App Economy Insights — The M&A Land Grab

Salesforce sells software businesses use to manage customers. Its big bet now is "Agentforce" — AI agents that do work automatically — but investors doubt the flashy marketing matches what the product actually does. So Salesforce bought Fin (formerly Intercom) for $3.6 billion: a proven AI agent that already handles customer-support questions for 30,000+ companies and brings in over $400 million a year.

The logic the article highlights: customer support is the cleanest place for AI to prove itself — huge volume, an obvious payoff (fewer human agents), and easy-to-measure results. By bolting a working agent onto its stack, Salesforce is buying the "agent layer" — the part that actually does the task — rather than just talking about it. It was the cheapest of the week's three deals and the one with the clearest near-term return; the catch is that mashing another product into an already-sprawling software suite is the hard part.

Fox owns live sports, news and broadcast TV but had no big foothold on the actual screen people stream from. So it agreed to buy Roku for about $22 billion to grab the "distribution layer" — the menu and platform millions of TVs open to, where ads get sold. As Americans keep shifting from cable to streaming (now nearly half of all TV watching), owning that doorway becomes valuable.

Investors hated it: Fox stock dropped 17%. The reasons are the price, the fact that Fox is borrowing $12 billion to pay for it, and that Roku is a contested prize — Amazon, Google, Samsung and Walmart all want the same living-room real estate. The article frames it as the riskiest of the three deals: a real strategic gap being filled, but possibly overpaying for a fought-over asset with debt, all while Fox still has to keep writing huge checks for sports rights.

Roku makes the operating system and home screen on tens of millions of TVs — it's the interface sitting between viewers, the streaming apps, and advertisers. It doesn't make most of its money from selling devices; it makes it from that platform (about $4.1 billion, ~88% of revenue) through advertising, taking a cut of subscriptions, and selling prime spots on the home screen.

That's exactly why Fox is paying ~$22 billion for it: Roku reaches 100M+ households (about a quarter of US connected-TV homes), and whoever controls the interface controls where attention — and ad money — flows. The risk flagged in the piece is that Roku's position is contested by much bigger players (Amazon, Google, Samsung, Walmart's Vizio), so it's a valuable but defendable, not unbeatable, control point.

Cursor is a popular AI-powered code editor — the tool many programmers now type into while an AI helps write the software (the "vibe coding" trend). SpaceX is buying its parent, Anysphere, for $60 billion entirely in SpaceX stock, just days after SpaceX went public. The prize is the "workflow": the place where developers actually do their work, which is where AI is being adopted fastest.

Owning that gives SpaceX the daily workflow, a stream of data on how people use AI to code, a built-in audience to push its own xAI models to, and the chance to run it all on its own Colossus computers. Two cautions the article raises: SpaceX is paying with stock that may be temporarily inflated from its hot IPO, and it may be paying a "frontier-model" price for what could just be a product that bigger AI suites (Claude Code, OpenAI's Codex, GitHub Copilot) eventually copy. The open question is whether Cursor stays a durable platform or becomes a feature.