1. Google Plans Up to $40 Billion Investment in Anthropic, Intensifying AI Giant Competition
On April 26th, Google officially announced its plan to launch a strategic investment of up to $40 billion in Anthropic, an artificial intelligence unicorn. The investment adopts a “phased injection + performance binding” model — with a valuation of $350 billion, $10 billion in cash will be immediately injected into Anthropic, and the remaining $30 billion will be gradually added based on Anthropic’s subsequent performance achievements. Meanwhile, Google will also provide approximately 5 gigawatts of high-performance computing power support for Anthropic, covering the entire process of its large model training, inference, and product iteration to help it break through computing power bottlenecks.
Notably, just one week before the announcement of Google’s investment, Amazon had just announced an additional $25 billion investment in Anthropic. Together with their previous cooperation layout, Amazon’s cumulative investment in Anthropic has exceeded $40 billion. Currently, supported by two of the world’s top cloud service providers — Google and Amazon — Anthropic has maintained a strong business growth momentum, with its annualized revenue successfully exceeding $30 billion. Among them, its self-developed programming tool Claude Code has become the core growth engine. With its accurate code generation, debugging, and optimization capabilities, it has gained favor from numerous global technology enterprises and developers, and its market share has continued to rise. (Sources: Titanium Media, Cailian Press)
2. NVIDIA Blackwell Fully Compatible with DeepSeek-V4, Domestic Chips Follow Suit, Accelerating Large Model Computing Power Adaptation
On April 26th, NVIDIA officially announced that its next-generation computing power platform, Blackwell, has completed full compatibility with the DeepSeek-V4 series of large models, including two core products: DeepSeek-V4-Pro and DeepSeek-V4-Flash. According to actual tests, under the GB200 NVL72 hardware configuration, the single-user generation speed of both models has exceeded 150 tokens per second, an increase of more than 30% compared with the previous generation adaptation scheme, which greatly improves the practical application efficiency of large models.
It is reported that the two models, DeepSeek-V4-Pro and DeepSeek-V4-Flash, have distinct positioning: the former has a parameter scale of 1.6T/49B, focusing on high-precision and high-complexity task processing, suitable for scenarios such as scientific research and high-end enterprise services; the latter has a parameter scale of 284B/13B, focusing on lightweight and high-concurrency needs, and can be widely applied to high-frequency scenarios such as daily conversations and intelligent customer service. Both models support million-level context length, which can better meet complex needs such as long text processing and multi-turn conversations.
More notably, on the day of DeepSeek-V4’s release, Atlas 350 chip, owned by Huawei Ascend, a leading domestic chip manufacturer, also completed the first-time adaptation of the series models. This makes DeepSeek the world’s first large model enterprise to achieve dual adaptation of NVIDIA Blackwell platform and domestic Ascend chips at the time of model release, marking a technological breakthrough of China’s domestic chips in the field of large model computing power adaptation and promoting the diversified development of the AI computing power ecosystem. (Sources: NVIDIA Official Website, IT Home)
3. OpenAI Launches ChatGPT Workspace Agents, Expanding into Enterprise-Grade Workflow Automation Scenarios
On April 26th, OpenAI officially released its new product, ChatGPT Workspace Agents. With the core positioning of “enhancing team collaboration + workflow automation”, the product optimizes core capabilities such as intelligent collaboration, task automation, and cross-platform linkage to meet the actual needs of enterprise users. It can realize functions such as automatic meeting minutes generation, task assignment and follow-up, and cross-software data synchronization, greatly improving team work efficiency.
The launch of ChatGPT Workspace Agents marks that OpenAI’s business layout has officially expanded from general dialogue scenarios to enterprise-grade workflow automation scenarios, and it is also an important measure to respond to the growing demand of the enterprise-level AI market. Currently, with the in-depth penetration of artificial intelligence technology in enterprise scenarios, enterprises’ demand for automated collaborative agents is becoming increasingly urgent. OpenAI’s move not only enriches its own product matrix but also further promotes the application of AI technology in enterprise office, project management and other fields, intensifying the competition pattern of the enterprise-level AI market. (Source: AI Industry Daily)