Cognition just raised $1 billion at a $25 billion pre-money valuation, more than doubling what it was worth eight months ago. The company behind Devin, the AI coding assistant, is now running at $492 million in annualized revenue.

That’s not a projection. That’s actual run rate. For an AI coding tool that didn’t exist two years ago, it’s a remarkable number. It tells you two things: developers are paying real money for AI tools that actually work, and the market for this stuff is much bigger than the hype suggested.

The valuation jump is aggressive, but it’s backed by revenue growth that justifies it. When a company more than doubles its valuation in eight months while approaching half a billion in revenue, investors aren’t betting on potential anymore. They’re buying into a business that’s already working.

The cloud bill just got bigger

Snowflake signed a $6 billion, five-year deal with AWS for AI CPU chips. That’s $1.2 billion a year, and it’s a signal that the AI infrastructure build-out isn’t slowing down. It’s accelerating.

The deal is notable because it’s focused on CPUs, not GPUs. Nvidia isn’t getting this money. Amazon is. That matters because it suggests the industry is finding ways to run AI workloads without paying Nvidia’s premium for every cycle. It also means Snowflake is betting big on a future where AI inference happens at scale on more cost-effective hardware.

For AWS, this is validation. The company has been building its own chips, Graviton and Trainium, for years. Now one of the biggest names in data infrastructure is committing $6 billion to use them. That’s not a pilot program. That’s a strategic shift.

AI’s productivity story gets a real example

Remote, a payroll startup, grew revenue 50% per employee without adding headcount. The company just crossed $300 million in annual recurring revenue and became cash-flow positive. The reason: AI adoption.

This is the story everyone wants to tell about AI, but few companies can actually back up with numbers. Remote can. They’re doing more with the same number of people, and the gains are substantial enough to change the economics of the business.

It’s also a useful data point for anyone trying to figure out what AI actually does for a company’s bottom line. A 50% increase in revenue per employee isn’t marginal. It’s the kind of improvement that changes how you think about headcount, growth, and profitability.

The Polymarket problem

A Google employee allegedly used internal search data to win $1.2 million on Polymarket. Federal prosecutors charged Michele Spagnuolo with fraud, claiming he accessed confidential Google data to predict outcomes before the market knew.

This is a preview of a problem that’s going to get worse. Prediction markets are growing, and they’re increasingly tied to data that someone, somewhere, has early access to. The more these markets matter, the more valuable that access becomes, and the more people will try to exploit it.

Spagnuolo was arrested and released on a $2.25 million bond. The case will test how existing insider trading laws apply to prediction markets, which operate in a regulatory gray zone. If the government wins, it sets a precedent. If it doesn’t, the message is clear: the rules don’t apply here yet.

What it adds up to

Cognition’s raise, Snowflake’s chip deal, and Remote’s productivity gains all point in the same direction. AI is moving out of the research phase and into the part where money changes hands at scale. Companies are paying billions for infrastructure, raising billions on real revenue, and posting numbers that show AI is doing something measurable.

The Polymarket case is the asterisk. As AI and data infrastructure get more valuable, the incentives to cheat get bigger. The industry is moving fast, but the rules haven’t caught up. That gap is going to create problems, and this won’t be the last time someone gets charged for using data they shouldn’t have.