Why Midjourney Bought an Astrology App

On July 23rd, Midjourney founder David Holz posted a short note on the company blog. It took less than a minute to read.

Here’s what it said: over the next 6 months, the company would announce about six more projects of similarly ambitious scale. Honestly, he wrote, it’s exciting and terrifying at the same time, and the one thing that’s certain is that they need help. So he was announcing the company’s first-ever acquisition.

The target was Co–Star. The astrology app.

On its face, the news that an AI image-generation company had bought an astrology app reads like a joke. Plenty of foreign outlets covered it with exactly that “wait, what?” tone. But what Midjourney bought wasn’t astrological content — it was a product and design organization that has spent years getting users to apply unverifiable outputs to their own lives. And when you line this deal up against the other acquisitions and contracts Midjourney has signed this year, it becomes clear why that particular organization was the one it needed.

---

A company that made an estimated $500 million in annualized revenue on Discord alone, without ever building an app, just unveiled a full-body ultrasound scanner

Midjourney launched as a Discord bot in July 2022. It wasn’t on the App Store. It didn’t even have its own website with a polished onboarding flow. To use it, you had to join someone else’s chat server, type commands in a channel full of strangers, and accept that your output images would be visible to everyone.

That choice produced results nobody would have predicted. With zero free tier — everyone paid — its Discord server grew past 21 million members, making it the largest server on the platform. Revenue is estimated to have climbed from $50 million in 2022 to $200 million in 2023, $300 million in 2024, and roughly $500 million (~₩700 billion, ~$500M) in annualized run rate by 2026. The company never took outside investment and was profitable within 6 months of launch.

Midjourney discloses almost nothing about its own basic operations. Employee-count estimates vary wildly by outlet — 11, 40, 60, 100, 163. Forbes’ company profile, updated in April 2026, listed 60. Since the company confirms neither revenue nor headcount, every figure here is an estimate. Keep that caveat in mind for everything that follows.

It opened a web interface in 2024, but as of 2026, Midjourney still has no iPhone app and no Android app. This is a company that made $500 million riding on top of Discord for more than four years without ever building an app of its own.

Then, on June 17th, Holz unveiled a full-body ultrasound scanner at a San Francisco event — alongside a new division called Midjourney Medical. At the event, he previewed 8 products total: 4 pieces of hardware and 4 pieces of software. He also said the company would open a 25,000-square-foot spa in Union Square by the end of 2027, housing 10 of the scanners.

A company that had never built so much as a single app had just declared it would simultaneously ship eight products — including consumer hardware and a physical retail space.

For Midjourney, Discord was never just a distribution channel. Onboarding, payment flows, even the peer learning loop where users studied each other’s outputs — all of it rode on infrastructure Discord had already built. Building on someone else’s platform means paying the price of dependency, but it also means you never accumulate the experience of designing your own interface. In effect, for four years Midjourney had never once designed the screen a user sees first.

A month later, the Co–Star acquisition was announced. Midjourney’s official account said it had named Co–Star founder Banu Guler Chief Design Officer, putting her in charge of a “company-wide cross-functional design, product, and front-end organization.” Holz’s blog post didn’t go this far — it only mentioned “building out an organization spanning product and design” — so the title details come from the company’s official account, not from Holz himself.

Two things stood out to me about this hire. First, she was named design chief, not Chief Product Officer. Second, front-end was explicitly folded into her mandate. This isn’t a placement built around overall product strategy — it’s one built around everything a user directly sees and interacts with: the screens, the copy, all of it.

What Costar Was Actually Selling

Costar launched in New York in 2017. Founder Banu Guler, born to a Pakistani father and a Turkish mother, moved to New York at 17 and studied psychology at NYU. While working as a bicycle messenger, she taught herself graphic design, then led design and product at Diane von Furstenberg, Michael Kors, and VFILES before founding Costar. To date, the company has raised roughly $20.9 million.

The Costar app is built on three layers.

• A layer that calculates real-time star positions using ephemeris¹ data from NASA’s Jet Propulsion Laboratory
• A layer of text written by human astrologers
• A natural-language AI layer that combines the two into personalized sentences for each user

The first layer is verifiable. Celestial positions are empirical measurements, and anyone can check the calculations. What can’t be verified are the second and third layers. And yet what users actually consume isn’t the first layer — it’s the interpretation layered on top of it.

That interpretation is also what made Costar famous. Blunt push notifications like “Stop texting your ex” got screenshotted and spread far beyond the app itself. It’s a case where the notification copy itself — with zero ad spend — became the channel that publicized the app.

On scale: Sensor Tower estimates Costar’s monthly active users² at roughly 4.3 million, and the company itself stated that, as of 2021, it had already reached 20 million downloads and a quarter of American women aged 18 to 25. When the acquisition was announced, Guler wrote that the app had reached 35% of young Americans — a company claim with no independent verification.

There’s a line Guler wrote on the company’s team page. Costar, she said, isn’t just another horoscope app — it’s about that irrational feeling of being completely understood. I think that single sentence explains nearly everything about why this acquisition happened.

Ultrasound Scanners and Star Charts Share the Same Structure

Let’s go back to the scanner unveiled in June.

Here’s how it works. A person stands in a shallow water tank and slowly descends, while ultrasound elements embedded around the tank fire sound waves at the entire body and capture the signals that bounce back. According to the company, there are about 500,000 elements, the computation runs at over 2 petaflops, and within 60 seconds it reconstructs more than 25 organs and anatomical structures in 3D. The core hardware is 40 of Butterfly Network’s ultrasound chips, and the licensing deal is worth up to $74 million over 5 years. The company calls this “ultrasound CT.”³

But there’s no regulatory approval yet. Only about twelve people have been scanned so far, and no imaging data comparing results against MRI has been released. That’s why the launch plan Midjourney has announced isn’t framed as “diagnosis” but as a “body composition map.” In other words, it’s meant to be something you get on your way out of the sauna and cold plunge — a wellness experience, not a medical procedure, at least at the start.

Let’s lay the two structures side by side.

Co-Star: The raw data (celestial positions) is real. The interpretation layered on top of it sits outside verification. Users take that interpretation as a story about their own life.

Midjourney’s scanner: The raw data (ultrasound signals) is real. The interpretation layered on top of it sits outside regulation. Users take that interpretation as a story about their own body.

It’s the same structure. And in this structure, success or failure isn’t decided by data accuracy. Accuracy is either already secured, or even if it is secured, users have no way to verify it themselves. What actually determines the outcome is which sentence, at which moment, on which screen, delivers that interpretation.

Come to think of it, Midjourney’s original business ran on this same structure. There’s no “correct answer” for an image generated from a prompt — there’s no standard by which to judge it right or wrong. And yet nearly 20 million users pay for it every month. According to Bloomberg, Midjourney is also reportedly preparing an image generator specialized in astrology.

Still, the three cases differ in how risky it is when a user misreads the result. If you don’t like an image, you just regenerate it; if your horoscope for the day is wrong, you laugh it off. But a scan of the inside of your body is different. The moment a user takes that as “a story about their own body,” it can lead to someone who should see a doctor not going, or someone who doesn’t need to worry being consumed by anxiety. Questions have already been raised about whether running routine full-body scans on healthy people is even clinically meaningful. The ability to build trust well is what makes a product sell — but it can also lead users to believe interpretations that have never been validated.

Oswarld’s Lens

I kept running into the same scene while working on GTM strategy: a product where the technology clearly works, but nobody’s buying it. What these products lacked wasn’t usually engineers — it was someone who could translate a model’s output into sentences and screens that users could actually understand. This role always gets pushed aside because the title is ambiguous. It doesn’t look like design, or product, or marketing.

Midjourney ran its service for over 4 years without its own mobile app, with annualized revenue estimated at $500 million. That’s because someone else’s infrastructure — Discord — handled the interface for them. But the experience of stepping into a scanning tank isn’t something Discord can design on your behalf. So I don’t see this acquisition as simply adding headcount — I see it as a shift where the user-facing design work Discord used to handle now becomes something the company’s own internal organization takes on directly.

Still, there’s a part of this I’m skeptical about. Costa’s real asset is that users genuinely trust the sentences the app sends them — and the fact that the app is now owned by an AI company threatens to unsettle that feeling. Right after the announcement, X filled up with posts from people saying they’d delete the app because of the acquisition. Holz’s insistence that Costa would remain entirely under Guler’s control was probably a preemptive response to exactly this kind of backlash. But the discomfort that comes from a change in company ownership doesn’t just disappear because of a promise to keep operations independent.

And all of this is my own interpretation. Midjourney hasn’t disclosed the deal amount, headcount, or what the other six products even are. All we actually know right now is one blog post and one personnel announcement.

Closing

Here’s the recap.

1. Midjourney made $500 million without an app. Discord served as its interface. But if you’re building four pieces of hardware plus a physical space, you have to fill that role yourself.
2. Costa is a team that’s done exactly that for 7 years. Layering unverifiable interpretation on top of real measured data, and making it trustworthy enough that people screenshot and share it.
3. The scanner follows the same structure. The ultrasound signal itself is real measurement, but the interpretation layered on top hasn’t received regulatory approval yet — and users take it as a story about their own bodies.

If you’re building a product with an AI feature bolted on, check whether you have even one metric that measures “how users actually receive this output” sitting next to your accuracy metrics. Most dashboards only have the former. But in my experience, it’s often the latter that determines adoption.

💬 There are moments when you don’t fully trust what an AI tool spits out, yet you keep using it anyway. What built that trust for you? The tone of the sentences, the way evidence was displayed, or just plain familiarity? Tell me in the comments. Once I’ve gathered enough cases, I’ll sort them into types for the next issue.

📨 If someone around you is wrestling with the user-trust problem in AI products, pass this along.

References & Further Reading

Primary sources

• David Holz, “Midjourney’s First Acquisition”, Midjourney Updates, July 23, 2026. Read original ··· This is where today’s piece starts. It’s a one-minute read, but the line “we’ll be announcing six more projects” explains the whole context of the acquisition.
• Matt Novak, “Midjourney Buys Astrology App Co-Star”, Gizmodo, July 24, 2026. Read article ··· The article where Midjourney itself confirmed that the actual deal closed in spring, with the announcement only coming in July.
• “AI Startup Midjourney Pivots to Health With Ultrasound Machine”, Bloomberg, June 18, 2026. Read article ··· On-the-ground coverage of the scanner announcement. The spa rollout plan and the 50,000-unit target both come from here.
• “Midjourney Medical Explained: What Is the 60-Second Full-Body Ultrasound Scanner?”, Iatrox, June 2026. Read analysis ··· A piece that filters the company’s announcement through a regulatory lens. Especially worth reading for why the name “Ultrasonic CT” is so misleading.
• Co-Star, “Team”, costarastrology.com. View page ··· Contains Gulher’s own product philosophy, written in her own words. This does more to explain the acquisition’s background than any news article.

Background

• Kia Kokalitcheva, “Astrology app Co-Star raises $15 million in new funding”, Axios, April 14, 2021. Read article ··· Covers the scale of Co-Star’s early growth and Gulher’s own framing of the product (“astrology as a language, a frame”).
• Lucas Matney, “Co-Star, the astrology app, raises $5.2 million seed round”, TechCrunch, April 17, 2019. Read article ··· Shows how investors viewed an astrology app at the time. Nobody could have guessed that an AI company would buy it seven years later.
• “Midjourney Shipped Inside Discord Instead of Building an App”, Built Plain, July 2026. Read analysis ··· Lays out what choosing Discord bought Midjourney, and what it cost. This is the backbone of section 2 in today’s piece.

---

The author is Oswarld (Kwangseob Ahn). Current roles: Adjunct Professor at Sejong University, Strategy Consultant at INLEVEL9. Career, research, books, and recent work are kept current on the About page. Latest · July 2026: HEMA-2: A Consolidation-Aware Tri-Memory Architecture with Multi-Channel Scheduling for Lifelong Conversational AI.

📝 Glossary

Footnotes

1. Ephemeris: a table of the positions of celestial bodies over time. Co-Star pulls this data in real time from NASA’s Jet Propulsion Laboratory to calculate star positions. Up to this point, it’s straightforward astronomical calculation that anyone can verify. ↩

2. Monthly Active Users (MAU): the number of people who actually opened an app at least once in a given month. Unlike download counts, this excludes people who installed the app but never use it. That said, external estimates are derived by reverse-engineering app-market data, so they can diverge from a company’s actual internal figures. ↩

3. Ultrasonic CT: the name Midjourney gave its scanner. Despite the “CT” in the name, it uses no X-rays or radiation — only sound. Because the name is easily confused with conventional CT scans, medical experts have pointed out that the term is misleading. ↩