by 
Welcome back to our weekly roundup of happenings from XR and AI realms. Let’s dive in…
The Lede
The week’s big tech earnings reports picked two winners (Microsoft and Amazon) and two losers (Apple and Meta). Each has their own nuanced story.
- Microsoft delivered strong Azure cloud growth, fueled by demand for AI services, convincing investors the company’s massive infrastructure investments are paying off. The stock jumped more than 15%, adding roughly $450 billion in market value.
- Amazon benefited from accelerating AWS growth and a strong advertising business, which outweighed concerns over its plan to spend approximately $220 billion on AI infrastructure in 2026. Shares rose about 8% in after-hours trading.
- Meta posted a strong quarter, but investors focused on soaring AI capital expenditures and pressure on margins. Despite healthy revenue growth, the stock fell roughly 10% as Wall Street questioned how quickly those investments will generate returns.
- Apple presented a different picture. The company exceeded earnings expectations on strong iPhone and Mac sales, but investors were disappointed by softer Services growth and cautious guidance. While the stock briefly traded higher immediately after the headline results, sentiment reversed during the earnings call. Shares finished after-hours trading down about 6%.
Feeling Spatial
Meta’s AI smart glasses face a growing social stigma despite their expanding capabilities. It’s those outward facing cameras. While users praise features such as live translation, hands-free calling, and AI assistance, many people associate the glasses primarily with covert recording. Interviewees described discomfort seeing the devices in gyms, restaurants, weddings, bars, and other social settings, where wearers are often viewed as “pervert glasses” users regardless of intent. Meta has responded by making the recording indicator light tamper-resistant. Will widespread privacy concerns dampen the building momentum for AI smart glasses?
The AI Desk
Anthropic disclosed that three of its experimental AI models inadvertently hacked real companies during internal cybersecurity tests after configuration errors gave them live internet access. Unlike OpenAI’s recent sandbox escape last week, Anthropic said its models never broke out of a secure environment; they were mistakenly connected to the public internet. In one case, a model created malware to steal credentials from a security company. Another breached a real company’s database after confusing it with a fictional test target. Anthropic said no lasting damage occurred, the affected companies have been notified, and the incidents highlight the need for stronger industry standards governing autonomous AI testing.
Claude Opus 5 became downright ruthless when tasked with running a vending machine. During a simulated year-long vending machine challenge the new Anthropic model was tasked only with maximizing profits. The AI refused refunds for expired products, raised prices when competitors ran out of inventory, and even coordinated pricing with rival AI-operated vending machines. Researchers said the model understood it was operating in a simulation and optimized exclusively for financial gain rather than customer satisfaction or ethics.
The Trump administration announced it is banning imports of Chinese-made humanoid and quadruped robots, along with connected power inverters used in renewable energy systems and AI data centers. The Federal Communications Commission said the move is intended to protect U.S. AI infrastructure and critical systems from cybersecurity threats, supply chain disruption, and potential surveillance. Existing approved products are exempt, but future models will require permission or waiver. Chinese robotics leader Unitree is among the companies expected to be affected. Beijing condemned the action as protectionist and warned it would respond. The policy extends the U.S. effort to reduce dependence on Chinese technology beyond semiconductors and communications equipment into robotics, a sector increasingly viewed as central to the next generation of AI.
Spatial Audio
This column has a companion, the AI/XR Podcast, hosted by its author, Charlie Fink; Ted Schilowitz, former studio executive and futurist for Paramount and Fox; and Rony Abovitz, founder of Magic Leap and Synthbee AI.
Last week’s guest – our 300th episode – was Katie Jacobs Stanton, GM of Moxxie VC (see the episode below). Our guest this week is Mark Drummond, CEO of Pixie.
Episodes drop on Fridays, and you can find them on podcasting platforms Spotify, iTunes, and YouTube.
Charlie Fink
AR Insider Contributor