DeepSeek reportedly raises $15 billion in massive funding round

A massive capital injection

DeepSeek, a prominent Chinese AI laboratory, is reportedly securing a massive funding round of approximately $15 billion. This capital influx is intended to bolster the company's position as a leader in the competitive artificial intelligence space. The funding comes at a time when the company is rapidly expanding its technical capabilities and infrastructure.

According to reports from recodechinaai.com, the funding round has reached about RMB100 billion against a pre-money valuation of approximately RMB500 billion. While the final terms of the deal could still change, the scale of the investment highlights the immense interest from the private sector. The round is expected to reach completion within the coming days.

The investor list includes some of China's most influential corporations. Tencent and CATL are among the largest backers participating in the round. Other notable participants include Meituan, Geely, and GigaDevice, representing a broad coalition of industrial and technology giants.

Scaling production and efficiency

DeepSeek has already demonstrated significant technical scale with its current production systems. The company reports that a single cluster can serve 155 trillion tokens every day. This high volume of processing is essential for maintaining the performance of its large-scale models across various applications.

The lab is also focusing on the economics of AI inference, which is the process of running a trained model to generate outputs. DeepSeek reports generating $38 in revenue per GPU-hour, even when pricing is set at off-peak rates. This level of efficiency is a key metric for the sustainability of large-scale AI operations.

To support these goals, DeepSeek recently released DeepInfer, its internally developed inference framework. This framework is now available to other developers and supports multiple hardware types, including Nvidia’s SM100 Blackwell GPUs and Huawei’s Ascend 950 chips. By opening this framework, DeepSeek aims to facilitate broader adoption of its high-efficiency production methods.

Navigating global restrictions

Despite its rapid growth and funding, DeepSeek faces a complex regulatory environment in the United States. A Department of Home Security order recently barred covered DeepSeek products from federal civilian procurement. This move requires government agencies to remove the products from their systems, including certain third-party implementations.

In contrast, the regulatory landscape in China remains more open to model exports. China’s proposed AI export controls are currently under consultation and are expected to continue through November 15. These provisions are designed to facilitate open-source releases while excluding technology that is already in the public domain.

State media reports that DeepSeek and MiniMax have already completed the registration process for overseas licensing and open-sourcing. These steps suggest a strategic effort by Chinese labs to maintain international reach while navigating the specific restrictions imposed by different governments.

The broader Chinese AI landscape

DeepSeek's funding round is part of a larger trend of institutional support for AI in China. The State Council’s General Office recently issued guidance on 18 tasks across six areas to adapt employment policies to AI development. These measures aim to pair AI-driven job creation with protections for workers whose roles may be displaced by new technologies.

The government is also establishing new corporate entities to support the data infrastructure required for AI. The National Data Group was recently launched in Xiongan as a state-owned enterprise. This organization is tasked with developing and utilizing data resources to support the national AI agenda.

Simultaneously, other Chinese firms are making moves in the hardware and model space. For example, Ant announced open weights for its healthcare-focused Ling-3.0-flash-Sante model. These diverse developments show that while DeepSeek is a major player, the entire Chinese AI ecosystem is undergoing significant structural and financial shifts.

Infrastructure and hardware demand

The demand for the hardware required to run these models remains incredibly high. Reports indicate that ByteDance, Alibaba, and Tencent have ordered more than three million Nvidia RTX PRO 5500 cards. These are export-compliant workstation GPUs primarily intended for inference rather than training.

However, the supply chain for these chips remains constrained. Nvidia has reportedly stated it cannot guarantee fulfillment for these orders because the current demand exceeds planned production levels. This highlights a bottleneck where software capabilities are outpacing the physical availability of the necessary hardware.

Domestic production is also a key focus for Chinese companies seeking to reduce reliance on external hardware. Cambricon reported a 148% year-on-year profit increase to about $210 million in the third quarter. While sales were below some expectations due to output constraints, the company remains a significant part of the domestic compute landscape.