Google I/O kicked off yesterday and it was a lot. Gemini Spark, smart glasses, a new $100 per month tier, and a proposed web standard for AI agents in the browser. Meanwhile Meta was busy axing 8,000 people and Google was cutting a $25 billion deal with Blackstone to build the data centers to power all of this. Busy Monday.
Google Unveiled Gemini Spark and Wants $100/Mo From You
The headline from I/O was Gemini Spark, which Google is calling a "24/7 AI agent." It sits behind the new AI Ultra subscription tier priced at $100 a month. That puts it square against OpenAI's Pro tier. Google also said Gemini is now processing 9.7 trillion tokens a month, which is a number so big it stops meaning anything.
The smart glasses demos were the other big moment. Google showed Gemini running on audio glasses without pulling out your phone, and announced at least three glasses coming this year from partners including Xreal. The AR glasses race is back, apparently.
For developers, Google introduced WebMCP, a proposed open web standard that lets browser-based AI agents call structured tools on websites. That one could be quietly important. If it sticks, it gives agents a cleaner way to interact with the web than scraping HTML.
Meta Cut 8,000 People to Pay for Its AI Bet
Meta announced roughly 8,000 layoffs this week, about 10% of the company. They also pulled the job postings for 6,000 open roles. At the same time, they bumped their 2026 capital expenditure guidance up by $10 billion to $145 billion.
The math here is pretty simple. Zuckerberg wants to spend big on AI infrastructure and the cheapest way to do that is headcount cuts. He has done it before and it worked. The stock usually rewards it.
The harder question is whether $145 billion in capex actually buys you a lead when Google, Microsoft, and Amazon are writing the same kinds of checks. At some point you are all just burning money to stay even.
Google and Blackstone Are Building the AI Power Grid
Speaking of big checks, Google and Blackstone announced a joint venture to build AI infrastructure. Blackstone is putting in $5 billion upfront with total investment potentially reaching $25 billion. The goal is 500 megawatts of compute online by 2027.
This is the part of the AI story that gets less attention than model releases but matters just as much. You can have the best model in the world and it does not help if you cannot run it fast enough for your customers. Power and cooling are the real constraint right now.
The Google-Blackstone deal is a sign that Big Tech is increasingly leaning on private equity to share the infrastructure cost. That works until interest rates make it not work.