Quick Hits — September 3, 2026
Five things worth knowing this midday: Microsoft finally puts a number on Azure, Chinese model subscriptions land on a shopping app, and a phone that calls itself the first real AI agent handset clears its last regulatory hurdle.
Microsoft will finally tell investors how much money Azure makes. The company reported $29.4 billion in Azure sales for its most recent quarter and $101.9 billion for the fiscal year ended June 30 — the first time it has disclosed actual revenue rather than just a growth rate. The disclosure lands alongside a restructuring of Microsoft's reporting from three segments into two: "Agents and Infra," which folds cloud, AI software, and the traditional business-software lines together, and "Devices and Consumer," which picks up Windows, Xbox, and advertising across Bing and LinkedIn. Microsoft also nudged its Azure revenue forecast slightly lower, attributing the change to shifting some GitHub sales into the M365 Cloud bucket rather than to any demand signal. The take: the number puts Azure clearly behind Amazon's $42.2 billion cloud quarter but ahead of Google's $24.8 billion — and more usefully, it ends two decades of investors guessing at Microsoft's most important business from a percentage.
Alibaba's Tmall is now selling AI model subscriptions next to the phone chargers. The platform launched a "token top-up center" and an accompanying "AI space station" storefront on September 3, stocking subscription and pay-as-you-go token plans from Alibaba Cloud, Zhipu, Kimi, and MiniMax, delivered either as a redemption code or as a direct account top-up. Zhipu opened its own official Tmall flagship store the day before; reports cite a 40x jump in searches for the brand on its first day. Users reach the storefront by searching "token" in the Taobao app. The take: moving model access onto a mass-market shopping app turns tokens into something consumers compare, coupon, and impulse-buy like a data plan — a distribution channel Chinese labs have never had, and one that quietly pushes their buyer from developer to everybody.
A phone billing itself as the world's first AI agent handset cleared its final regulatory hurdle. Nubia's NaviX Ultra received its network access license on September 1, according to ZTE terminal chief Ni Fei, who said the device has completed both the model-filing and device-certification processes and will go on sale this month. The phone is a joint project between ZTE and ByteDance's Doubao team, and follows the December 2025 Nubia M153 — a 3,499-yuan engineering preview that sold out and doubled in price on the secondary market before some mainstream apps blocked its screen-reading automation over authorization and privacy concerns. Ni Fei says the new model runs a "native agent" approach, executing whole tasks rather than bolting AI features onto existing apps. The take: the interesting part isn't the hardware, it's that the compliance path now exists — the last attempt died on permissions, not capability.
India's top science advisor says AI rules should be "nimble, not over-constraining." Principal Scientific Adviser Ajay Sood told ET that regulation needs a cautious approach without rigid laws that would stall innovation, pointing to February 2026 amendments to the IT Act's intermediary rules — the country's only AI-specific norms to date, and ones that apply to platforms rather than model developers. India still has no standalone AI legislation; MeitY acts as the coordinating ministry and issues advisories as needed. Sood framed the approach as continuously evolving and risk-based. The take: it's the same "flexible, wait-and-see" line most governments have settled on, but from India it carries weight — the country is large enough that its choice of intermediary rules over model regulation will shape what global products ship there.
Shanghai's Bund Summit draws 50,000 registrants around an "AI new economy" theme. The 2026 Inclusion·Bund Summit runs September 9–12 at the Huangpu Expo park with one main forum, more than 40 breakout sessions, and over 600 speakers, including Nobel economics laureates Thomas Sargent and Philippe Aghion alongside executives from Alibaba, Stripe, and SenseTime. Organizers say attendees are coming from more than 50 countries and regions, with over 300 companies exhibiting across 15,000 square meters — including more than 40 embodied-AI firms showing robots doing actual work in pharmacies, factories, and inspection rounds rather than demoing backflips. The take: the programming shift from "what AI can do" to "who captures the growth" is the tell — the conference circuit has stopped selling capability and started selling economics.
Sources: Reuters · The Verge · Lianhe Zaobao · East Money · DIGITIMES · Beijing Business Today (via Sina) · ETTelecom · Economic Times · AI Era (Xin Zhi Yuan) · Phoenix Net
