According to individuals with inside knowledge of the situation, Nvidia’s ambitious plan to pour up to $100 billion into OpenAI, aimed at bolstering the training and operation of its cutting-edge AI models, has hit a roadblock. This setback arises from internal skepticism within Nvidia. The two tech giants officially unveiled their partnership in September of the previous year. Under this agreement, Nvidia was tasked with constructing a computing infrastructure boasting at least 10 gigawatts of power capacity to cater to OpenAI’s needs and investing in the enhancement of its computational capabilities. In return, OpenAI pledged to lease Nvidia’s chips for its operations. At present, both entities are engaged in a thorough reevaluation of the trajectory of their collaborative efforts.

On January 30, it was reported that the super political action committee 'Leading the Future,' which supports the development of artificial intelligence, has completed fundraising of over US$125 million. Greg Brockman, former president and co-founder of OpenAI, Joe Lonsdale, founder of 8VC, and Silicon Valley venture capital firm Andreessen Horowitz (a16z) have all contributed to the donations.

Individuals with in - depth knowledge of the situation have disclosed that Nvidia initially had grand plans to pour up to $100 billion into OpenAI. This substantial investment was aimed at bolstering OpenAI's efforts in training and running its cutting - edge AI models. Nevertheless, the collaboration pact has ground to a halt, primarily due to concerns voiced by certain figures within Nvidia.

Back in September of the previous year, the two tech giants made an official announcement regarding their partnership and inked a memorandum of understanding. At that time, it was anticipated that the negotiation process would wrap up within a matter of weeks. However, since then, there has been a notable lack of significant progress.

At present, both parties are engaged in negotiations centered around Nvidia's proposed equity investment, which amounts to tens of billions of dollars, as part of OpenAI's ongoing funding round. Jensen Huang, the CEO of Nvidia, clarified that the original agreement was non - binding. Moreover, both Nvidia and OpenAI have expressed their keenness to continue collaborating in the future.

On the evening of January 29, Pang Tianyu, a PhD holder from the Department of Computer Science at Tsinghua University, shared on social media that he would be joining Tencent. He is scheduled to officially assume his new position on February 4, serving as the Chief Research Scientist for the Hunyuan Large Model Team. In this capacity, he will spearhead the development of reinforcement learning technology tailored for multimodal models, initially concentrating on pioneering research into state-of-the-art algorithms within the realms of generative models and reinforcement learning.

Pang Tianyu boasts an outstanding research background, having previously contributed his expertise at Sea AI Lab in Singapore. His prolific output includes over 70 papers published in prestigious conferences such as ICML and NeurIPS. His exceptional contributions have earned him numerous accolades, including the Microsoft Research Fellowship and the NVIDIA Pioneer Academic Award, among others.

Recently, several AI researchers and executives from Apple's Siri team have left the company to join competitors such as Meta and Google.

On the tech podcast TBPN, Bloomberg journalist Mark Gurman divulged that when Apple was in the process of choosing a technology partner for the next - generation Siri, it gave serious consideration to using Anthropic's Claude large language model. In the highly competitive tech landscape, this was a significant decision - making point.

Ultimately, though, Apple made the announcement that the new iteration of Siri would be built on Google's Gemini platform. This choice was mainly shaped by cost - related negotiations. In the business world of tech giants, cost - effectiveness plays a crucial role in determining partnerships.

Apple and Google have forged a multi - year cooperation agreement. Under this pact, Google's Gemini model will serve as the foundational support for Apple's intelligent features, which of course includes the new version of Siri. This kind of long - term collaboration is not uncommon in the tech industry, where companies often join forces to leverage each other's strengths.

The new version of Siri is anticipated to make its debut alongside iOS 26.4 in 2026. It is set to boast more robust personalized interaction capabilities. In today's digital age, users expect highly customized experiences from their devices, and this enhancement aims to meet those expectations. Moreover, it will support multimodal interaction, allowing users to interact with Siri through various modes such as voice, touch, and perhaps even gesture in the future. Additionally, it will be capable of ultra - long text processing, a feature that can be extremely useful for tasks like drafting long documents or analyzing extensive reports.

In this collaborative endeavor, Apple has adopted a paid licensing model. It will pay approximately $1 billion annually to acquire the usage rights of Google's Gemini model. This move is a strategic one for Apple, as it enables the company to swiftly make up for the shortcomings in its self - developed large models. By doing so, Apple can keep up with the ever - growing user demands for intelligent and seamless experiences on their Apple devices.

The pro-artificial intelligence super political action committee (Super PAC) named "Charting the Future" has successfully secured over $125 million in funding. Its primary objective is to exert influence on the 2026 U.S. midterm elections and shape federal policies concerning AI regulation. Notably, OpenAI's president, the founder of 8VC, a16z (a prominent venture capital firm), and other influential figures have already made investments. The founder of SV Angel and the AI startup Perplexity have also joined the ranks of investors. The tech industry giants are eager for Washington to institute cohesive, industry-friendly federal standards, which would serve as a unified framework to replace the current patchwork of state-level regulations.

At the close of last year, NVIDIA unveiled the DGX Spark mini AI supercomputer. This innovative system is powered by a custom-designed GB10 super SoC processor, which seamlessly integrates a 20-core CPU based on the ARM architecture with a GPU that leverages the Blackwell architecture. The processor supports a substantial 128GB of unified memory and enables dual-machine clustering, allowing it to scale up and handle models with up to 405 billion parameters. Recently, Jensen Huang, NVIDIA's CEO, disclosed that the company is partnering with MediaTek to develop the energy-efficient N1 series of SoC processors. These processors will be crafted using TSMC's advanced 3nm process technology. The CPU segment of the N1 series will feature a 20-core ARM architecture, comprising 10 Cortex-X925 cores and 10 Cortex-A725 cores. Meanwhile, the GPU will incorporate the Blackwell architecture, delivering FP4 performance that can reach up to 1 PFLOPS. Additionally, the processor will be equipped with fifth-generation Tensor cores and fourth-generation RT ray-tracing cores, specifically targeting desktop-level AI computing scenarios. However, the exact launch date for the N1 series remains undisclosed at this time.

Microsoft once basked in a moment of glory, a feat largely attributable to CEO Satya Nadella's visionary foresight. At the close of 2022, the advent of ChatGPT presented Microsoft with a golden opportunity. Nadella, with unwavering determination, invested in OpenAI, facilitating the seamless integration of cutting-edge AI models into Microsoft's software ecosystem. These models were then exclusively offered through the Azure cloud platform, giving Microsoft a significant head start.

However, the once-cordial relationship between Microsoft and OpenAI began to cool. Subsequently, Microsoft redirected its investment towards Anthropic. In the meantime, OpenAI forged partnerships with Oracle, Google, and Amazon. Microsoft's strategy of casting a wide net, while effective in diversifying risks, has inadvertently diluted its initial advantage in the AI arena. Since reaching its peak stock price in October of the previous year, Microsoft's shares have witnessed a cumulative decline of roughly 11%. By early 2026, Microsoft's stock price had been trading in a sideways pattern. Following the release of the earnings report, after-hours trading took a nosedive, plummeting by 8.1%, primarily due to capital expenditures surpassing market expectations. Despite total sales exceeding forecasts and profits soaring by nearly a quarter, these positive developments were insufficient to counterbalance the adverse impact of the unexpectedly high capital expenditures.

In the current competitive landscape of the AI field, the dynamics have shifted dramatically. Google has made aggressive strides in consumer application scenarios with its Gemini model, while Anthropic's prowess in code generation stands at the pinnacle of the industry. For Microsoft to reclaim its former glory and assert itself as a formidable contender, it must emulate Google's approach and seek a breakthrough.

On January 30, Ali QianWen made an official announcement regarding the launch of the DeepPlanning benchmark. This particular benchmark has been meticulously crafted to assess an Agent's prowess in global planning when confronted with real-world, intricate scenarios. These scenarios encompass tasks such as multi-day travel planning and shopping for multiple products. Currently, the benchmark has been made openly accessible as an open-source resource on both Hugging Face and ModelScope platforms.

Music publishers, including Concord Music Group and Warner Music Group, have initiated legal action against AI company Anthropic, alleging that the firm unlawfully downloaded in excess of 20,000 copyright-protected musical compositions. The damages claimed in the lawsuit could potentially exceed $3 billion, which, if realized, would make this one of the most substantial non - class - action copyright lawsuits in the annals of U.S. legal history.

It's worth noting that Anthropic has previously faced legal challenges from novelists. In those cases, while the judge determined that using copyrighted content to train models was within legal bounds, acquiring such content through piracy was deemed illegal. As a result of those earlier lawsuits, Anthropic was slapped with a $1.5 billion fine. On average, each affected writer received a compensation of $3,000 per work.

On January 29, 2026, the AI intelligent travel platform Hongtu Zhixing App was officially launched. As the officially designated tourism and travel service platform of Dreame Technology, it enables precise itinerary planning, real-time crowd warnings at tourist attractions, and personalized experience recommendations. Its services cover the domestic high-net-worth market and cross-border services, offering users tailored multi-scenario solutions and providing a full-scenario closed loop from itinerary planning to destination services.

In a mere three days after its launch, Kimi K2.5 has swiftly captured the hearts of developers and users across the globe. On the well - known overseas programming tool platform Kilo Code, its call volume has eclipsed that of all its competitors, securing the top spot.

On the globally renowned large - model aggregation platform OpenRouter, its call volume also stands out, ranking among the top three worldwide. It trails only behind Claude Sonnet 4.5 and Gemini 3 Flash.

Moreover, the introduction of K2.5 has spurred significant growth. Kimi's overseas revenue has now surpassed its domestic revenue. And the number of paying users worldwide has surged, multiplying four times over.

On January 24th, the China Judicial Artificial Intelligence Conference (CJAI2026) convened in Shanghai, where the Internet Judicial Research Institute of Tsinghua University officially launched LegalOne-R1. This innovative open-source legal large language model (LLM) was developed leveraging Huawei Cloud Ascend AI cloud services. LegalOne-R1 is available in three parameter configurations: 1.7 billion (1.7B), 4 billion (4B), and 8 billion (8B), catering to diverse computational needs.

As a generative legal reasoning model, LegalOne-R1 has undergone extensive training on authentic Chinese judicial data. Its development incorporated a sophisticated "mid-stage training + post-training" dual-phase enhancement approach, enabling the model to acquire a comprehensive array of legal capabilities. This meticulous process ensures that LegalOne-R1 excels in real-world scenarios, delivering unparalleled stability, accuracy, and practicality.

The model's development was generously supported by the "Huawei AI Hundred Schools Initiative" and esteemed industry collaborators. Notably, the LegalOne-R1-8B variant demonstrated exceptional legal acumen on public evaluation datasets, outperforming existing open-source models in key tasks and setting a new benchmark in the field.

LegalOne-R1 has been fully optimized for training and inference on the Huawei Ascend computing platform, utilizing the MindSpore framework. On January 23rd, the model's parameters were made publicly available on an open-source platform, accompanied by a comprehensive technical report and an application guide to facilitate seamless integration and utilization.

Open-source access: https://huggingface.co/CSHaitao/LegalOne-1.7B.

Xiaohongshu is discreetly carrying out internal testing for a novel 'Voice Q&A' function. Some users have observed that the 'Trending' tab, previously located beneath the search bar in the app, has now been substituted with a dedicated button for this new feature. When users pose questions using their voice, the search results page not only showcases pertinent notes but also offers AI-crafted summarized answers. This innovative function harnesses AI technology to systematically arrange and amalgamate notes and experience-sharing content from the platform.

According to Xiaohongshu's official statement, the function is still in the internal testing stage and might be rolled out to a wider user group in the future. Insiders close to the platform assert that this represents Xiaohongshu's fresh foray into the Q&A realm. Xiaohongshu has a competitive edge in cultivating users' inclination towards 'proactive search' and 'practicality'. The introduction of this new function is geared towards reinforcing its pivotal position in users' decision-making processes. This function zeroes in on generating responses based on genuine notes within the platform, setting it apart from AI search products that predominantly depend on general information or publicly accessible data online.

On January 29, the pilot-scale verification platform at the Beijing Innovation Center for Humanoid Robots was officially inaugurated. During the event, the 1000th customer-customized prototype humanoid robot was unveiled as it rolled off the production line right there on the premises. Spanning an impressive 9,700 square meters, this platform boasts an annual production capacity of 5,000 units. Moreover, it offers comprehensive one-stop services, encompassing six core functionalities tailored to meet diverse industry needs.

Fengwu Technology and Zhongqing Robotics have jointly declared the successful attainment of a strategic partnership, culminating in the formation of a joint venture company. Following the establishment of the joint venture, Fengwu Technology will seamlessly integrate its advanced algorithms and 'Property Brain' functionalities into Zhongqing Robotics' framework. This integration will leverage in - management scenario data to enhance the robot's decision - making prowess and response times, thereby establishing a 'human + robot' collaborative operational paradigm. Looking ahead, both entities intend to expand the application of this paradigm into sectors such as home services and elderly care, constructing a comprehensive, intelligent service system. Their ambition is to introduce this system not only to domestic markets but also to international markets, including North America and the Middle East.

According to Business Today, the open-source personal AI assistant Clawdbot has finally been named OpenClaw. Previously, the project underwent two renamings due to trademark issues. The new name not only retains the lobster element but also emphasizes its open-source nature.

Sources close to the situation reveal that SpaceX is exploring the possibility of merging with Tesla, an idea championed by certain investors. Simultaneously, SpaceX is also contemplating an alternative strategy: merging with artificial intelligence firm xAI ahead of its initial public offering (IPO). A merger involving xAI could draw significant attention from infrastructure funds and sovereign investors in the Middle East, potentially necessitating substantial financing arrangements. Nonetheless, no definitive decisions have been reached, and the specifics could evolve, with the possibility that each company may continue to operate independently. As of the time of reporting, neither Musk nor representatives from the involved companies have provided comments in response to inquiries.

On January 30th, news emerged that Xiaohongshu is currently conducting internal tests for its "Voice Q&A" feature. For a select group of users, the familiar hot list entry located beneath the search box within the app has been replaced by a voice-activated "Q&A" button. Once users pose their questions through voice input, the search results page not only presents relevant notes but also showcases summary answers crafted by AI. It's important to note that this feature is still in the midst of internal testing, with the potential to reach a broader user base in the future. This strategic move is designed to delve into novel Q&A scenarios that seamlessly blend "real-life community experiences with AI-driven insights," ultimately empowering users to swiftly access the information they seek.