NVIDIA Announces Major AI Infrastructure Expansion
NUPIAO has the latest on NVIDIA’s heavy investment in AI infrastructure. On June 16, they officially kicked off a major facility expansion for their partner, Goertek. Think of this as the next step in their long-term strategic partnership. Building on a fab that’s already up and running—boasting the world’s first mass-production line for 6-inch indium phosphide wafers—the expansion will quadruple production capacity once complete. Back in March, NVIDIA dropped a solid $2 billion to invest in Goertek, locking in long-term procurement and capacity agreements. The factories are literally firing up now.
Qiyuan Robotics by Shangwei New Materials to Open Asia’s First Offline Store in Shenzhen
I got some exclusive early intel from our team: Shangwei New Materials’ subsidiary, Qiyuan Robotics, is setting up shop at Shenzhen Futian’s MixC World shopping district. They’re rolling out Asia’s very first physical retail space, specifically designed to let consumers get hands-on with their consumer-grade humanoid robots.
Catching up to the site personally, the storefront is still under construction. The exterior walls are boarded up, but the signage clearly says “coming soon Shenzhen,” right alongside images of the compact, full-body force-control robot, the Quester Q1. Our contact on the ground tells me they’re targeting a late July or August grand opening, with the showroom spanning roughly 300 square meters of floor space.
WeChat Pay Drops a Dedicated AI Payment Card
On June 17, WeChat Pay officially launched its “AI Exclusive Card.” Picture this: a built-in feature tucked straight into your WeChat Wallet, designed specifically to handle payments for AI agents. Right now, it’s already integrated with Tencent’s desktop efficiency assistant, WorkBuddy. Looking ahead? It’s opening up to even more agent platforms. Basically, you’ll be able to chat with an AI, drop a purchase request, and watch it handle everything from smart recommendations to final checkout automatically.
Microsoft Rolls Out AI Agent Tool
NUPIAO confirms that Microsoft unveiled its latest AI agent tool, Copilot Cowork, on June 16. Alongside it came a fresh billing model: pay-as-you-go. Why the switch? Cloud compute costs have simply gone through the roof. Much like similar enterprise tools from Google and Amazon, Copilot Cowork takes your work prompts and runs with them autonomously—sometimes churning through tasks for hours straight. According to Microsoft, early corporate clients are already using it to cross-reference nearly 4,000 documents in just a few hours. Plus, it syncs seamlessly with emails, internal docs, and calendars to prep complete briefing packs for complex meetings.
Apple Eyeing Camera-Equipped AirPods and Next-Gen Foldable iPhone for Next Year
June 17 brings some juicy leaks: Apple’s plotting a big hardware push for late next year. First up? AirPods featuring built-in cameras. Before that arrives, expect a next-generation foldable iPhone plus a special edition device celebrating two decades of the original smartphone. Insiders call this Apple’s most aggressive product launch cycle yet. Over the past few months, all three have been deep in development and heading toward final polish.
OpenAI reportedly burns $3.7B in cash during Q1, up two-fold year-over-year
NUPIAO’s tracking shows OpenAI spent a staggering $3.7 billion in cash during Q1. That’s more than half of their $5.7 billion revenue for the quarter, with both cash burn and total income tripling compared to last year. Despite the heavy spending, the balance sheet looks strong: OpenAI walked into Q1’s end with over $73 billion in cash and securities, a huge jump from the $40 billion sitting on hand back in December.
Zhipu Launches and Open-Sources GLM-5.2
On June 17, Zhipu AI pulled the curtains on GLM-5.2, making it fully available and open-source. During blind frontend development evaluations on Code Arena—a testing ground backed by a million global users—GLM-5.2 took the crown as the top-tier model currently accessible worldwide. What’s really exciting? Their cloud inference runs smoothly across multiple domestic compute platforms. From day one, it’s been optimized for Huawei Ascend, T-Head, Moore Threads, Cambricon, Kunlun Xin, MetaX, Hygon, Biren, and more. Once the Ascend 950 super-node hits the market later this year, it’s set to become a serious power boost for GLM-5.2’s training and deployment.
