Summary
Between September and early October 2026, a cluster of seemingly unrelated events landed in the same window: NVIDIA agreed to acquire Hugging Face, the main open-source model hub, for about $12.9 billion; AMD bought Fei-Fei Li's world-model lab World Labs for roughly $8.2 billion; Meta launched Muse, billed as the "world's first personal AI agent," and put it into its AI glasses; OpenAI packed more than 20 announcements into DevDay, headlined by agents that can take on ongoing responsibilities; and in the same week, the Asia New Vision Forum, Tech Week Singapore, and — in early October — AI Everything Abu Dhabi pulled half of Asia and the entire Gulf toward one theme.
Read together, these are not seven separate news items but different faces of one story. This article argues that what is happening in 2026 is a regrouping of AI power, wrapped from three threads: chipmakers shifting from "selling compute" to "enclosing the entire ecosystem"; giants moving the battlefield from chatbots to agents (AI assistants that autonomously execute long-running tasks); and regional players in the Gulf and Southeast Asia trying to play a more important hub role outside the US-China axis. Understanding these three threads explains where the next 12 to 24 months are headed — and what it all means for ordinary people: your work, your privacy, your consumer decisions, and your emotional life.
I. The Scene: What Happened
1. NVIDIA Buys the Open-Source Home Base: $12.9 Billion for Hugging Face
On September 3, 2026, NVIDIA announced it would acquire Hugging Face — the hub where developers download, host, and share open-source AI models — for $12.93 billion. The news was confirmed by NVIDIA's own blog and by CNBC.
The oddity is a role reversal: NVIDIA has long cast itself as open source's biggest champion, yet it is now buying the collective hall of the open-source world. More importantly, this is the first time a chip company has taken hold of the model-distribution gateway itself. Previously, downloading a model from Hugging Face and deploying it on NVIDIA GPUs were two separate steps; after the acquisition, NVIDIA holds both ends at once — where models come from and where they run — forming a complete chain from distribution to deployment.
2. AMD's Parallel Move: $8.2 Billion for Fei-Fei Li's World Labs
On September 28, AMD announced it would acquire World Labs, the AI lab founded by Fei-Fei Li, in an all-stock deal worth about $8.2 billion (roughly 55 billion RMB). Li will join AMD, reporting directly to CEO Lisa Su. Su's explanation was terse: "The deeper you understand the entire end-to-end process, the better systems you can build."
This deal stems from the same logic as NVIDIA's Hugging Face acquisition: hardware vendors extending into software and models to fill in their own gaps. NVIDIA has CUDA, a software moat built over a decade; AMD has strong hardware, but is comparatively weaker in software ecosystem and frontier research talent — so this deal rapidly strengthens both, while betting on "spatial intelligence and world models," the frontier track that most drives compute demand. World Labs focuses on "world models" (letting AI simulate, in its "mind's eye," how the physical world will change before it acts) — the core technology behind robots and autonomous driving, i.e. "physical AI" (AI with a body that acts in the real world).
3. Meta Launches Muse: Putting an Agent in Your Glasses
On September 8, Meta released Muse, officially "a secure, private personal AI agent that proactively helps with people's goals and suggests ideas." The key word is "proactively" — it no longer waits for your question but continuously learns your habits and goals and acts on your behalf. On September 23, at Connect, Meta announced it would bring Muse to its AI glasses; on September 29, it extended Muse to small businesses and added integrations including Shopify and Dropbox.
Commercially, this means Meta wants the agent to go from "an app on your phone" to "an entry point always in front of your eyes" — an assistant that sees you, hears you, and acts for you at any moment.
4. OpenAI DevDay: 20+ Announcements, Headlined by Agents
On September 29, OpenAI held its "biggest yet" DevDay, with more than 20 announcements. Its framing: "today, we introduced agents that can take on ongoing responsibilities." Two drew the most attention: an agent toolset called dots, and a new model, GPT-6.1 Sol. OpenAI also opened ChatGPT as a shared surface where humans and agents collaborate, letting developers ship native experiences directly to its 1.2 billion weekly users.
One line from OpenAI is worth remembering: it aims to "expand our commitment to an open ecosystem." In business terms: whoever owns the agent gateway owns the next generation of distribution power.
5. Three Summits in One Week: Regional Players Contesting the "Third Pole"
First, a definition. Beyond the two AI camps of the US and China, the EU, Gulf states, India, Japan and South Korea, and Singapore are all openly or semi-openly vying to lead a "third path." "Third pole" is not a rigorously defined technical term but a shorthand for this multilateral trend — actors unwilling (or unable) to pick sides between the US and China and trying to become a force of their own. It refers both to relatively independent governance and technology options and to regional hubs of capital and compute.
The three summits chosen here are representative cases of this trend in the September–October 2026 window — but they are not the whole picture. At least two others belong to the same story: India's AI Impact Summit (February 2026, the first held in the Global South) and China's World Artificial Intelligence Conference in Shanghai (July 17–20, 2026). These three are chosen because they are geographically concentrated and closely timed, best revealing the division of labor and mutual echoes within one region:
- Asia New Vision Forum (ANVF Innovation 2026): held September 29 in Singapore, co-hosted by Caixin Media and the international conference platform DLD, themed "Disruption Meets Resilience." Scholars including those from Peking University's Institute for International and Strategic Studies attended, discussing "how Asia can shape a sustainable future with innovation and resilience amid accelerating technological disruption and global realignment."
- Tech Week Singapore 2026: September 29–30 at Marina Bay Sands, with more than 28,000 attendees and 550+ speakers, focused on AI infrastructure (the data centers, chips, and networks that run AI).
- AI Everything Abu Dhabi 2026: October 6–7 at the ADNEC Centre, hosted by a UAE government department, billed as "the world's largest AI-first expo and summit," aiming to help the UAE become "the world's first AI-native society," gathering 400+ AI companies.
Three different countries, three different phrasings — but their dense clustering in the same week, in the same region, reflects a shared regional aspiration.
II. The Hidden Business Logic: Three Threads
Thread One: From "Selling Compute" to "Enclosing the Entire Ecosystem"
NVIDIA's and AMD's acquisitions tell the same story: the chip war has finished the hardware round; what is being fought now is the full-stack loop of "hardware + models + developers."
By buying Hugging Face, NVIDIA bought "the marketplace where the gold diggers gather" — it is no longer just selling shovels but becoming the mandatory route for everyone to find, dig, and refine gold. AMD's purchase of World Labs uses one giant number to instantly patch its weakness in "software + top talent" while betting on the most frontier and most compute-hungry direction.
Behind it is a stunning capital ledger: in 2026, Google, Amazon, Microsoft, and Meta together plan to spend about $725 billion in capital expenditure (capex — long-term spending on equipment, data centers, and the like), up roughly 77% from 2025's record ~$410 billion. When more than $700 billion a year is poured into compute and data centers, hardware itself is no longer scarce; what is scarce is the "gateway" and "ecosystem" that make that compute actually used, and used repeatedly. That is precisely why chip giants are shifting from selling compute to enclosing ecosystems.
Thread Two: Leading Companies Move the Battlefield to Agents
The moves by OpenAI, Meta, and others point to one shared consensus: from chatbots to agents.
- OpenAI's dots targets "agents that take on ongoing responsibilities" and opens ChatGPT so developers can reach 1.2 billion weekly users directly.
- Meta's Muse leads with "proactive" and moves into AI glasses, making the agent a "portable gateway."
- AMD bets on world models via World Labs — the prerequisite technology for agents that "work in the physical world."
Why is everyone fixated on agents? Because the two business models are entirely different. A chatbot is a pay-per-use toy — you use it once, pay once, and leave. An agent is a subscription entry point that takes over your daily workflows — it keeps doing things for you, and you keep paying for it. More critically, whoever becomes the default dispatcher of your daily tasks holds a position equivalent to a mobile operating system — and that is a genuinely durable business.
But there is one enormous obstacle on this path: trust. A 2026 survey by consultancy BCG found that only 4% of consumers would let an AI agent complete a purchase alone; a March 2026 McKinsey report likewise found that security and risk concerns are the number-one barrier to scaling agentic AI. In other words, the technology is no longer the bottleneck — whether people trust it is the bottleneck.
Thread Three: The Gulf and Southeast Asia Take Their Positions as Regional Hubs
Lay the summits above on a map and a line of division of labor emerges:
- The Gulf supplies money and compute. Saudi Arabia and the UAE have together committed more than $50 billion to GPU (graphics processing unit, the core chip for AI training) infrastructure; Saudi AI firm Humain was cleared to import 18,000 of NVIDIA's most advanced GB300 servers. What they want is "sovereign AI" (independently controlled national AI capability) — using petrodollars to buy a ticket into the AI era.
- Southeast Asia supplies scenarios and neutrality. Singapore hosting three summits in a week is itself a positioning as a "neutral hub outside the US-China rivalry" — connecting both capital and application scenarios.
- East Asia supplies market and manufacturing. Through the Caixin-hosted forum in Singapore, Chinese institutions and scholars export agenda and discourse.
Together, these three form a concentrated display of regional power beyond the two US and Chinese AI camps: the Gulf handles capital and energy, Southeast Asia handles transit and scenarios, East Asia handles manufacturing and market. Amid intensifying US-China tech competition, any force that can offer a third path to nations that "don't want to choose sides" may gain strategic space. It is worth stressing that "third pole" here is closer to an emerging trend than an already-formed camp.
III. Trends (the Next 12–24 Months)
Open source is being absorbed by capital. Hugging Face being bought by a chip giant is a landmark signal: the open-source door will not close, but the keys to distribution will be held by the giants. Over the next 12 months, expect more "open community absorbed by a giant" deals — and the accompanying debate about "whether open source is still open."
The focus shifts from "how smart the model is" to "who owns the gateway." What is contested is no longer parameter counts or benchmark scores, but whose agent first lives inside your phone, glasses, and office software as the default dispatcher of your daily tasks. This is why Meta rushed Muse into glasses and OpenAI rushed to open the ChatGPT platform — they are racing for the next "operating system" position.
Hardware becomes sexy again. AI glasses, on-device chips, and robots will replace "yet another chat box" as the focus of hardware innovation. The world models AMD bought via World Labs are exactly the foundation for robots.
Compute and power become the invisible bottleneck and investment theme. The $725 billion in capex ultimately lands on the power grid. Data-center electricity, energy supply, and grid upgrades will be the most underrated constraints and investment themes of the next two years.
"Trust" turns from a soft topic into a hard metric. When only 4% of people would let an agent spend money alone, whoever offers auditable, revocable, explainable agent-authorization mechanisms will truly open the market. This will spawn a new industry: agent identity verification and liability insurance.
Sovereign AI turns from slogan into purchase order. Gulf states are proving with real money that "AI sovereignty" can be bought. This sets an example for every middle power: build your own AI capability, or be "rented" in the next round.
IV. What It Means for Ordinary People
1. Your Way of Working Is Being Reshaped
This is not the future; it is happening within a year or two. Agents will book, compare prices, fill forms, draft, and run repetitive workflows for you — someone who knows how to use agents may be several times more productive than someone who does not.
The debate over how big the impact will be is fierce. Anthropic CEO Dario Amodei predicts AI could wipe out half of entry-level white-collar jobs and push unemployment to 20%; BCG's judgment is gentler, holding that AI "will reshape more jobs than it replaces" — many people keep their jobs but face radically new ways of working. Academia sits in between: research from Stanford and others reminds us that the real shock may not be "losing a job overnight" but "being asked to do the same job in an entirely new way."
The most practical conclusion for individuals: rather than worry about when you will be replaced, turn agents into your leverage now.
2. Your Privacy Boundary Is Blurring
Muse's selling point is "proactively helping you," and "proactive" requires long-term access to your data — your calendar, habits, conversations, preferences. Convenience and surveillance are two sides of the same coin.
The data captures this tension: 81% of US consumers worry about how companies use their data. McKinsey also ranks security risk as the top barrier to scaling agents. For an agent to act for you, it must know enough; knowing enough means entrusting your life to a system.
3. Consumption Is More "Arranged" — But Emotional Value Still Matters
Future consumption will be increasingly "arranged": prices, recommendations, and decisions handled by agents, with your power of choice quietly taken over by platforms. This is the other side of that "only 4% would let an agent order alone" figure — people want convenience yet instinctively resist handing over decision-making entirely.
But here is the easily overlooked flip side: emotion and emotional value are precisely what agents can least "take over" — and least should. AI companion apps have exploded in recent years — companion apps have reached about 220 million cumulative downloads and have pulled in more than $400 million. Yet they have always carried a sharp critique: when companionship comes from a program designed to "make you satisfied" rather than an existence with its own joys and sorrows — a self of its own — is it really companionship?
Because of this, some products marketed as "AI companions" have been pointedly called out in the Chinese internet as having "no 'self'" and "just playing a role." The critique touches a deeper question: an agent can compare prices, schedule, and execute tasks for you, but it can hardly feel for you, or make you truly feel understood. Emotional value, genuine human connection, being seen and cared about — these are scarce goods between humans, and may become more precious, not less, as AI spreads.
In other words, the better AI gets at "doing things," the more people need "real people and real emotion." This is both a risk (getting hooked on the instant gratification an algorithm serves up) and an opportunity (giving the time saved back to the people who truly matter).
V. Conclusion: What Ordinary People Should Hold Onto
Taken together, this cluster of events tells one story: AI is turning from "a technology that chats" into "a business that wants to take over your workflows and decision gateways." Chipmakers extending into ecosystems, giants racing for agents, the Gulf and Southeast Asia betting on regional hubs — competition has moved from technology itself to gateways and infrastructure.
For ordinary people, both the upside and the risk are real: the productivity tools genuinely work, and the erosion of privacy, the squeeze on jobs, and the outsourcing of your power of choice are genuinely happening. The most practical response is two principles that seem contradictory but are in fact complementary:
- Actively learn to use agents — treat them as leverage, raise your own productivity a notch first, and do not passively wait to be replaced.
- And hold one line: never let any single platform fully act on your behalf — use multiple platforms, stay clear-eyed about assistants that "proactively help," and reserve your most important decisions and deepest emotions for real people.
Technology will keep sprinting. But the power of choice is best kept in your own hands.
References
- NVIDIA. "NVIDIA to Acquire Hugging Face." NVIDIA Blog, 2026-09-03. https://blogs.nvidia.com/blog/nvidia-to-acquire-hugging-face/
- CNBC. "Nvidia agrees to buy Hugging Face for almost $13 billion." 2026-09-03. https://www.cnbc.com/2026/09/03/nvidia-agrees-to-buy-hugging-face-for-almost-13-billion-ai-expansion.html
- Securities Times (证券时报). "AMD acquires Fei-Fei Li's AI lab for ~55 billion RMB; Li to join AMD reporting to Lisa Su." 2026-09-28. https://www.stcn.com/live/video-detail/25857.html
- Meta. "Introducing Muse: The World's First Personal AI Agent." 2026-09-08. https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/
- Meta. "Everything We Announced at Meta Connect 2026." 2026-09-23. https://www.meta.com/blog/meta-connect-2026-everything-we-announced/
- TechCrunch. "Meta is expanding its AI agent Muse to small businesses." 2026-09-29. https://techcrunch.com/2026/09/29/meta-is-expanding-its-ai-agent-muse-to-small-businesses/
- OpenAI. "DevDay 2026 Recap." 2026-09-29. https://openai.com/index/devday-2026-recap/
- Caixin Global. "ANVF INNOVATION 2026." https://www.caixinglobal.com/caixin-anvf-2026/
- Caixin. "2026 Asia New Vision Forum Innovation Summit opens Sept 29." https://topics.caixin.com/2026/asia-new-vision-forum_2026/
- Tech Week Singapore. Official site, 2026-09-29~30. https://www.singaporetechnologyweek.com/
- AI Everything Abu Dhabi. Official site, 2026-10-06~07. https://www.aieverythingabudhabi.com/
- Yahoo Finance. "Google, Microsoft, Meta, and Amazon capex spending to hit $725 billion in 2026." 2026-04-30. https://finance.yahoo.com/sectors/technology/articles/google-microsoft-meta-amazon-capex-131823436.html
- Statista. "Big Tech's AI Spending to Reach $760 Billion in 2026." 2026-07-31. https://www.statista.com/chart/35046/capital-expenditure-of-meta-alphabet-amazon-and-microsoft/
- ClusterBid. "Middle East GPU: Saudi Arabia & UAE AI Buildout." https://clusterbid.com/blog/middle-east-gpu-investment-saudi-arabia-uae-ai-infrastructure
- Phenomenal World. "Nvidia in the Gulf." https://phenomenalworld.org/analysis/nvidia-in-the-gulf/
- BCG X. "Personal AI Agents Compete on Trust, Not Capability." 2026-09. https://www.bcg.com/x/the-multiplier/personal-ai-agents-compete-on-trust
- McKinsey. "State of AI trust in 2026: Shifting to the agentic era." 2026-03-25. link
- IT Home (citing Brigham Young University et al.). "AI companion apps pull in over $400M; nearly 70% of users hide use from partners." 2026-07-15. https://www.ithome.com/0/977/200.htm
- Incave. "A History of AI Companions: From ELIZA to the 2026 Boom." 2026-06-18. https://incave.io/blog/ai-companion-industry/
- Stanford SIEPR. "What is really happening to jobs? Separating AI hype from reality." 2026-07-06. link
- BCG. "AI Will Reshape More Jobs Than It Replaces." 2026-04-03. https://www.bcg.com/publications/2026/ai-will-reshape-more-jobs-than-it-replaces
- Anthropic. "Labor market impacts of AI: A new measure." 2026-03-05. https://www.anthropic.com/research/labor-market-impacts