China's DeepSeek has released its V4 model, and Wall Street is paying close attention. In a report published April 24, Goldman Sachs described the launch as a milestone comparable in significance to the startup's earlier breakthroughs — one that could reshape the competitive landscape between Chinese and American AI developers.
The bank's report, titled "Navigating China Internet: DeepSeek V4 Launch — Implications for China AI Models/Cloud/Data Centers," found that V4 performs on par with state-of-the-art models currently available in the United States, while requiring roughly 7 to 10 percent less memory. Goldman Sachs said that combination could substantially lower deployment costs and accelerate broader AI adoption worldwide.
How DeepSeek V4 Improves on Its Predecessor
Compared with the previous V3 model, V4 delivers meaningful gains in memory efficiency, according to the Goldman Sachs report. The new model is also built to run on domestically produced Chinese chips — a notable shift that carries pricing implications.
DeepSeek has said it expects the Pro version's cost to fall sharply from the second half of 2026 onward, contingent on the mass availability of Huawei's new Ascend 950 chip. The company's ability to deliver competitive performance on domestic hardware underscores a broader trend of Chinese AI development advancing despite U.S. export restrictions on advanced semiconductors.
Why Goldman Sachs Sees This as a Turning Point for China AI
Goldman Sachs framed the V4 release not just as a product update but as a signal of accelerating competition within China's AI sector. As the pace of new model releases quickens, the bank said rivalry among Chinese AI developers is set to intensify — and DeepSeek's open-source V4 is not immune to that pressure.
Looking ahead, Goldman Sachs identified coding performance, task-completion success rates, and multimodal capabilities as the key factors likely to determine pricing power for next-generation models.
Cloud and Data Center Stocks Positioned to Benefit
On the infrastructure side, Goldman Sachs recommended that investors maintain exposure to China's cloud and data center sub-sectors. The bank highlighted GDS, VNET, Alibaba Cloud, and Kingsoft Cloud as companies well-placed to benefit as enterprise adoption, AI agent deployment, and consumer AI assistant usage drive rising demand for compute capacity and push cloud pricing higher.
"We believe improvements in computing cost efficiency will further drive broader adoption, exploration, and popularization of AI applications," the bank said.
Big Tech Eyes DeepSeek as Valuation Tops $20 Billion
For China's large internet conglomerates, Goldman Sachs noted that strong operating cash flows give them a structural advantage in pursuing AI infrastructure and cloud opportunities. However, the bank flagged the challenge of competing with independent, AI-native developers for top talent, and suggested these companies will need to sharpen incentive structures to attract and retain leading researchers and engineers.
Separately, Goldman Sachs noted that according to media reports, Chinese tech giants Tencent and Alibaba may be in talks to invest in DeepSeek at a valuation exceeding $20 billion. （Related： Opinion | Da Vinci Had AI? He'd Have Built a Guild. So Should We. ｜ Latest ）
