☁️ Together AI ($800M), 🔒 Anthropic & 🏗️ Meta Cloud — infrastructure as battleground

July 1, 2026 has handed us three headlines that, like pieces of the same puzzle, sketch out the immediate future of artificial intelligence. On one side, the multi-billion dollar investment in open-source infrastructure; on another, the shadow of regulation and export controls; and finally, the social media giant that decides to become a cloud provider. All point to the same truth: AI is being reshaped around who owns the hardware, who controls access, and who sets the rules of the game.

☁️ Together AI: $800 million to democratize the open-source cloud

The cloud platform for open-source models, Together AI, has closed a massive $800 million funding round, reaching a valuation of $8.3 billion. The company, which provides infrastructure for running and fine-tuning open-source AI models, has become the clear beneficiary of the developer exodus seeking alternatives to proprietary APIs from OpenAI or Google. Its pitch is simple but powerful: offering computing power optimized for models like Llama 3, Mistral, or Falcon, letting startups and enterprises customize their own algorithms without relying on a single vendor.

The backing of heavyweight investors like General Catalyst and existing backers shows the market strongly believes in the "open but managed AI" model. Together AI doesn't just sell GPUs; it sells an ecosystem where data control and flexibility are the main draws. At a time when big tech companies are tightening their usage policies, having a cloud specialized in open-source becomes a strategic asset.

🔒 Anthropic pulls its flagship models due to new export restrictions

In a move that has shaken the global AI community, Anthropic has removed its most advanced models — Fable 5 and Mythos — from public access to comply with new US government export control regulations. The decision immediately affects developers and researchers outside the country, who can no longer interact with these frontiers of AI. Anthropic, known for its focus on safety and alignment, has prioritized legal compliance over accessibility, sparking an uncomfortable debate about technological sovereignty.

This move is not an isolated one. With the US administration tightening restrictions on technologies deemed critical for national security, AI companies find themselves caught between global innovation and geopolitical demands. The immediate result is a new kind of digital divide: no longer just countries with or without internet, but countries with or without access to frontier models. Companies and governments in Europe, Asia, or Latin America will have to look for alternatives — perhaps in the open-source ecosystem that Together AI is boosting — or resign themselves to being a step behind.

🏗️ Meta builds its own cloud business to sell its AI surplus

Mark Zuckerberg has done it again: Meta, the parent company of Facebook and Instagram, is building a cloud computing business to sell its spare AI infrastructure capacity to external customers. The play follows Amazon's AWS playbook, but with a twist: Meta already owns one of the largest GPU fleets in the world, acquired to train and run its own models (like Llama 4 or its recommendation systems). Now, instead of letting those resources sit idle during low-demand periods, it will rent them out to third parties.

If the plan succeeds, Meta will go from being a social media and metaverse company to a top-tier cloud infrastructure provider. Competition in the cloud market is heating up: AWS, Azure, and Google Cloud will no longer be the only giants. Meta could offer aggressive pricing thanks to its scale, and most importantly, native integration with its open-source AI ecosystem. Startups wanting to fine-tune models based on Llama could have a direct incentive to host on Meta Cloud, creating a virtuous (or vicious, depending on your perspective) cycle that reinforces the company's dominance.

🌐 The thread that connects them

These three stories, though seemingly from different domains, share a common nerve: control of AI infrastructure as a lever of power. Together AI shows that money flows to those offering independence from a single vendor (open-source in the cloud). Anthropic, for its part, reminds us that independence can be cut short by geopolitical decisions that leave half the world locked out of the most advanced models. And Meta, with its new cloud business, adds another layer: the company that once only consumed AI now produces and sells it, creating a closed ecosystem around its own models. Taken together, we're witnessing AI stop being just a product and become critical infrastructure, with winners and losers defined by who has the chips, who writes the rules, and who decides who gets to rent them.