According to Beating AI, Tencent's Hunyuan team reportedly took a week-long leave after releasing the Hy4 preview. The team's workspace, which was previously always occupied, was almost deserted this week.

This vacation is related to the release of the latest version of Hunyuan. On August 28, Tencent open-sourced Hy4 preview, which is currently the largest open-source model version of Hunyuan, with a total of 770B parameters, 49B activation parameters, and a context length of over 1M.

Prior to the release of Hy4 preview, the team worked intensively on the project for several consecutive months. After the version was launched and received internal approval, the company arranged for the team to take a break.

With the competition for large-scale models intensifying, long-term, high-intensity R&D by model teams has become the norm in the industry. Many teams immediately move on to the next round of development after releasing a model, making this arrangement of taking a concentrated break after completing a phase goal relatively rare.

According to BlockBeats, on September 1st, Goldman Sachs believes that the market's discussion on AI capital expenditure is far from over. As large tech companies continue to vie for computing power, land, electricity, and data center resources, the supply-demand imbalance for AI infrastructure may persist until the first half of 2028, meaning investors will continue to see higher capital expenditure guidance in the coming years.

Eric Sheridan, head of TMT research at Goldman Sachs, stated that there remains a significant gap between current AI computing power demand and available supply, with rising storage prices, chip prices, and data center construction costs. The pressure points across the AI industry chain have expanded from a simple GPU shortage to include memory, land, electricity, data center enclosures, and delivery cycles.