Meta Shuts Down Core Metaverse Project — An $80 Billion Lesson Learned

On March 20, 2026, the AI industry witnessed several major developments: Meta officially announced the shutdown of its metaverse flagship product Horizon Worlds on VR, Google tightened free access to Gemini CLI, OpenAI accelerated its programming ecosystem through an acquisition, and the Turing Award was given to the quantum field for the first time. These moves collectively point to one trend — tech giants are comprehensively reallocating resources, with AI becoming the undisputed strategic priority.

Meta Shuts Down Horizon Worlds VR: A Strategic Pivot After $80 Billion

Meta announced it will shut down the VR service of its core metaverse product, Horizon Worlds. According to the official timeline, the product will be removed from the Quest app store on March 31 and will completely cease VR operations on June 15, retaining only its mobile service.

The weight of this decision cannot be understated. To date, Meta's Reality Labs division — the unit carrying its metaverse ambitions — has invested more than $80 billion cumulatively. From Zuckerberg's high-profile rebranding of Facebook to Meta in 2021, going all-in on the metaverse, to the quiet exit of its core product today — barely five years have passed.

This business contraction marks a further tilt of Meta's strategic focus toward AI. For the broader industry, it also serves as a sobering reminder: capital-intensive, long-cycle frontier tech projects can be forced to scale back before achieving commercial viability, even with megacorp-level funding. The metaverse isn't a false premise, but under the impact of the AI wave, resource allocation priorities have fundamentally shifted.

Gemini CLI Tightens Free Access: Compute Costs Force Faster Monetization

Google officially released an update announcement for Gemini CLI services, with new policies taking effect from March 25. Key changes include:

• Free users lose access to the Gemini Pro model — free users are now limited to the lightweight Gemini Flash model
• Pro model access is restricted to Google AI paid subscribers, with AI Pro plans starting at $19.99/month
• Simultaneous upgrade of abuse detection mechanisms, specifically targeting "compute freeloading" behavior where users connect Gemini CLI to third-party software via OAuth authentication

The logic behind the adjustment is clear: inference costs for large models remain stubbornly high, and freely offering high-performance models is no longer sustainable. Google's move aims to standardize developer usage and control AI compute costs, directly impacting the global community of developers using Gemini CLI. It's foreseeable that other LLM providers will follow with similar tiered access strategies.

OpenAI Acquires Astro: Accelerating the Codex Programming Ecosystem

OpenAI announced plans to acquire Astro, a Python developer tools startup. Upon completion, the Astro team will be integrated into the Codex project. Astro previously developed UV, Ruff, and other open-source tools widely popular in the Python community. This acquisition will significantly enhance Codex's developer service capabilities.

Codex's growth momentum is currently very strong:

• Weekly active users have surpassed 2 million, tripling since the beginning of 2026
• India is one of Codex's fastest-growing international markets
• OpenAI has entered a strategic partnership with India's Tata Group, with Tata Consultancy Services fully adopting Codex tools across its engineering operations

OpenAI's intent is crystal clear: by acquiring quality developer tool teams, it's filling in the ecosystem puzzle for AI-powered programming and pushing AI-native software development toward standardization. This isn't just a product-level expansion — it's a battle for developer mindshare.

Turing Award Goes to Quantum for the First Time: The Future Direction of AI's Underlying Compute

The Association for Computing Machinery (ACM) officially announced the 2025 Turing Award recipients, honoring IBM researcher Charles Bennett and University of Montreal professor Gilles Brassard for their foundational contributions to quantum information science.

This is the first time since the Turing Award's establishment in 1966 that it has been given to a quantum physics-related research field. The BB84 quantum cryptography protocol they proposed laid the technical foundation for quantum-secure communications. Their work holds milestone significance not only in quantum computing and quantum communication but also provides important theoretical support for the evolution of AI's underlying compute. As classical computing architectures gradually approach physical limits, quantum computing may become the key breakthrough for next-generation AI compute.

Domestic Developments: Xiaomi Voice LLM and Alibaba Cloud's ¥100 Billion Milestone

Xiaomi Releases Self-Developed Voice Synthesis LLM

Xiaomi officially released its self-developed voice synthesis large model MIMO VR-PTS, designed specifically for full-modality interaction scenarios in the Agent era. The model is built on a proprietary Audio Tokenizer and multi-codebook joint speech-text modeling architecture, pre-trained on hundreds of millions of hours of voice data and refined through multi-dimensional reinforcement learning.

Its core highlight: beyond natural text reading, it can accurately reproduce non-verbal features like sighs, coughs, and laughter, support tonal shifts and emotional gradients within a single sentence, and intelligently recognize text formatting and punctuation to automatically match natural pauses and intonation. It also supports dialect rendering and song performance, marking an important step forward in humanized expressiveness for voice synthesis.

Alibaba Cloud External Revenue Breaks the ¥100 Billion Mark

Alibaba Group CEO Wu Yongming announced during the earnings call that Alibaba Cloud's external commercial revenue for FY2026 (through end of February) has officially surpassed ¥100 billion RMB, making it the first domestic cloud provider to cross this threshold. Even more noteworthy is the strategic target: within the next 5 years, Alibaba Cloud's annual cloud and AI commercial revenue — including MaaS — will exceed $100 billion USD.

Over the past three months, Token consumption on the Bailian MaaS platform's public model service marketplace grew 6x, fully demonstrating the powerful growth momentum of AI commercialization.

AI Copyright Disputes Continue to Heat Up: Anthropic Sued Again by Music Giant

Another major lawsuit has been filed in the global AI copyright arena. International music giant Bertelsmann Music Group (BMG) officially sued Anthropic in a California federal court, alleging that Anthropic used hundreds of copyrighted lyrics from artists including The Rolling Stones and Bruno Mars without authorization when training the Claude series of large language models.

This is yet another major music copyright lawsuit against Anthropic, following earlier actions by Universal Music and other rights holders. The copyright compliance issue surrounding AI training data is moving from industry discussion to a critical phase of legal adjudication.

Other Notable Developments

• NVIDIA doubles down on AI inference chip investments: Continued investment in AI inference architecture company Groq, whose proprietary LPU chips are expected to ship 4–5 million units cumulatively in 2026–2027
• Ant Digital launches AI security protection system: Tianjian 2.0 focuses on three core capabilities — combating ideological drift, purifying Skills repositories, and risk intelligence broadcasting
• 2026 Zhongguancun Forum Annual Conference: To be held in Beijing from March 25–29, expected to release multiple major scientific and technological achievements in cutting-edge fields including global AI and integrated circuits

Summary

From Meta shutting down the metaverse to Google restricting free access, from OpenAI expanding its programming ecosystem to the Turing Award embracing quantum computing — the tech industry in 2026 is undergoing a profound resource reallocation. AI is no longer just a track; it's becoming the core strategic fulcrum for all tech giants. For developers and industry practitioners, understanding the logic behind these changes matters more than chasing any single technology trend.