December 11, 2025 · 24-Hour AI Briefing: Meta Turns Closed-Source, Apple Bets on Glasses + iPhone, and Microsoft Pours $17.5B into India
In the last 24 hours, three tech giants made telling moves across the AI stack:
Meta is pivoting from open-source champion to closed-source powerhouse, Apple is doubling down on a “glasses + iPhone” transition path instead of sci-fi headsets, and Microsoft is making its largest-ever Asia investment to turn India into a cloud and AI infrastructure hub. Here is what actually matters behind the headlines.
1. Meta pivots from open-source champion to closed-source AI giant, and buys Rivos for $4B
Meta is shifting its AI strategy toward a more closed-source, commercial model.
The company is reportedly preparing a new closed-source model for release next spring, trained using multiple third-party tools and data providers, including Alibaba’s Qwen (Tongyi Qianwen).
At the same time, Meta and Intel engaged in a bidding war over AI chip startup Rivos, driving its valuation up to around $4 billion. Meta ultimately won the deal. Rivos focuses on custom AI chips and data center acceleration, aiming to become a key building block in the next-generation compute stack.
Commentary: Meta looks like it is completing its transformation from “open-source idealist” to “full-blown commercial AI giant.”
With the Llama series, Meta briefly became the symbol of resistance against the closed models of OpenAI and Google. But Llama 4 failed to deliver a decisive breakaway moment, and the developer ecosystem never fully translated into durable commercial power. Now Meta is aggressively hiring, acquiring, and quietly tightening control over its models — effectively admitting that “free open-source goodwill” alone cannot finance a trillion-dollar AI arms race.
The $4 billion acquisition of Rivos is really about owning the bottom of the stack: whoever controls their own AI chips has a shot at reducing their dependence on Nvidia in the next phase. But it is also a high-stakes bet: if the architecture works, it becomes a defensible moat; if it fails, it turns into a very expensive sunk cost written into an entire generation of hardware. For developers and the broader open-source community, the practical question is: if Meta is no longer the friendly open ally, who becomes the new anchor for truly open AI?
2. Apple Glasses: from sci-fi MR headsets back to “shippable consumer electronics”
Apple plans to launch its first smart glasses, Apple Glasses, in 2026, with commercial availability in 2027.
The device is positioned as a lightweight smart wearable accessory without full standalone compute power. Instead, it will offload part of the processing to an iPhone. At the same time, Apple has reportedly paused development of the lighter Vision Pro variant (once codenamed Vision Air) to concentrate engineering resources on Apple Glasses.
Commentary: Apple has effectively admitted that, within a 1–2 year window, it is almost impossible to build a device that is light, comfortable, has full compute on board, and still delivers premium visual quality at a consumer-friendly price. Rather than doubling down on bulky, expensive MR headsets with unclear mainstream appeal, Apple is stepping back to a more pragmatic combo: “glasses + iPhone.”
Fully enclosed MR headsets are unlikely to become mass-market devices in the near term. By contrast, Apple Glasses can plug directly into an ecosystem of 1.7 billion active iOS devices and a mature App Store. Users get a new form factor with minimal friction — they keep their apps, accounts, and workflows. This kind of end-to-end integration is exactly what Meta and Google struggle to match.
Apple’s move looks less like a retreat and more like a reset: from sci-fi hardware to something that can actually be mass-produced, priced, and shipped at scale. If Apple Glasses can hit the right balance of comfort, functionality, and price, it could quietly lock up a huge share of the “everyday AR” market before its rivals are even ready.
3. Microsoft’s $17.5B bet on India: cloud, compute, and a “friend-shoring” experiment
Microsoft has announced a $17.5 billion investment in India over the next four years, focused on cloud services and AI infrastructure.
This is Microsoft’s largest investment in Asia to date and will fund new data centers, cloud platforms, AI deployment, and local developer ecosystem initiatives.
Commentary: This is a very deliberate bet on India as both a digital economy and a geopolitical construct.
On one side, India combines population scale with a massive pool of software engineers, already ranking as the world’s third-largest digital economy. It is not just a consumption market, but a lab for AI model deployment, cloud adoption, and talent development. On the other side, as U.S. controls on advanced chips and cloud services to China tighten, India is being positioned as a “friend-shored” alternative growth engine.
India’s data localization and “data trust” narrative adds another layer: data should stay in India; compute should be provisioned locally; services can then be exported globally. By pouring money into local infrastructure, Microsoft is effectively buying into India’s digital sovereignty story — and helping shape it.
Yet none of this is frictionless. Power reliability, land costs, regulatory approvals, tax complexity, and evolving tech regulation all introduce serious execution risk. This $17.5B commitment is both a land-grab for future growth and a live test of whether “sovereign-flavored cloud + AI” can scale in emerging markets.
To zoom out and connect the dots, you may also want to revisit our recent briefings on the most important developments of the last 72 hours:
2025 is quickly turning into the year when AI moves from pure model races to a full-stack game: models, chips, data centers, regulatory positioning, and everyday devices all start to intertwine. The players that can control more than one layer of this stack are the ones that will still matter five years from now.