But the money story is not the one I find most interesting today.
Google Drops Gemma 4 and It Actually Matters
Google released Gemma 4, their latest open model series, and this time they built it specifically for reasoning and agentic workflows. It ships under Apache 2.0, which means you can actually use it for real work without a legal team hovering over your shoulder.
Why does this matter? Because the gap between closed and open models keeps shrinking. Every time Google drops a capable open model, it puts pressure on the whole ecosystem to be more accessible. If you are building agentic apps and you have not looked at what open models can do lately, you are leaving options on the table.
My take: Google has figured out that winning the AI race is not just about having the best closed model. It is about being the foundation that everyone else builds on. Gemma 4 is a chess move, not a charity project.
MCP Just Became Infrastructure
Anthropic's Model Context Protocol hit 97 million installs in March. Let that sink in for a second. Every major AI provider now ships MCP-compatible tooling. This is not an experiment anymore. This is plumbing.
If you have been on the fence about building with MCP, the fence is gone. When something crosses the 90-million-install mark and every big player supports it, you are not adopting early anymore. You are just adopting.
I have been saying for months that the real unlock in AI is not smarter models but better tool integration. MCP is proving that out. The models are smart enough. What they need is access to your actual systems, your actual data, your actual workflows. That is what MCP gives them.
The OpenAI IPO Question
Back to that $25 billion number. OpenAI going public would be one of the biggest tech IPOs in recent memory. But here is the thing that nobody is talking about enough: an IPO changes incentives. Public companies answer to shareholders. Shareholders want growth. Growth in AI right now means enterprise contracts and consumption-based pricing.
That is not necessarily bad for developers and builders. But it does mean the era of generous free tiers and loss-leader API pricing has an expiration date. If you are building on OpenAI's APIs, now is a good time to think about what your plan B looks like. Not because the sky is falling, but because smart builders always have options.
Between Gemma 4, MCP maturity, and the shifting business dynamics of the big AI labs, the message this week is clear: the tools are ready, the standards are settling, and the business models are hardening. Time to build something with all of it.