Midjourney Built a Cash Machine—Now It Wants to Take It Apart to Build Hardware

An AI Company That Does Not Need Funding Is Entering the Business That Needs It Most

The Information pairs two unusual facts: Midjourney is profitable and it is chasing hardware. The publication previously reported annual revenue was “significantly” above $200 million, generated without conventional venture funding.

Most frontier labs raise billions before explaining how scale will cure losses. Midjourney proved demand through subscriptions ranging from roughly $10 to $120 per month, then decided what to fund next. No preferred shareholders force it toward another valuation mark.

Hardware reverses that advantage. Software is delivered when a subscriber clicks. Hardware requires design, procurement, manufacturing, channels, inventory, returns and support. Cash flow moves from collecting first to spending first and guessing demand later.

Google’s 50 Billion Images Are Turning Aesthetic Competition Into Distribution War

Google disclosed at I/O 2026 that Gemini passed 900 million monthly users, more than doubling in a year, and Nano Banana models generated over 50 billion images. Live TikHub/X discussion poses a blunt consumer question: why pay Midjourney monthly when “good enough” generation sits inside tools people already use?

Midjourney can still charge professionals for style consistency, control, privacy and workflow. Google does not need to win every image. It can push ordinary demand to zero price and distribute through Search, Android, Workspace and Gemini.

Midjourney’s licensing partnership with Meta offers a different path: sell aesthetic technology to a platform with billions of users. Licensing is a low-capital extension. Building hardware is a high-capital extension with unproven distribution.

The Co–Star Acquisition Is a Test of Reversibility

Midjourney made its first acquisition in July, buying astrology app Co–Star and adding roughly two dozen employees. Compared with hardware, this is a lighter consumer experiment: it acquires a product team, user relationships and mobile experience.

The optimistic case is that Midjourney is learning to turn personalized AI into emotionally sticky consumer software. The pessimistic case is that strong cash flow has started searching for stories rather than letting customer demand direct capital.

Investors should demand reversibility. A small acquisition or prototype can be shut down with bounded loss. Inventory and supply-chain commitments cannot. Great allocation does not require always being right; it keeps the cost of being wrong small.

Five Master Investors Interrogate the Hardware Dream

Buffett would ask about the circle of competence. Midjourney has proved expertise in generative models, aesthetics and subscription communities—not procurement, retail or support. A circle is what the company can repeat with high certainty today.

Peter Lynch would warn about diworsification. Cash-rich growth companies often enter lower-return businesses. Hardware is an extension only if it strengthens the core; a larger market alone can turn a simple excellent story into a complex average one.

Philip Fisher would support genuine long-term R&D. He would still require a bridge to customer demand: what problem cannot existing devices solve, and do prototypes improve retention?

Howard Marks would separate possibility from odds. A huge upside does not guarantee high expected value. Probability, failure loss and occupied capital belong in the same calculation.

Charlie Munger would watch the man with a hammer. Holz’s Leap Motion experience may provide unique insight—or make hardware the most natural answer to a problem that does not require it.

The Best Anti-Google Strategy May Be to Stay Unlike Google

Midjourney cannot defeat Google in user scale, capital spending or default distribution. It can win by being smaller, more focused, closer to creators, willing to charge and free from the need to turn every product into an advertising or cloud funnel.

Track five proofs: core subscription growth, professional retention against free substitutes, high-margin Meta licensing, reusable consumer capability from Co–Star, and explicit budgets, milestones and stop conditions for hardware. Without those boundaries, the hardware dream deserves a higher discount rate.

Do not add a giant hardware market to valuation. Separate existing software cash flow from new-business burn. Valuefocus’s DCF calculator can model subscription, licensing and hardware independently, while the margin-of-safety tool tests whether the core still works if hardware fails.

Midjourney’s hardware move is not necessarily wrong. A dedicated device could create a new creative behavior and a physical gateway for its models. Until evidence arrives, however, it looks like a founder tax on a rare AI cash machine. A moat is not doing more. It is knowing which seductive opportunities to refuse.