Amazon announced that GPT-5.5, GPT-5.4, and OpenAI’s Codex coding agent are entering limited preview on Bedrock, with wider rollout expected soon. AWS is also introducing Bedrock Managed Agents powered by OpenAI, making it easier for businesses to build and deploy enterprise-ready AI agents without managing the heavy infrastructure themselves.

In simple terms: OpenAI is no longer tied to one cloud giant.

This is bigger than just product access-it signals a power shift. Companies can now choose where they want to run OpenAI tools, instead of defaulting to Microsoft’s ecosystem. For AWS, this is a major strategic win, giving Amazon a stronger position in the AI cloud race alongside Google and Microsoft.

It also adds context to Amazon’s reported multibillion-dollar OpenAI investment from earlier this year-something that now looks far more strategic.

Why this matters:
The AI industry is moving away from exclusive “one lab, one cloud” partnerships. As demand for advanced AI infrastructure keeps exploding, no single provider may be able to dominate access forever.

The bigger question now:
If AI becomes too large for one cloud to contain, are exclusive alliances becoming obsolete?

What’s changing?
Code found inside the Gemini app suggests Google is testing a monthly credit system, where users would get a set number of credits each month and could buy extra if they run out. Instead of strict prompt caps based on your plan, those credits could be used more flexibly across different Gemini models and features.

At the same time, a new “Images” section has appeared in Gemini’s web interface, labeled “NEW,” which could point to a built-in image editor or upgraded image-generation tools coming soon.

Why this matters:
This would bring Google closer to the pricing strategies already used by OpenAI, Anthropic, and Notion where usage-based systems are gradually replacing rigid subscription limits.

Google already uses this kind of credit structure in products like Flow, Whisk, and Antigravity, so bringing it into Gemini would create a more unified AI billing system across its ecosystem. It could also help bridge the huge pricing gap between AI Pro ($19.99) and AI Ultra ($249.99), giving users more options without forcing them into an expensive tier jump.

A likely reveal could happen at Google I/O on May 19–20.

The bigger picture:
The era of simple unlimited AI subscriptions may be fading. As AI models become more advanced, agentic, and expensive to run, companies are increasingly shifting toward consumption-based pricing.

For heavy users, this could mean more freedom and customization. For casual users, though, it may also mean one new challenge: figuring out how much their AI habits actually cost.

In short: AI pricing is starting to look less like Netflix… and more like your cloud bill.