Meta Discloses Nearly $700 Billion in Future Commitments for AI Data Centers and Cloud Computing
Meta announced on July 30 local time that it has committed nearly $700 billion in future spending through long-term and short-term agreements, targeting AI data centers, cloud computing, and other areas. The company has signed irrevocable contracts totaling $349.3 billion, primarily covering third-party cloud agreements, servers, and network infrastructure. Meta noted this figure is a conservative estimate since, for agreements with variable terms, “we do not estimate total obligations beyond the minimum purchase quantities.” Additionally, the company holds $347 billion in lease commitments that have yet to commence and are therefore not reflected on its balance sheet. Notably, $68 billion was added in July alone, with related payments beginning in 2027 and 2028. These expenditures are separate from existing leases and mainly involve data centers, colocation facilities, and “specific network infrastructure.”
Anthropic in Talks for $15 Billion Funding to Build Texas AI Data Center, with Google Providing Guarantees and Chips
As of July 31, reports indicate that data center developer Nexus Data Centers is in advanced negotiations to raise approximately $15 billion for an AI data center project in Texas serving Anthropic, with Google offering financing guarantees and supplying chips. Google has agreed to provide billions of dollars in guarantees for Anthropic’s commitments if the company fails to meet lease and electricity payment obligations. The guarantee covers four data center lease agreements signed by Anthropic, as well as power purchase agreements linked to an on-site power plant. In return for supporting the project, Google is expected to receive roughly a 20% equity stake in the data center and power project. Furthermore, Anthropic plans to deploy Tensor Processing Unit (TPU) chips co-designed by Google and Broadcom at the facility, with chip procurement to be handled through a separate vendor financing agreement between Anthropic and Broadcom.
Qwen Undergoing Testing Inside Tesla Vehicles
On July 31, reports surfaced that the Doubao LLM assistant has been integrated into select new Tesla vehicles. Multiple insiders revealed to the news outlet that Qwen has also entered the deep-testing phase within Tesla’s China infotainment systems, with integration imminent.
According to these insiders, Qwen has already completed extensive on-vehicle testing in real Tesla environments, and capabilities such as “listening and responding, controlling the car, navigating, and handling tasks” are all part of the plan. Going forward, the features Qwen could deliver on Tesla’s infotainment system may far exceed the conventional definition of a “voice assistant.”
Tesla China Officially Integrates Doubao LLM into Vehicle Systems
On July 31, Tesla China began rolling out version 2026.14.13 of its vehicle software update in batches, covering the Model 3, Model Y, Model S, and Model X. The primary upgrade in this update is the integration of the Doubao LLM. According to the release notes, the Doubao LLM not only delivers real-time information but also enables natural, fluid conversations. This feature requires an active premium in-car entertainment subscription.
MiniMax H3 Officially Launched, with Model Weights to Be Released in the Coming Days
On July 31, MiniMax announcedthe official release of MiniMax H3. Described as a general-purpose full-modality generative model, it supports unified understanding of multimodal contexts composed of text, images, video, and audio, and can output audiovisual content with native dual audio channels, supporting up to 15 seconds at 2K resolution.MiniMax stated that it plans to open up the model weights in the coming days, subject to compliance with relevant laws and regulations.
ByteDance Unveils Seedance 2.5 Video Generation Model
On July 31, ByteDance announced the release of its next-generation video creation model, Seedance 2.5. According to the company,Seedance 2.5 can generate 30-second video clips in a single pass and supports multi-round extension; it allows users to input up to 30 images, 10 videos, and 10 audio clips as reference material in one go; and it supports precise timestamp control for targeted video editing. Seedance 2.5 is gradually rolling out on platforms like Jimeng AI and Doubao Professional Edition, with API services expected to go live on Volcano Ark in the near future.
DeepSeek-V4-Flash Official API Enters Public Beta
As of July 31, the official API for DeepSeek-V4-Flash has entered public beta. The model architecture and size of DeepSeek-V4-Flash-0731 remain consistent with DeepSeek-V4-Flash-preview, with only post-training being redone. It was noted that only the API interface for DeepSeek-V4-Flash has been upgraded; the DeepSeek-V4-Pro API and the APP/WEB endpoints remain unchanged. The official version of DeepSeek-V4-Pro is expected to be released shortly.
Amazon Invests $13.7 Billion in OpenAI’s Series C Preferred Stock in Q2
On July 31, a filing by Amazon revealed that over the six months ending June 30, the company invested a total of $28.7 billion in OpenAI’s Series C preferred stock, with $13.7 billion of that commitment funded during the second quarter.
