducing costs.

Reinforcement Learning Algorithm Breakthrough

The new reinforcement learning algorithm GRPO (Generalized Reward-Penalized Optimization) optimizes the model training process, achieving a balance between performance improvement and cost reduction while minimizing unnecessary Computing Power.

These innovations have formed a complete technological system that comprehensively reduces Computing Power requirements from training to inference, enabling ordinary consumer-grade graphics cards to run powerful AI models, significantly lowering the threshold for AI applications.

Impact on Chip Supply

DeepSeek optimizes algorithms through the PTX (Parallel Thread Execution) layer of a certain chip company, achieving more precise performance tuning. The impact on chip suppliers is twofold: on one hand, DeepSeek is more deeply integrated with hardware and the ecosystem, and the lowering of the AI application threshold may expand the overall market size; on the other hand, algorithm optimization may change the market demand structure for high-end chips, with some AI models that originally required high-end GPUs now potentially running efficiently on mid-range or even consumer-grade graphics cards.

The Significance of AI Industry in China

DeepSeek's algorithm optimization provides a technological breakthrough path for China's AI industry. In the context of high-end chip constraints, the idea of "software compensating for hardware" reduces reliance on top imported chips.

Upstream, efficient algorithms reduce the pressure on computing power demand, allowing computing power service providers to extend hardware usage cycles and improve return on investment through software optimization. Downstream, the optimized open-source models lower the barriers to AI application development, enabling numerous small and medium-sized enterprises to develop competitive applications based on the DeepSeek model, fostering more vertical AI solutions.

The Profound Impact of Web3+AI

Decentralized AI Infrastructure

The algorithm optimization of DeepSeek provides new momentum for Web3 AI infrastructure. The MoE architecture is suitable for distributed deployment, allowing different nodes to hold different expert networks, eliminating the need for a single node to store the complete model, thus reducing the storage and computing requirements of a single node. The FP8 training framework further reduces the demand for high-end computing resources, enabling more computing resources to join the node network.

Multi-Agent Systems

1. Intelligent Trading Strategy Optimization: By analyzing market data, predicting price fluctuations, and executing on-chain trades through the cooperation of multiple agents, it helps users achieve higher returns.

2. Automated execution of smart contracts: Achieving automation of complex business logic through the collaborative operation of agents for contract monitoring, execution, and result supervision.

3. Personalized Portfolio Management: AI helps to find the best staking or liquidity provision opportunities in real-time based on the user's risk preferences, investment goals, and financial situation.

DeepSeek seeks breakthroughs through algorithm innovation under computing power constraints, opening up differentiated development paths for China's AI industry. Lowering application barriers, promoting the integration of Web3 and AI, reducing reliance on high-end chips, and empowering financial innovation, these impacts are reshaping the digital economy landscape. The future development of AI will no longer be just a competition of computing power, but a competition of collaborative optimization between computing power and algorithms. On this new track, innovators like DeepSeek are redefining the rules of the game with Chinese wisdom.