The final week of September 2025 turned out to be one of the most eventful weeks in the technology sector, with artificial intelligence once again dominating global discussions. Investments worth hundreds of billions of dollars were announced, governments deepened their partnerships with leading firms, and new breakthroughs were introduced that could shape industries from healthcare to entertainment. At the centre of these events were issues of infrastructure, sovereignty, ethics, and innovation. From energy-hungry data centres to lightweight scientific models designed for ordinary devices, the past week showcased the scale and diversity of the AI drive across nations and companies.
The week made clear that artificial intelligence is no longer confined to research labs or niche applications. It is now a matter of national strategy, industrial planning, and social debate. The announcements highlighted how private companies and governments are racing to secure hardware, talent, and influence in order to gain an edge. Concerns about energy use, privacy, and economic balance are growing, but so too is the expectation that AI could reshape healthcare, security, business, and daily life.
1. NVIDIA Commits Up to $100 Billion to OpenAI Partnership
NVIDIA’s partnership with OpenAI set the tone early in the week. On September 22, NVIDIA confirmed that it would commit up to 100 billion dollars in a long-term investment in OpenAI. This deal is one of the largest financial commitments ever seen in artificial intelligence. It includes not only the supply of advanced data centre chips but also the joint deployment of at least 10 gigawatts of compute infrastructure starting in 2026. The partnership is designed to address OpenAI’s growing training needs for next-generation models such as GPT-5, at a time when global projections suggest that AI demand could triple overall data centre capacity by the end of this decade. Jensen Huang, NVIDIA’s chief executive, described the relationship as a union of hardware and software that could unlock new levels of scaling.
We’re proud to announce a landmark partnership with @OpenAI to build new gigascale AI factories using millions of NVIDIA GPUs. ?
This partnership will supply 10 gigawatts of GPUs to fuel @OpenAI‘s data center growth. pic.twitter.com/CYEB2PdfWY
— NVIDIA (@nvidia) September 22, 2025
The implications of this agreement go well beyond corporate growth. By tying OpenAI’s model development more tightly to NVIDIA’s hardware ecosystem, the deal effectively creates a closed loop of innovation, profit, and dependency. Proponents argue that this is vital for maintaining U.S. leadership at a time when China is accelerating its own AI strategy. Critics warn that the concentration of power in so few hands could reduce competition and leave smaller players struggling to keep up. The move also reignited debates about energy consumption, as the compute required for future models will demand power equivalent to that used by millions of households. Still, industries from autonomous systems to drug research stand to gain from faster progress, making the deal one of the most consequential developments in years.
2. OpenAI, SoftBank, and Oracle Unveil $500 Billion Stargate Expansion
Only days later, another monumental project was revealed. On September 25, OpenAI, SoftBank, and Oracle announced a large expansion of their “Stargate” initiative. With a target budget of 500 billion dollars, this project aims to establish a 10-gigawatt network of AI data centres by 2030. The latest update confirmed five new sites across the United States, particularly in Texas and Arizona, with over 7 gigawatts of new capacity planned. These centres will host millions of GPUs and serve as the backbone for OpenAI’s enterprise tools and advanced agentic systems. SoftBank’s Masayoshi Son described the effort as the foundation of the “AI century,” while Oracle contributes its cloud expertise to ensure scale and stability.
More compute in the making.
Announcing 5 new Stargate sites with Oracle and SoftBank, putting us ahead of schedule on the 10-gigawatt commitment we announced in January. https://t.co/AWJQt02XZ4
— OpenAI (@OpenAI) September 23, 2025
This expansion comes against the backdrop of U.S.-China tensions and the growing desire for sovereign AI infrastructure. By reducing dependence on external providers, OpenAI and its partners seek to ensure that American firms remain at the centre of global AI compute. Supporters highlight the potential economic impact, including over 100,000 jobs created, while environmental campaigners have raised alarms about the carbon footprint. The project leaders have promised that at least half of the energy will come from renewable sources, though critics doubt the timeline. For developers and governments, the practical outcome is expected to be cheaper API access, stronger data security, and faster service delivery. For the broader industry, it shows how hardware ambitions are now central to AI’s direction, as firms like OpenAI move away from exclusive reliance on Microsoft Azure.
3. UAE President Meets OpenAI CEO for AI Collaboration Pact
Geopolitical collaboration also featured prominently during the week. On September 27, UAE President Sheikh Mohamed bin Zayed Al Nahyan met with OpenAI’s Sam Altman in Abu Dhabi to sign a new cooperation pact. The discussions centred on building Arabic-language models, setting ethical standards for AI governance, and expanding sovereign cloud infrastructure through G42’s network. With more than 100 billion dollars already invested in AI, the UAE has positioned itself as a global hub for the sector. Hosting OpenAI’s first Middle East data centre marks a further step in that plan. The focus on energy-efficient cooling technologies is particularly relevant given the desert climate, and it reflects the UAE’s broader attempt to blend technical progress with environmental responsibility.
For OpenAI, the arrangement offers more than just financial backing. It opens a path to reach the 400 million Arabic speakers worldwide through improved natural language processing, supporting education, healthcare, and government services. It also diversifies OpenAI’s funding and infrastructure base beyond U.S. hyperscalers, reducing reliance on a small circle of American partners. Critics caution that data governance in authoritarian contexts could create risks, but both sides have agreed to include UN-aligned safety audits in the collaboration. Strategically, the UAE is presenting itself as a neutral broker in the global technology competition, countering Chinese influence in the region and offering a model for other nations seeking to carve out a place in the AI economy.
4. Alibaba Surges AI Spending Beyond $50 Billion Amid Chip Race
Meanwhile in Asia, Alibaba made headlines on September 24 by announcing that it would raise its AI spending above 50 billion dollars. This sparked an 8 percent jump in its stock price, the highest in four years. The investment will support the Qwen3-Max model, which surpasses benchmarks in mathematics and programming and boasts a parameter count beyond one trillion. It will also finance new data centres across Asia and Europe. CEO Eddie Wu described the effort as an act of defending China’s AI sovereignty against U.S. restrictions on advanced chip exports.
The push extends to hardware as well, with Alibaba backing Huawei’s Ascend clusters and photonic processors that promise one hundred times higher efficiency. Qwen3’s two million token context window allows highly complex simulations, with direct benefits for e-commerce, logistics, and scientific applications. Alibaba’s video generation tool Tongyi Wanxiang 2.5 is also drawing attention for producing cinematic-level output that rivals OpenAI’s Sora. Domestically, these tools strengthen Taobao’s recommendation engines and are projected to raise revenues by 20 percent. Internationally, they position Alibaba as a central force in the multipolar AI world, where China seeks to challenge American dominance and widen access to advanced models through open-sourced variants.
5. Meta Launches Vibes: AI-Generated Video Feed in 40+ Countries
The entertainment industry was also shaken this week, as Meta launched its new feature called “Vibes” on September 25. Available in more than 40 countries, Vibes is a short-form AI-generated video feed integrated into Meta’s AI app and website. It allows users to create personalised clips from text prompts, which are then tailored using data and trends. Features include automatic dubbing into multiple languages, audio syncing, and direct links to e-commerce products. Early reports suggest that engagement has risen by 30 percent since the feature’s launch.
Excited to share Vibes — a new feed in the Meta AI app for short-form, AI-generated videos. pic.twitter.com/iVMa86Li7s
— Alexandr Wang (@alexandr_wang) September 25, 2025
Meta’s chief executive, Mark Zuckerberg, presented Vibes as a way to expand creative expression and offer tools to ordinary users. However, traditional creators have voiced concern that the tool could replace human labour, prompting Meta to set aside one billion dollars for a creator support fund. On the technical side, the system runs on Llama 3.1 combined with advanced diffusion models, and it uses edge AI to reduce server loads by 40 percent. Watermarking has been introduced to meet new EU rules and to reduce the risk of misuse through deepfakes. For advertisers, Vibes promises improved returns through precise targeting, while for users in emerging markets, it allows easier access to personalised content in non-English languages.
6. Apple Releases SimpleFold: Efficient Protein Folding AI
Closing the week, Apple announced a major scientific achievement. On September 27, the company released SimpleFold, a protein folding model with only 100 million parameters that can match 95 percent of the accuracy of AlphaFold2 while requiring 1,000 times less compute. What makes this achievement striking is its ability to run directly on iPhone Neural Engines. Researchers have described it as bringing advanced protein modelling to the masses. With applications in personalised medicine, vaccine development, and real-time diagnostics, SimpleFold opens doors for both professionals and smaller laboratories.
