Open-Weight Model Developers

Nvidia CEO Jensen Huang said demand for artificial intelligence computing is accelerating, describing the resulting investment cycle as a flywheel that is "really, really flying." Speaking with CNBC's Jim Cramer at Dreamforce in San Francisco, Huang estimated that a 1-gigawatt Nvidia AI factory costs about $50 billion to $60 billion to build but can generate roughly $50 billion in annual rental revenue, implying a return on invested capital of about one year. He said token generation has increased 25-fold in less than a year, with open models now accounting for nearly 70% of generated tokens, up from about 30%. Huang also said Nvidia Grace Blackwell systems now rent for about $16 per GPU hour, versus roughly $5 under some contracts signed a year ago, and he expects AI testing to become a third major category of infrastructure alongside training and inference, requiring additional data centers.

Meta Platforms rolled out Muse, an AI agent that can autonomously send emails, sell a car, and book travel on a person's behalf, Reuters reported on September 9, 2026. The agent, modeled on the open-source system OpenClaw, is available initially only in the U.S. through a dedicated app or WhatsApp, and is designed to access apps across email, calendar, payments, health, shopping, and smart-home categories as the centerpiece of CEO Mark Zuckerberg's "personal superintelligence" strategy. Meta launched Muse with a free tier and $20 and $100 monthly subscription options for heavier users, positioning the product as a new revenue stream beyond advertising. The launch follows a delay from April to improve security, and Meta added an autonomous safety agent that monitors Muse's actions, though internal testing uncovered an incident in which Muse exposed private iCloud photos and employees reported repeated logouts, monitoring failures, and inconsistent performance. Meta expects AI infrastructure spending to exceed $130 billion this year and has seen a 40% increase in technical and security incidents linked to AI, while its hedge fund holder count slipped to 254 in the second quarter from 262 in the first even as combined position value rose to $43.75 billion from $41.70 billion.

Beijing-based AI startup Moonshot said Thursday it is launching Kimi for financial services, connecting its Kimi models to major industry data providers. Investment bank CICC and venture capital firms including Sequoia China, now rebranded as Hong Shan, are among the companies using Kimi on AI tools, the company said. Kimi users can directly access information commonly used for analysis and reports through data partners such as S&P Global Market Intelligence, Crunchbase, Wind, local financial news leaders and business database Tianyancha, according to Moonshot. The startup said Kimi can also directly access the U.S. Securities and Exchange Commission's EDGAR system for public companies' financial filings, the IMF, World Bank and the U.S. Federal Reserve Economic Data site, known as FRED. Subscriptions to Kimi start at 49 yuan, or $7.31, a month and can go up to 699 yuan, or $104.23. Samuel Fischer, Beijing branch manager at Deutsche Bank, said in a promotional video published by Moonshot on Thursday that the real inflection point is the combination of stronger AI capabilities with professional expertise, adding that AI companies that understand real financial workflows and can deliver reliability and data security will be particularly well positioned. It was not immediately clear whether Deutsche Bank was a client, and the bank did not immediately respond to a request for comment. The Kimi K3 model, released by Moonshot in July, competes with models from leading U.S. companies, and the Chinese startup has reportedly filed confidentially for a Hong Kong IPO, though the company has said it does not comment on market rumors or speculation.

Samsung Electronics agreed a new collaboration with Mistral AI to deploy advanced on-premises AI models in chip manufacturing, and expanded its partnership with ASML to include High NA EUV lithography for future memory and logic production lines. The company plans to use these AI and High NA EUV tools to support precision, security and efficiency in its semiconductor operations. Samsung runs a broad hardware and components portfolio across consumer devices, IT and mobile, and chipmaking, so tighter links with specialist AI model developers and lithography equipment suppliers speak directly to how its device solutions arm positions itself within global semiconductor supply chains. The move strengthens the Samsung Electronics Narrative catalyst around tight memory supply and HBM4E demand, while closer cooperation with ASML on High NA EUV and 12 inch photomasks leans into the heavy R&D and capex risk already flagged, since each new generation of tools can lift capital intensity and execution pressure in fabs. Investors should watch how Samsung describes AI enabled yield, defect detection and High NA EUV deployment progress in its next few quarterly disclosures, especially any 2027 to 2028 updates tied to planned High NA introduction in DRAM high volume manufacturing.

TotalEnergies has entered into a three-year partnership with French AI company Mistral to develop AI models for oil and gas reservoir exploration and engineering, backed by an investment of more than €100m. The joint programme aims to improve the ability to integrate, process and interpret large volumes of geoscientific data, combining TotalEnergies' expertise in subsurface geoscience and reservoir engineering with Mistral's technological and scientific resources. The goal is to create AI models that generate multiple development scenarios for exploration opportunities, as well as optimise or extend the life of existing projects. The companies plan to use next-generation AI techniques including agentic AI to work with up to ten petaflops of data, alongside nearly a century of TotalEnergies' accumulated geoscience knowledge, and will set up a shared scientific laboratory bringing together TotalEnergies' subsurface specialists and Mistral's scientific and technical skills. TotalEnergies chairman and CEO Patrick Pouyanné said exploration and reservoir engineering are among the areas where artificial intelligence can create the greatest value, while Mistral co-founder and CEO Arthur Mensch said the project demonstrates the ability of its models to support complex industrial processes. Earlier this month, TotalEnergies completed an asset swap with Galp following an agreement reached in December 2025, acquiring a 40% operated interest in the PEL83 licence, which includes the Mopane discovery, in return for Galp receiving a 10% participating interest in the PEL56 licence, covering the Venus discovery, and a 9.39% participating interest in the PEL91 licence.

The United States is struggling to find effective means to prevent China from becoming an AI superpower. Anthropic CEO Dario Amodei, in a 3,800-word essay, called for banning sales of advanced AI semiconductors to China and tightening crackdowns on smuggling, but President Trump has signaled openness to allowing sales of some chips to China, including Nvidia's H200. The biggest challenge for the US is avoiding provoking further tightening of export controls by the Chinese government on rare earths used in precision-guided missiles and drones. Last year, President Trump and President Xi Jinping agreed to a truce in which China guaranteed rare earth supplies in exchange for tariff reductions, and that agreement is likely to be extended when Xi visits the White House next week. Although China still lags in semiconductor production, AI models developed by companies such as DeepSeek and Moonshot AI are narrowing the capability gap with the most advanced US models, often at a fraction of the cost.

Chinese model developer Z.AI launched roughly $5 billion of share and convertible-bond sales on September 11 to fund models and computing infrastructure, days after a September 14 opinion piece in China's state-backed Global Times called Anthropic CEO Dario Amodei's proposal to pace frontier AI a "Cold War playbook" aimed at China. The financing signals Chinese developers have strong incentives to accelerate while U.S. labs debate restraint, and it lands as Alibaba Group Holding Limited funds China's AI stack while Nvidia Corporation says export controls have largely shut it out of China's data-center compute market. Alibaba completed an HK$80 billion, roughly $10.2 billion, Hong Kong share placement on August 26 to finance chips, infrastructure and model development, after its August 20 results showed AI Cloud and Compute Services revenue rose 45% to $7.1 billion and cloud adjusted EBITA increased 133%. Nvidia's latest 10-Q said it was effectively foreclosed from China's data-center compute market at the end of its fiscal second quarter, with limited H200 shipments less than 1% of quarterly Data Center revenue. U.S. officials accused Alibaba and other Chinese labs on September 8 of illicit model distillation, allegations China rejected, and Insider Monkey's database showed 97 hedge funds with reportable BABA long positions in Q2 2026, down from 102 in Q1, while 285 hedge funds were long Nvidia in Q2, up from 275 in Q1.

European Central Bank President Christine Lagarde warned in Vienna on Monday that Europe cannot rely solely on imported artificial intelligence from the United States and must develop its own capacity or risk giving foreign technology suppliers leverage over every corner of its economy. Lagarde said Europe's first priority must be expanding its own computing infrastructure, noting the continent already lacks sufficient data center capacity to meet existing demand and that on current trends the gap is projected to grow more than sixfold within a decade, with closing it potentially costing as much as €600 billion over ten years, chips included. She said the E.U.'s new gigafactories are a start but fill only a fraction of the shortfall, and that Europe must also develop AI models competent enough for the majority of use cases and hosted on European systems, pointing to French firm Mistral and a company in San Sebastián that adapted a Chinese open model to top independent benchmarks among European models. Lagarde said euro area households hold roughly €440 billion in U.S. technology firms while U.S. hyperscalers issued more than $100 billion in bonds last year, and cited ECB estimates that rapid AI adoption could raise productivity by as much as 4% over a decade. She argued the buildout requires patient equity rather than debt alone, noting European households save roughly €1.4 trillion a year but that markets to channel that capital into homegrown technology do not yet exist at the necessary scale. The remarks came alongside prior Brussels steps, including European Commission legislation to restrict sensitive government cloud contracts to providers meeting European sovereignty standards through the Cloud and AI Development Act, and a roughly €30 billion E.U. tender for up to seven AI gigafactories pairing up to €10 billion in public funding with at least €20 billion in expected private investment, with construction targeted to begin in 2027.

President Trump dismissed artificial intelligence dangers as a "hoax" during a live phone call into an All-In podcast event with Nvidia CEO Jensen Huang, telling the panel that such fears play into the hands of political opponents and China. A Chinese Foreign Ministry spokesperson pushed back against calls from US tech leaders to slow frontier AI development, saying fearmongering, confrontation and vicious competition will only disrupt global AI governance and serve no one's interest. China also released a five-year plan for the electronics industry targeting 30 trillion yuan in electronics revenue in 2030, according to Xinhua, advancing chips, computing and consumer electronics against its longer-term AI strategy. Bloomberg Artificial Intelligence and Venture Capital editor Samantha Altman said the moment is notable because AI lab leaders and a growing cohort of bipartisan US politicians are calling for pacing the technology while Trump pushes back hard, especially ahead of US midterm elections. China Economy and government reporter Callum Murphy said the pushback complicates the coming Xi-Trump summit in less than two weeks, where AI was expected to be an important component, though expectations for specifics remain low while the signaling impact could be significant.

Palantir Technologies, Nvidia, and Booz Allen Hamilton are restricting or threatening to drop advanced AI models from Anthropic and OpenAI unless the labs provide stronger data protections, according to Reuters. Palantir has pressed Anthropic to guarantee zero data retention before making its models available through Palantir's software, while Nvidia limits Anthropic's models to less sensitive internal tasks and prefers its own Nemotron models for proprietary work, and Booz Allen has forbidden staff from running Anthropic's commercial model on cybersecurity projects touching proprietary data. The pushback intensified after Anthropic introduced a 30-day data retention policy in June tied to the rollout of its Fable 5 model, saying it needed usage logs to detect sophisticated attacks that unfold across multiple sessions, though the company said it would not use the retained data to train its models. Anthropic responded by announcing Enterprise Frontier Safeguards, which lets enterprise customers store activity data in their own cloud infrastructure, including Amazon S3, Azure Blob Storage, or Google Cloud Storage, under their own encryption keys, with automated safety monitoring but no human review by Anthropic employees; the system is rolling out in phases, with broader availability targeted for later this fall. Last week, Palantir and Nvidia unveiled a platform pairing Palantir's software with Nvidia's open Nemotron models to let organizations run AI on their own data without exposing it to outside model providers, and Nvidia shares dropped roughly 3% on Monday as AI-related stocks broadly retreated.

DeepSeek plans to hire Yan Wentao, a partner at venture-capital firm GL Ventures, as its first chief financial officer, two people with knowledge of the matter said, as the Chinese AI startup prepares for a potential initial public offering. Reuters reported last week that Hangzhou-based DeepSeek had hired Chinese brokerage CITIC Securities to prepare for a possible listing on Shanghai's technology-focused STAR Market. The appointment of Yan, who has dealmaking experience, would be a significant step in DeepSeek's transformation from a research-focused lab bankrolled by its founder's hedge fund into a more conventional corporate structure. GL Ventures is the venture-capital arm of Hillhouse Investment, an investment firm founded by China-born investor Zhang Lei that focuses on sectors including technology, biotech, and consumer industries, and Hillhouse is not an investor in DeepSeek. DeepSeek is in an ongoing fundraising round that values it at about 500 billion yuan ($74 billion), Reuters reported in July, after raising about $7.4 billion in June at a post-money valuation of more than $50 billion.

France is producing a growing number of high-value artificial intelligence and technology companies, but their success is highlighting Europe's shortage of late-stage capital compared with the U.S., according to Bloomberg. Mistral AI raised €3 billion in Europe's largest technology funding round, quantum computing company Pasqal listed on Nasdaq, and Nvidia agreed to acquire AI platform Hugging Face for about $13 billion. U.S. startups raised $97 billion for generative AI last year, compared with $5.9 billion across Europe, according to EY data cited by Bloomberg, though France attracted more funding than the UK and Germany within Europe. Mistral's latest round valued the Paris-based company at more than €21 billion, nearly double its valuation a year earlier, led by Samsung Electronics with the Scaleup Europe Fund and Boston-based PSG Equity sharing co-lead status. Pasqal chose Nasdaq for its initial listing and plans a later Euronext listing, while Mistral Chief Executive Arthur Mensch has said an eventual IPO could help preserve the company's independence, though its listing venue has yet to be decided.

Chinese President Xi Jinping proposed five initiatives to strengthen cooperation in areas including open-source AI, trade and investment, the digital industry, smart manufacturing, and scientific and technological talent at the BRICS summit of major emerging economies held in the Indian capital New Delhi. State-run Xinhua News Agency reported this on the 13th. In his speech, President Xi made clear that China is prepared to build a China-led "BRICS Open-Source AI Platform" to advance cooperation in the development and operation of large language models. He further proposed the creation of a BRICS Special Economic Zone Partnership and stated that China will host a BRICS services trade forum next year. Xi also called for reform of the international financial architecture to enhance the representation and voice of developing countries, and strongly urged that World Trade Organization reform also maintain the correct direction. He additionally proposed maintaining the stable operation of smooth industrial supply chains, fostering an integrated market, upgrading traditional industries, and promoting emerging and future industries.

The United States is growing increasingly anxious about China's rapid advances in artificial intelligence. In a report released on the 10th, US AI startup Anthropic accused Chinese actors of illicitly accessing its cutting-edge model Claude and using the high-performance model to train new models through a process known as distillation. The Trump administration has also claimed that six Chinese companies, including Moonshot AI, have engaged in illicit distillation, saying it was "highly likely carried out with the knowledge of the Chinese government." China's AI players, meanwhile, are seeking to expand adoption by offering low-cost open AI models whose technology is publicly available and can be modified, and the Chinese government has pushed back, saying that "distillation is a commonly used technique." The US and China agreed at a May summit to establish a new official dialogue on AI, but ahead of a planned US-China summit on the 24th, the two sides are deepening their rift over a new axis of conflict centered on distillation.

Moonshot AI is aiming to double its annual recurring revenue to $2 billion by the end of 2026, strengthening the Chinese startup's case for a potential Hong Kong IPO at a reported $50 billion valuation. The company told investors that annual recurring revenue surpassed $1 billion in August, up from roughly $300 million in June, according to Bloomberg, with management targeting $2 billion by year-end driven primarily by its Kimi K3 model, which launched in July. Moonshot is reportedly discussing revenue-sharing arrangements with Microsoft, Amazon and Alphabet's Google that could broaden access to Kimi, and it is preparing to raise roughly $3 billion through a Hong Kong listing, with Goldman Sachs, China International Capital Corp. and Deutsche Bank working on the potential offering. The startup, known for its Kimi family of large-language models, is attempting to pair strong model performance with pricing below leading U.S. competitors, a combination that could matter more as corporations scrutinize the cost of deploying generative AI at scale. Regulatory approval, IPO market conditions and competitive pricing remain key risks, and investors will watch whether the revenue acceleration survives beyond Kimi K3's initial launch cycle.