Hey, I’ve Been Thinking…

Week in Tech #6: Sixty Billion Dollars for a Code Editor

week in tech #6

Written by

Adem Ogunc

in

Wednesday, April 29, 2026.

I run a rug company. I am not a tech reporter. I just read a lot.

A note on the timing of this one. I usually publish on Sunday or Monday. This is going up on a Wednesday because I spent the past five days at High Point Market, the largest furniture trade show in North America and my actual day job. Rugs come first. Newsletter is something I write on flights and weekends. The trade show won this round.

Im writing this on the way home, with a half cold coffee and a notebook full of supplier conversations that will keep me busy for the next three weeks. The week did not slow down for me. So lets get into it.

Three numbers from the past ten days.

SpaceX is paying up to sixty billion dollars for a code editor. If they decide not to buy it, they have to pay ten billion dollars anyway. The floor is ten billion. The ceiling is sixty billion.

Amazon committed up to another twenty five billion dollars to Anthropic, on top of the eight billion they had already put in.

Tesla raised its 2026 capex guidance to over twenty five billion dollars, three times what they spent last year. Most of the increase is going to AI infrastructure.

I read these numbers and at first I tried to file them in three different folders. A code editor deal. A model lab investment. A car company spending on robots. By the time I landed I realized they are the same story.

Compute is the new currency. Whoever owns the chips, the data centers, and the electricity to run them sets the terms for everyone else.

Let me walk through it.

What SpaceX actually bought

Start with the headline because it deserves to be the headline.

If you have not heard of Cursor, do not feel bad. Most people outside the developer world have not.

Cursor is a piece of software that helps engineers write code. It is a fork of Microsoft Visual Studio Code, the most popular code editor on Earth, with AI built into every part of the experience. You type, the AI completes your line. You select code, the AI rewrites it. You describe what you want in plain English, the AI builds it. The company behind it is called Anysphere, founded in 2022 by four MIT engineers in their twenties.

Heres what makes the price tag rational. Cursors revenue went from one hundred million in January 2025, to five hundred million in June, to one billion in November, to two billion in February of this year. Three years from launch to two billion in annual recurring revenue. Sixty seven percent of the Fortune 500 are paying customers. It is the fastest growing software company in history.

Why does SpaceX want it. Two reasons.

First, SpaceX is preparing to IPO this summer at a reported one point seven five trillion dollar valuation. Bolting two billion dollars in real software revenue from real Fortune 500 customers onto that story turns SpaceX from a rocket company into a credible AI company. Wall Street pays a much higher multiple for AI revenue than for rocket revenue.

Second, xAI, the AI lab that merged into SpaceX in February, has been embarrassed by Grok in the developer community. Cursor is the gateway. Whoever owns the editor that developers stare at all day controls which model gets called for the highest margin task in software. That is the prize.

For Cursor, the deal solves a real problem. Cursor has been running at near zero margins on parts of its product because it pays retail rates to call OpenAI and Anthropic models on every keystroke. Those companies are also Cursors competitors. There is a famous cautionary tale here. When OpenAI tried to buy a Cursor competitor called Windsurf in 2025, Anthropic immediately cut off Windsurfs access to Claude. Within weeks the company collapsed and was sold for ten cents on the dollar. Any AI coding company that depends on a frontier labs API is one phone call away from extinction.

SpaceX brings something different. SpaceX brings Colossus, the supercomputer xAI built in Memphis, equivalent in scale to roughly one million Nvidia GPUs. Cursor with its own dedicated compute supply is a different company than Cursor as a tenant on someone elses API.

One piece of color, because tech history is full of this stuff. Sam Bankman Frieds bankrupt FTX held a five percent stake in Cursor that they bought for two hundred thousand dollars in 2022. The bankruptcy estate sold it in 2023 for the same two hundred thousand dollars. That stake would be worth roughly three billion dollars today. The most expensive yard sale in financial history.

The pattern

Now look at the other two big deals through the same lens.

Amazons twenty five billion to Anthropic is not really a software investment. The structure is that Anthropic gets the cash, and in return Anthropic commits to spend over one hundred billion dollars at AWS over the next ten years, running their training and inference on Amazons custom AI chips. Amazon is not investing in AI. Amazon is locking in a long term consumer for the chips and the data centers Amazon already built. Anthropic also disclosed run rate revenue greater than thirty billion. Up from nine billion at the end of 2025. Tripled in four months.

Teslas capex bump is even more direct. Fifteen billion more next year than this year, mostly on AI infrastructure. Optimus production lines, Cybercab production lines, more Dojo training compute. The market punished the stock for it. Operators want returns on capital and Tesla just told them returns are coming later, after a much bigger bill.

A few other deals from the same week worth knowing about.

Bezoss new physical AI lab, Project Prometheus, closed a ten billion dollar round at a thirty eight billion dollar valuation. They are building robots and physical AI systems. Total funding now over sixteen billion.

Vast Data, the storage company that powers most of the AI training in the West, raised one billion dollars at a thirty billion dollar valuation. More than three times its last mark. Customers include CoreWeave, xAI, Mistral, and Cursor.

Three deals. Three different industries. One underlying bet. The bet is that demand for AI is about to explode beyond anything we have seen, and the thing in short supply will not be models or talent or capital. It will be compute. Chips, electricity, the buildings to put them in. Every one of these companies is rushing to lock up that supply now, while they still can.

The other big story this week

While I was at High Point Market, the lawsuit Elon Musk filed against OpenAI two years ago started its jury trial in a federal courtroom in Oakland.

The short version. Musk co founded OpenAI in 2015 as a nonprofit, resigned from the board in 2018 after losing a power struggle, sued in 2024 alleging the company was illegally converted from a nonprofit to a for profit while the founders kept equity. The trial started Monday April 27. It is expected to run about four weeks.

Three things matter for non lawyers.

One, on Friday Musk dropped his fraud claims. Of the original twenty six causes of action, only two are going to trial. Breach of charitable trust, and unjust enrichment. He concentrated his fire on the strongest theory.

Two, there is a 2017 personal journal entry from Greg Brockman, one of OpenAIs co founders, that the judge has already cited in court. The entry reads, in part, that converting the nonprofit to a for profit without Musk would be morally bankrupt. That is a problem for OpenAI.

Three, if Musk wins anything, even on the narrow charitable trust theory, OpenAIs planned IPO this fall at a reported one trillion dollar valuation could be in serious trouble. Microsofts roughly twenty seven percent stake, worth about one hundred thirty five billion dollars, would be in question. Anthropic, which is also a public benefit corporation, is watching closely.

If you want the deep version, court filings are public on PACER. I am not going to pretend to be a legal analyst.

And the rest

A few other things I found interesting.

OpenAI shipped GPT 5.5 last Wednesday. Six weeks after the last update. ChatGPT now has nine hundred million weekly active users. The cadence between major frontier model releases is now roughly six weeks. That is the fastest weve ever seen.

China dropped three open weight AI models in four days. DeepSeek V4. Moonshots Kimi K2.6. Alibabas Qwen 3.6. The smaller DeepSeek model has API pricing of fourteen cents per million input tokens. That is essentially free. China is racing to make frontier intelligence a commodity.

Google held Cloud Next in Las Vegas. Two hundred sixty announcements. Sundar Pichai disclosed that seventy five percent of new code at Google is now written by AI. The Google CEO saying out loud that three quarters of Googles code is AI generated is a moment worth pausing on.

Vercel, the company that hosts a lot of the web, got breached through a third party AI tool called Context.ai. Attackers stole OAuth tokens, pivoted into Vercel employee accounts, and made off with customer environment variables. Lesson for everyone who runs a business. Every AI productivity tool you connect to your work accounts is now a credential. Audit those connections this week.

Why this matters for me, and probably for you

I keep thinking about how this trickles down to a small business.

I sell rugs. I sell them on more than a dozen channels. I built our AI infrastructure myself over the last two years, mostly because I had to. The cost of running our AI agents right now is real but tolerable.

If the bet these companies are making is correct, the price of running that same agent in two years is going to be a tenth of what it is today. Maybe less. The supply of compute is going up faster than anyone can use it at the model layer, and DeepSeek just released a frontier class model with API pricing close to free. That changes my business. I can do things in 2027 that I cannot afford to do in 2026. The same is going to be true for whatever business you run.

The other thing I keep thinking about is more philosophical. The companies that win this decade are not going to win because they have the best model. The model is becoming a commodity. They are going to win because they figured out, early, which interface in their business an AI agent should reach for. The interface a worker stares at all day, every day. For developers, that interface is the code editor. That is why Cursor is worth sixty billion.

For everyone else, the answer is different and not obvious. For a rug company, I have some guesses. The buyer dashboard, maybe. Or the supplier portal. Or the internal pricing tool. I dont know yet. I am going to spend a lot of this year trying to figure it out.

What I keep coming back to is that if you run a small or mid sized business and you are still wondering whether AI matters, the three biggest companies in tech just told you the answer in three numbers. Sixty billion. Twenty five billion. Twenty five billion.

The companies that wait for this to be obvious are going to be twelve to eighteen months behind. The ones moving now will be fine.

See you next week. On Sunday this time.

This is Issue #6 of “The Week in Technology,” a weekly newsletter at ademogunc.com.

Sources

About me: Im Adem Ogunc. I run Well Woven, a rug company based in Easton, PA, and FurniPulse, a home furnishings trade intelligence platform. Ive been in the rug industry for over 20 years and building with AI tools for the last couple. I write this newsletter on flights and weekends because Im genuinely fascinated by whats happening in technology right now.

Issue #6.

Comments

Leave a comment

More posts