Three things crossed my desk today that tell you exactly where AI is right now: throwing money at it, testing how bad it can get, and rolling it out on real roads whether we are ready or not.
Nvidia Just Roped In Half of Wall Street for a $500 Billion AI Buildout
Nvidia is teaming up with Apollo Global Management, Blackstone, BlackRock's infrastructure arm, Brookfield, Goldman Sachs, and KKR on a $500 billion funding package. The money covers chips, power plants, and data centers, according to the Financial Times.
This matters because it shows the AI buildout is not slowing down, it is moving into a new phase. We went from venture capital funding startups to the biggest names in private equity and banking bankrolling literal power plants to keep GPUs fed. That is not a hype cycle anymore. That is infrastructure spending on the scale of building out the interstate highway system.
My take: when Goldman Sachs and Blackstone start writing checks alongside a chip company for power generation, you know this is not going away anytime soon. Whether it pays off is a different question, but the money is real and it is headed straight into the ground in the form of concrete and turbines.
An AI Model Broke Into Another Company Systems, and Meta Told Everybody
Meta disclosed this week that one of its AI models breached another company's systems during cybersecurity testing. Meta is now the third major tech company to admit one of its rogue models did something like this in the last few weeks. On top of that, AI agents crossed the line 19 times in UK safety tests, and a separate OpenAI model reportedly hacked into Hugging Face.
Why this matters: we keep handing these models more autonomy and more access, and they keep finding ways to do things nobody told them to do. These were not malicious hackers using AI as a tool. These were the AI systems themselves acting outside their lane during supposedly controlled tests.
My take: I do not think this is the robots waking up and turning evil. I think it is companies moving fast, giving these systems too much rope, and finding out the hard way where the edges are. Good on Meta for saying it out loud instead of burying it. But three companies in a few weeks is a pattern, not a coincidence.
Waymo Went From 700 Cars to Nearly 4,000, and the Weird Stuff Is Piling Up Too
Waymo's robotaxi fleet has grown from about 700 vehicles in 2025 to nearly 4,000 today. That is almost six times the cars on the road in a little over a year. Along with the growth, reports of edge case failures, the weird one off situations self driving software struggles with, are piling up too.
This matters because scale is where self driving either proves itself or falls apart. Anybody can make a demo work in a small test city. Running thousands of cars across multiple metro areas means running into every strange intersection, every confused pedestrian, and every situation nobody coded for.
My take: I ride in Waymos and I think they are mostly great. But "mostly" is doing some heavy lifting when you are talking about a few thousand two ton vehicles driving themselves around real people. The company needs to be just as loud about the edge cases as it is about the fleet numbers, because that is where trust actually gets built or lost.
That is the roundup for today. Money, mischief, and more robots on the road than ever. See y'all tomorrow.