Two significant positive developments are emerging in the artificial intelligence sector.
First, major technology firms, including Alibaba, are reportedly in discussions to invest in AI startup DeepSeek. This potential investment could value the company at over $20 billion. According to sources, this would represent DeepSeek's first external funding round. Previous reports indicated that DeepSeek was initially seeking to raise at least $300 million with a valuation of no less than $10 billion. However, following strong preliminary interest from investors, the company is now pursuing a valuation exceeding $20 billion.
There are indications that DeepSeek's valuation could potentially surpass $20 billion. For context, two of its publicly-listed competitors, Shanghai-based MiniMax and Beijing-based Zhipu, were valued below $10 billion during their Hong Kong listings in January. Zhipu's valuation has since surpassed $50 billion, while MiniMax's valuation exceeds $30 billion.
DeepSeek gained global attention last year with the launch of its R1 model, which achieved performance close to top-tier models at a significantly lower training cost compared to its U.S. counterparts. The company had previously maintained a stance against accepting external capital. However, this position is changing due to increasing technical challenges and competitive pressures. The current market is more intense than a year ago, characterized by faster model iterations, higher training and operational costs, and intensified competition from well-funded tech giants in both China and the United States.
In pre-market U.S. trading, shares of Alibaba experienced a sharp upward movement and are currently up over 1%.
Second, Google has unveiled its latest generation of Tensor Processing Units (TPUs), which are proprietary chips designed to make AI computing services faster and more efficient. Announced at the Google Cloud Next conference, the new TPU generation will be available in two versions: TPU 8t, primarily for training AI software, and TPU 8i, for running AI services after model completion, a phase known as inference.
In an industry long dominated by NVIDIA, Google has become one of the few companies to successfully develop its own AI chips. Demand for TPUs has been strong in Silicon Valley in recent months, and Google aims to extend its advantage with this new generation. This initiative is part of Google's broader AI strategy to reduce deployment costs and energy consumption for AI software while improving service response times. The new TPUs store more data on the chip itself to enable the "faster response" users expect, although the growing complexity of software layers continues to drive demand for computing power.
Mark Lohmeyer, Vice President of Compute and AI Infrastructure at Google, stated, "The core objective is to achieve the lowest possible latency response at the lowest possible cost. The number of transactions is growing rapidly, and to achieve scale, the cost per transaction must decrease significantly."
This news also contributed to a sharp rise in Google's share price during pre-market trading.
Comments