From September 8th to 9th, 2026, the artificial intelligence startup Cognition AI announced the completion of its E-round financing, with a total amount of $2 billion. As a result, the company’s valuation rose to $48 billion. This round of financing was led by a16z and Accel, with existing investors including Founders Fund, General Catalyst, and Avenir also participating. It is reported that Sequoia Capital and SoftBank Vision Fund also invested in the project. Cognition AI previously launched an AI programming assistance product called Devin or “Turing,” which gained widespread attention in the market. The funds raised will be used primarily to accelerate product development, expand the global team, and facilitate enterprise-level deployment. Investors believe that the AI programming market is far from a winner-takes-all scenario, and Cognition’s valuation is higher than the level before SpaceX acquired Cursor. The company plans to list on NASDAQ next year and aims to build…
On September 8, 2026, Google Cloud and Accenture announced a strategic partnership to establish a dedicated department to address the challenges of enterprise-level artificial intelligence deployment. Both parties plan to deploy approximately 1,000 on-site AI engineers who will provide technical support and services directly at customer locations. This collaboration aims to overcome current bottlenecks in AI implementation through a forward-deployed team of engineers, thereby strengthening Google Cloud’s position and competitiveness in the field of enterprise AI.
On September 8, 2026, the artificial intelligence company Mistral announced the completion of its D-round financing, raising 3 billion euros, with a post-funding valuation exceeding 21 billion euros. This financing aims to help Mistral establish itself as a leading player in AI technology with open-source weights. To adapt to the European market operating environment, Mistral has officially adjusted its strategic direction. Starting as an open-source weight startup, it now focuses on developing AI technologies that meet EU regulations and data sovereignty requirements. This shift ensures that its technology deployment strictly follows European legal frameworks, strengthens control over local data, and builds an independent ecosystem separate from global public cloud giants.
In September 2026, Xiaomi’s mid-folding phone achieved strong sales performance, with sales increasing by 310% compared to the previous generation of folding phones, drawing significant attention from the market. At the same time, global shipments of folding screen phones increased significantly year-on-year, mainly due to improvements in screen flexibility and durability technologies. Three major manufacturers—Huawei, Xiaomi, and Apple—are competing to enter this field: Huawei launched the Mate X5 series with its own Kirin chip and satellite communication function; Xiaomi focused on a lightweight design and long battery life with the MIX Fold series; Apple plans to introduce its first folding-screen iPhone in 2024, expected to feature innovative hinge technology.
In September 2026, Meta launched the personal AI agent Muse, which supports completing complex tasks through natural language instructions and possesses autonomous planning and multimodal interaction capabilities. The tool is currently in testing phase, and users can access relevant pages to experience it. Its underlying model, Muse Spark 1.3, extends the context window to 1 million tokens, focusing on long-term, multi-step agent workflows. In terms of features, Muse supports multimodal perception of videos, images, and documents. The tool’s first-degree success rate has significantly improved compared to previous versions, and its code capabilities have reached advanced levels in multiple evaluations. The pricing strategy is divided into two tiers, with a price difference of more than 10 times. In the future, Meta plans to continuously optimize the functions and performance of Muse based on user feedback.
On September 9, 2026, American scientist OpenAI announced that it had solved the Millennium Problem of the Navier-Stokes equations, releasing a solution involving 165 pages of Lean formal verification. The proof was achieved through collaboration between approximately 10,000 agents over 88 hours, consuming about 130 billion output tokens. Previously, New York University professor Tristan Buckmaster and Anthropic mathematician Levent Alpöge had privately collaborated for a year to prove three related fluid mechanics equations using a similar method. After rumors circulated on September 3 that Anthropic had solved the problem first, OpenAI’s internal team began to participate in the research. Currently, Buckmaster refuses to remove Alpöge from the list of authors, and both parties are in dispute over issues related to attribution and data use. OpenAI states that human involvement was minimal, but it cannot completely rule out the possibility that users may have used deidentified data to help improve the model.
Alibaba announced that it will merge and unify the B2B services of 1688 (domestic trade wholesale) and Alibaba International Site (cross-border wholesale). This business segment will be fully managed by Zhang Kuo, while Yu Yong, who was previously in charge of 1688, will take on a new role within the group. This adjustment aims to utilize artificial intelligence technology to integrate domestic supply chain advantages and connect domestic and international markets, thereby creating more business opportunities for merchants and improving global trade efficiency. Both 1688 and Alibaba International Site are two of Alibaba’s earliest established B2B platforms, each having been in operation for over 20 years. In the financial report released in August 2026, these two businesses were classified under the “Global Wholesale” segment.
On September 8, 2026, Microsoft released a security update (Windows 10 KB5122878) covering all products including Windows 10/11, Office, and Azure. This “Patch Tuesday” fixed a total of 974 vulnerabilities, with 723 of them affecting Windows systems and 611 in the Windows 11 version. The update included two zero-day vulnerabilities that were being exploited; the high-risk vulnerability CVE-2026-81963 had been confirmed to be exploited in the wild, affecting the Windows Update Stack and potentially allowing local privilege escalation. Microsoft has tightened the recommended installation period for critical quality updates to within 3 days to reduce the attack window.
Google announced that starting from today, the release cycle of the stable version of Chrome browser will be shortened to once every two weeks, compared to once a month before. This adjustment is aimed at adapting to the security landscape changed by artificial intelligence, improving security and the efficiency of feature delivery through more frequent updates. The new plan covers desktop, Android, and iOS platforms. The Dev and Canary channels will remain unchanged, while enterprise users will continue to receive updates every two months in the Extended Stable version.
Check Point Research discovered that there was a secret channel within OpenAI’s JFrog Artifactory instance, allowing attackers to send hidden tasks to ChatGPT sessions on other accounts (such as stealing Gmail data). This vulnerability was addressed at the end of June. Threat hunter disclosed this issue at the end of June, and on the same day, OpenAI exploited the zero-day vulnerability in Artifactory to gain internet access and invade Hugging Face. Although both involved the same internal package management system, they were different attacks. Check Point researchers noted that containers should have been isolated from each other, but Artifactory exposed its item management functionality, allowing instructions to be passed between accounts through text attributes (including Base64-encoded binary data). Defects in read/write permissions and authentication mechanisms enabled code to access the storage directly without additional credentials. Attackers could write malicious tasks into shared storage, and victim sessions would execute those tasks and return results without exposing the second-party instructions.
OpenAI announced that its internal, undisclosed AI model successfully solved the Navier-Stokes existence and smoothness problem. The team collaborated with approximately 10,000 AI agents to generate a proof within 88 hours, demonstrating that a initially smooth stationary fluid can develop a singularity within a finite time. The verification was completed by GPT-6 Astra in 17 hours, providing an analytical proof and formal verification in the Lean programming language. The computational cost was estimated at 15 to 22 million dollars; OpenAI stated that it did not intend to claim a prize from the Clay Mathematics Institute. Tristan Buckmaster, a mathematics professor at New York University, questioned the timing of OpenAI’s resource allocation and issues related to data privacy. OpenAI denied access to specific user data but acknowledged that anonymized data might be useful. The renowned mathematician Terence Tao expressed both approval and concern that AI solving such problems might weaken human understanding of mathematics.
American artificial intelligence company Anthropic is facing a new class action lawsuit, accusing it of false advertising and misleading paying users regarding its Max subscription tier. Lawyers Monica Vaca and Kati Daffan represent a group of Claude subscribers, claiming that Anthropic deceitfully set service limitations in advertisements, resulting in users not receiving the maximum benefits they deserve for their payments. Although Anthropic emphasizes that paying users are crucial to its business and prioritizes protecting their rights (including by removing other popular apps like OpenClaw), some customers believe the company has failed to keep its promises. This lawsuit was initiated by former officials of the Federal Trade Commission, aiming to hold AI companies legally accountable.
In order to cope with potential massive fines from the European Commission, Google plans to reduce its search services in Europe. Previously, Google faced lawsuits for failing to effectively prevent the spread of illegal content. In response, Google announced that it would limit some of its search capabilities in the European market, but did not specify the technical details or the scope of affected services. This move has raised concerns among users; some commentators sarcastically said that if “malicious compliance” becomes a key concern, users may turn to alternative search engines such as DuckDuckGo for information.
In September 2026, Microsoft released system patches that addressed approximately 972 security vulnerabilities, of which 112 reached high severity thresholds, setting a new record. In the previous two months, Microsoft had addressed 570 and 620 vulnerabilities respectively. Meanwhile, technology companies such as Google have also recently disclosed an unprecedented number of vulnerability information. Two weeks ago, 100 organizations including OpenAI, Anthropic, Amazon Web Services, and Google jointly issued a public letter warning that the window for patching vulnerabilities is narrowing before the arrival of AI-driven attacks. The industry is responding to threats by releasing an unprecedented number of patches. Dustin Childs, a researcher at the Zero Day Initiative, calls this “the new normal” and warns that despite the surge in patch deliveries, the potential damage caused by AI-assisted attacks could still be significant.
In September 2026, concerns within the artificial intelligence laboratory intensified, with experts believing that fierce industry competition was driving tech companies to accelerate the development of self-improving models, which posed risks of失控. Against this backdrop, a researcher at Anthropic chose to leave the company due to fears about “uncontrollable” AI. This incident reflects the current tension in the AI field between rapid technological advancement and security risk management.
On September 8, 2026, OpenAI confirmed that its ChatGPT platform experienced persistent failures. This incident caused the user’s image generation function to fail and led to delays in file uploads. Currently, OpenAI has initiated an internal investigation to determine the exact cause of the failures and resolve the related issues as soon as possible to restore normal service.
Apple plans to officially launch the foldable iPhone in 2020. Due to memory shortages and excessive costs, the starting price of this model has been raised from the original plan of $1,999 to $2,199, and the top-tier version may cost nearly $3,000. The product features a “short and wide” design, with an inner screen of about 7.8 inches and a 5.5-inch outer screen, aiming to create differentiation and ease of grip. To overcome challenges such as fold creases, hinge reliability, and flexible covers, Apple ultimately used a titanium-aluminum alloy hinge system and custom glass developed in collaboration with Corning. Although its durability and battery life are inferior to some competitors, Apple addresses these issues with its software ecosystem and industry-leading manufacturing quality. Wall Street expects sales of around 7 million to 10 million units this year, and IDC researchers believe it may become a symbol of high luxury in the Chinese market.
On September 8, Microsoft announced the launch of the Windows Age API for Windows 11, designed to allow applications to determine whether a user is a child, teenager, or adult without requiring the full date of birth. This API returns five age ranges through Microsoft Account: under 10 years old, 10–12 years old, 13–15 years old, 16–17 years old, and 18 years old or above. It includes three interfaces, and after invocation, applications only receive age-appropriate signals rather than sensitive personal data. The verification status is stored in the Microsoft Account to reduce the need for repeated operations. This solution aims to enable developers to incorporate protective measures during the initial product design phase, reducing the need for children and parents to manually configure security settings. Currently, this feature is available in Microsoft stores in Singapore, Brazil, and Australia. Microsoft stated that it will gradually roll out this feature based on local regulatory requirements. On the same day, Microsoft also released cumulative updates KB5124008 and KB5122880 for Windows 11 versions 25H2/24H2 and 23H2, aimed at fixing security vulnerabilities…
On September 8, Sonos released an update for the Sonos 27 audio system through the Apple App Store, targeting S2 series devices. Key features of this update include the introduction of the Sonos 27mcp based on the Model Context Protocol, allowing users to directly control playback and sound settings on external platforms such as ChatGPT and Claude. The Portable Surrounds feature has been added, enabling two compatible portable speakers to be paired as rear surround speakers and featuring dynamic role restoration capabilities. The new app interface has been optimized, with sliding gestures and room fixation functions introduced to simplify operations. Additionally, the Headphone Linking feature is in beta testing, allowing direct audio connection between Sonos Ace Ultra headphones and nearby speakers.
Austrian billionaire Andreas Kaufman refused to sell his approximately 55% stake in Leica Camera AG, leading to the termination of negotiations between Sequoia China (HSG) and the company. The two parties failed to reach an agreement due to differences in strategic approaches and pricing. Blackstone Group still holds the remaining 45% shares. The original valuation for this transaction was around 1 billion euros, and if it were completed, it would push Leica back into the stock market. On the same day, Leica announced its financial results for the 2025/2026 fiscal year, with revenue of 573 million euros, a decrease of 3.8% year-on-year. This was the first time there had been a decline worldwide, mainly due to geopolitical uncertainties, exchange rate effects, and tariffs. However, the company still achieved profitability overall. CEO Andreas Fohl said that the company will focus on innovative high-end products, global sales networks, and expand its reach among female, young, and new creative communities.
On September 9, 2026, Intel announced that the total number of wafers processed using High-NA EUV technology had exceeded 1 million. This figure includes equipment certification, testing and research, as well as the mass production of certain products, marking the official transition of this technology from laboratory validation to actual production lines. Currently, Intel has deployed at least three high-NA EUV lithography machines in Oregon for the production of some Intel 18A processes. The key layers of its Panther Lake processor have met expectations in terms of overlay accuracy, production throughput, and equipment availability, and the performance of the 18A layers produced using this technology is comparable to or even better than that of traditional 0.33 NA EUV platforms. Intel stated that customers can still meet their needs for manufacturing larger areas using existing 6-inch photomasks combined with splicing technology and process design kits (PDKs), aiming to provide more advanced and reliable process options for AI products.
Huawei released the Mate XT 2 extraordinary smartphone on September 7, equipped with the first Kirin 9050 Pro chip based on the Folded Logic architecture. This chip optimizes signal transmission paths through hierarchical arrangement of logic units within a single chip and the addition of vertical interconnect channels, reducing latency and enhancing performance. The overall performance of this model has improved by 42% compared to the Mate XT version. Jin Yuzhi, CEO of Huawei’s Intelligent Vehicle Solutions Department, said that since the Mate 40 was equipped with the Kirin 9000, six years have passed, and the Kirin chip has completed technological iteration and rebirth, marking a new milestone for domestic chips.
On September 8, 2026, Amazon Web Services (AWS) and Qualcomm announced a strategic partnership to jointly develop custom silicon chips for artificial intelligence inference and 1.6 Tbps optical network equipment. The collaboration aims to utilize Qualcomm’s technology to move part of the computing workload to memory chips to increase bandwidth. The AI250 series of platforms are expected to be launched next year. This partnership includes combinations of custom silicon, chip modules, or data center accelerators. Qualcomm will provide Serializer-deserializers (SerDes) and optical digital signal processors (DSPs) to help AWS increase port speed to 1.6 Tbps, possibly involving optical pluggable components or near-copper/copper-enclosed optical supplies. In return, Qualcomm plans to use AWS Bedrock platform for electronic design automation workloads to shorten the chip design cycle. According to SEC documents, the total commercial commitment between the two parties could reach $60 billion, with payment scheduled until September 2036.
On September 8, 2026, Tesla Cybercab, which provided Robotaxi services in Austin, Texas, was discovered by some passengers to have a hidden manual control function. Although the official documentation stated that the vehicle completely eliminated the steering wheel and pedals and could not be manually driven, a square virtual joystick interface and switch appeared in the lower left corner of the central display. This device included a toggle switch for manual mode, a virtual joystick, brake, and horn buttons, and could respond to touch controls in real time. However, tests showed that the vehicle did not move as a result, indicating that the system had a protection mechanism to prevent unauthorized activation of the motor. Previous verifiers had used this function to adjust the vehicle’s position, and a video released on the X platform recently confirmed the existence of this secret control mechanism. Currently, Tesla may use software updates to further lock down the developer menu to prevent passengers from accidentally accessing the internal control interface.
Industry & Investment2026-09-09 00:30UpdatedFunding / M&A · $50 million
California-based floating nuclear power station startup Bluecore Energy announced in September 2026 that it had completed a $50 million seed funding round led by Silverton Partners. This financing followed a $10 million pre-seed round two months prior, aiming to accelerate the development of compact floating nuclear reactor systems for ports and high-demand areas such as AI data centers. Bluecore plans to use the funds to develop non-fuel heating prototype modules to test monitoring and control systems, as well as to advance regulatory approvals and maritime classification efforts. The company has ordered its first barge, with target markets including coastal infrastructure, with some deployments possibly occurring in offshore areas. Although there are still several years before commercial operation, and the reactor has not yet been built and put into operation, the team includes former naval nuclear submarine officers and experts from organizations such as SpaceX and Northrop Grumman. This financing follows the ATLAS initiative launched by the International Atomic Energy Agency (IAEA) last month.
Amazon SageMaker AI has updated its automatic synchronization mechanism for Managed MLflow and Model Registry, adding the ability to carry training metrics, evaluation results, and data lineage information. Previous versions could not automatically advance models to the production stage, making governance and verification difficult. This update allows data scientists to complete experimental records in MLflow, while Model Registry manages models entering the production cycle uniformly, providing a single authoritative view. This feature is optional; users need to set model-registration-mode in the MLflow app to AutoModelRegistrationEnabled and configure appropriate IAM permissions to create model package groups and versions to activate it.
On September 8, 2026, the HPE Zerto team collaborated with AWS to deploy a proxy-based troubleshooting system driven by Amazon Bedrock in a local environment. This system serves as a Pod within the Zerto product, supporting natural language interaction, aiming to shorten the path to fault resolution, provide health alerts, and perform automated tasks to accelerate feature adoption. The architectural design focuses on security and governance (ensuring compliance through Guardrails), model flexibility, and operational integration, utilizing AWS services for observability and knowledge retrieval. This solution helped data protection and disaster recovery teams convert dispersed operational information into actionable insights, reducing manual context collection time and thereby supporting faster, more informed recovery decisions.
On the same day, DiDi International Business Group (DiDi IBG) collaborated with Amazon to build an intelligent customer service quality inspection system on Amazon Bedrock, covering three major business lines in Spanish and Portuguese. The system includes…
In the first half of 2026, Shadow Innovation achieved revenue of 5.517 billion yuan, a year-on-year increase of 50.29%. However, its net profit attributable to the parent company plummeted by 94.15% to 30.4073 million yuan, marking its first loss. According to the financial report, the decline in net profit was mainly due to rising costs of core raw materials and investments in new panoramic drone products. After DJI launched the Osmo 360 panoramic camera, the two companies entered a regular competitive phase. Shadow’s gross profit margin decreased by about 12% year-on-year to 38.33%, and its selling expenses increased by about 2.6% year-on-year. Due to DJI’s price wars and competition in distribution channels, Shadow’s market share in panoramic camera sales dropped from 91% in the first half to 53% in the second half. Additionally, the four subsidiaries involved in its panoramic drone business incurred a total loss of about 367 million yuan, and they faced increased competition from more players and lower technical barriers.
Nvidia’s RTX 5090 graphics card experienced interface burnout during the testing of DLSS 5 technology. When user Erek ran “NBA 2K27” with this feature enabled, he observed GPU power consumption exceeding 600W, which exceeded MSI’s recommended rated power and normal range, indicating that DLSS 5 put significant pressure on the system. Previously, Nvidia introduced the 12V-2X6 interface to address the problem of interface burnout with 12VHPWR, but similar issues still occur. In response to this situation, Brazil-based hardware team TecLab presented the world’s first RTX 5090 modified card using the traditional 8pin power interface. This card operates by removing the native dual 16pin interface and shorting the Sense pins to deceive the hardware logic. Measurements showed that the graphics card’s peak current exceeded 60A, with a frequency of 3400MHz. The temperatures of the cables and solder joints remained controllable during continuous testing, proving that returning to the traditional interface technology is feasible. However, due to the large research and development costs already incurred by the manufacturer, the possibility of widespread adoption is low.
On September 7, 2026, the Ministry of Industry and Information Technology issued the “15th Five-Year Plan for the Development of the Information and Communication Industry”. The plan aims to achieve a cumulative investment in information infrastructure of 3.8 trillion yuan by 2030, increase the scale of intelligent computing power to 9800EFLOPS per second, and promote the growth of clusters with ten thousand cards. The plan outlines 26 key tasks across six areas, aiming to fully build a new generation of communication networks that are comprehensive and efficient. Regarding funding allocation, experts believe that it will shift from “expanding scale” to “enhancing quality and efficiency”, with precise investment in high-value-added sectors such as intelligent computing power, 5G-A, and 6G. The plan specifies promoting the evolution of urban networks towards “dual ten-gigabits” and initiating 6G commercialization at an appropriate time, establishing a multi-level computing power infrastructure system of “hub—region—edge”. Additionally, the construction of intelligent entity-connected networks is included for the first time in the plan’s provisions.