He Rejected Apple Twice Before He Created Midjourney

David Holz's path from a childhood of obsessive tinkering to founding Midjourney reveals that his real education came from the failure of his earlier venture, Leap Motion, a pioneering hand-tracking company whose technology was genuinely ahead of its time. Despite raising over $40 million, attracting top investors, and even drawing acquisition interest from Apple, Leap Motion never achieved product-market fit because the market, user habits, and ecosystem weren't ready for touchless 3D interaction. A key theme is that user education and timing matter as much as innovation: Holz struggled with teaching people how to learn new interactions, and the VR/AR market collapsed just as the company pivoted toward it. Rejected Apple buyouts, a disastrously low $30–50 million offer, and a final fire-sale acquisition for $30 million underscored the brutal gap between technological vision and commercial reality. The ultimate lesson—that a great product cannot succeed if the market isn't ready—directly shaped Holz's approach with Midjourney, which he bootstrapped with no external funding and built into a profitable, tiny-team success by waiting until the world had caught up with his ideas.

• ▶ 0:16 Midjourney, founded by David Holz, is highlighted as one of the most interesting AI companies—a seeming overnight success that actually took years to build and now earns tens of millions in profit with a tiny team.
• ▶ 1:19 David’s unusual childhood—his father was a dentist working from a boat—led to early access to a computer, obsessive learning, and hacking games like Star Wars: Dark Forces, a classic path for future tech founders.
• ▶ 3:03 Despite impressive research work at NASA and the Max Planck Institute, David grew restless with academia and sought a bigger impact, setting the stage for his later leap into entrepreneurship.
• ▶ 3:12 David started a company called OcuSpec to make a bigger impact, but it was quickly renamed Leap Motion.
• ▶ 3:31 The leap’s core concept came from a middle school typing class, where David realized that seamlessly transmitting fast-moving thoughts into technology was a fundamental problem.
• ▶ 4:12 This thinking produced the Leap, a small device that tracks hand motion and translates it into the computer, with the interaction model being the most important design factor.
• ▶ 4:17 Leap Motion arrived years before the market was ready: the iPad had just launched, Windows had no touch support, and most people only associated touch with phones and trackpads.
• ▶ 5:13 Alan Kay advised David to “think about how the user learns” rather than features—a problem Leap Motion never truly solved.
• ▶ 5:27 User education remains the lasting challenge, both for people adopting new tech and for training algorithms, explaining why Leap Motion’s vision was ahead of its time.
• ▶ 5:38 Leap Motion's 2011 seed round attracted top investors like Andreessen Horowitz and Founders Fund, but not all VCs grasped the technology's vision.
• ▶ 5:45 During a pitch, David demoed a 3D "Minority Report"-style interaction, only for a lead investor to suggest giving away the first two dimensions for free and charging for the third, revealing a fundamental misunderstanding.
• ▶ 6:36 The anecdote illustrates how poor VCs jump to monetization without understanding the technology, ignoring that David's real challenge was novel hardware—not pricing—and likely fueling his later preference for bootstrapping.
• ▶ 6:58 Midjourney is totally bootstrapped, having raised no external funding.
• ▶ 7:09 Leap Motion raised a $12.75 million Series A in 2012.
• ▶ 7:13 Leap Motion raised a $30 million Series B in 2013, led by Founders Fund, to fund product release.
• ▶ 7:17 Leap launched in mid-2013 thanks to new funding, selling about 500,000 units—below expectations but still solid for a new hardware product.
• ▶ 7:32 Early launch problems included tricky tech support and weak user retention, indicating engagement issues after purchase.
• ▶ 7:36 The sales figure may have been inflated by partnerships, meaning it combined direct consumer sales with bulk or bundled deals.
• ▶ 7:42 Apple expressed interest in acquiring Leap Motion, but founder David rejected the offer.
• ▶ 7:49 David criticized Apple as “no longer innovative” and praised Android, causing the deal to collapse.
• ▶ 8:00 In an alternative history, Apple instead acquired other companies like PrimeSense, whose tech later powered the Vision Pro—unlike what might have happened with Leap Motion.
• ▶ 8:21 Leap Motion sent about 12,000 devices to developers to spur hand-tracking apps, marking a major developer-outreach push.
• ▶ 8:28 The company pivoted from desktop input to VR, but integrating the tech into existing headsets was unclear; the Oculus add-on kit added complexity to an already hard-to-understand device.
• ▶ 8:46 After Facebook acquired Oculus in 2014, the VR industry cooled down, weakening the market for Leap Motion’s VR-focused direction.
• ▶ 8:53 Apple made a second acquisition attempt in 2018, offering only $30–50 million—a fraction of Leap Motion's $306 million Series B valuation.
• ▶ 9:17 The offer would have given investors cents on the dollar, leaving founders and employees with almost nothing.
• ▶ 9:22 The deal nearly closed, with Apple sending offer letters to employees, but it ultimately fell through.

• ▶ 9:37 Leap Motion open-sourced its AR headset North Star—a smart, collaborative strategy aimed at tackling core VR/AR hardware challenges, but the product ultimately lacked market fit.

• ▶ 9:51 VR/AR headsets are extremely difficult because they require seamless software-hardware integration, real-time responsiveness, comfort, and a compelling app ecosystem—illustrated by how costly and prolonged Apple's Vision Pro development was.

• ▶ 10:24 Despite having genuinely novel technology and never being a "bad company," Leap Motion and North Star never took off because product-market fit was simply never achieved.

• ▶ 10:39 Leap Motion was caught unprepared for a sudden competitive or market-driven "attack," leaving them strategically vulnerable.
• ▶ 10:43 The North Star VR headset's technology "just wasn't quite mature enough" and could only be built inside a massive tech company like Facebook, not a standalone firm.
• ▶ 10:49 The central lesson: if the market isn't ready for your product, no matter how good it is, success is out of reach—timing and ecosystem preparedness are essential.
• ▶ 10:51 Leap Motion was acquired by UltraHaptics in May 2019 after financial trouble.
• ▶ 10:56 The $30 million deal was a dramatic decline from peak valuation, likely wiping out employee stock holders.
• ▶ 11:04 Strategically, the acquisition made sense: UltraHaptics had worked with Leap Motion for ~6 years and combines hand tracking with haptic feedback for VR.
• ▶ 11:19 Leap Motion is now rebranded as UltraLeap, but the long-term direction of its hand-tracking technology is unclear.
• ▶ 11:28 The tech is still growing and has found niche real-world adoption, such as gesture control in cars (e.g., adjusting volume by hand movement).
• ▶ 11:33 Widespread hand-based computer interaction has not yet arrived—most people still don't use hand control in daily computing.
• ▶ 11:35 David led Leap Motion for nearly a decade, building new technology, raising money, and growing the company into a significant venture.
• ▶ 11:40 Leap Motion reached a notable milestone when Apple attempted to acquire it twice.
• ▶ 11:45 After that chapter, David set up a studio for his next adventure, deliberately waiting for the right opportunity—and he didn't have to wait long.
• [12:09–12:16] In June 2020, a Berkeley paper on diffusion models caught David Holz's attention, and he saw the technology would boost AI image quality by an order of magnitude—the opportunity he'd been waiting for.
• [13:19–13:35] David was testing early Midjourney models by January 2022 and opened the Discord publicly a few months later, enabling extensive user testing and key trade-off decisions like balancing generation time against image quality.
• [14:29–15:13] Midjourney hit a turning point when an image won the Colorado State Fair contest, sparking intense backlash from artists who saw it as plagiarizing real artistry—a debate that foreshadowed broader concerns about AI and misinformation.
• ▶ 15:55 The hardest part of building an AI company is the intense public scrutiny, with Sam Altman’s congressional testimony cited and many people believing AI programs like ChatGPT should be illegal.
• ▶ 16:04 Despite this hostile environment, David Holz has built Midjourney into the best AI image generator available, setting up the question of how he did it.
• ▶ 16:12 Midjourney’s first growth strategy was embedding the product inside Discord, with server size as the key growth metric rather than revenue.
• ▶ 16:37 David leveraged Discord for virality through volume—putting millions of users together creates a feedback loop where more images attract more users.
• ▶ 18:14 The Discord community also trains the model: users select the best of four generated images, giving direct feedback that improves Midjourney.
• ▶ 18:35 Midjourney prioritizes product-market fit, using real user learning and feedback to shape the product.
• ▶ 18:41 David's core philosophy is a bias for action: “thinking overwhelms doing” is a trap, so make many probably-correct moves quickly rather than waiting for perfection.
• ▶ 19:12 Because generative AI moves extremely fast (Midjourney is already at V5), the winning strategy is to ship a minimum viable product and iterate rapidly with a large, engaged user base.
• ▶ 19:38 Many AI companies chase platform/infrastructure status instead of building a great consumer product.
• ▶ 19:58 Building a platform without a compelling product fails—Leap Motion’s hand-tracking never took off.
• ▶ 20:12 Midjourney succeeds by prioritizing an incredibly easy, fun product aimed at everyone online, achieving strong product-market fit.
• ▶ 20:35 David deliberately bootstrapped Midjourney rather than raising VC money, identifying compute—not cash—as the scarce resource needed to scale.
• ▶ 20:56 Midjourney secured GPU capacity through his network and used global load balancing, routing work to idle GPUs in nighttime regions to keep rendering fast for users.
• ▶ 21:34 The industry is shifting toward valuing compute over capital, exemplified by the Andromeda cluster built for startups that need processing power more than funding.
• ▶ 22:02 The section frames learning as a core value for Midjourney, directly linking the company's improvement to learning from past experiences.
• ▶ 22:07 David’s background—nearly a decade at Leap Motion, years in academia, and active participation in the AI community—is highlighted as a key source of this learning mindset.
• ▶ 22:21 These accumulated experiences are credited with making him a better founder, CEO, and engineer, tying Midjourney’s success to a culture of iterative learning.
• ▶ 22:28 David hosts four-hour Discord sessions with thousands of users, constantly iterating on Midjourney while pondering AI's broader role in society.
• ▶ 22:48 He compares AI to water: dangerous yet civilization-building, with no will or spite—an opportunity that shouldn't be banned despite the risks.
• ▶ 23:20 It took him 20 years from college to Leap Motion, but only ~1 year to grow Midjourney into the leading AI image tool, generating millions monthly with zero venture capital.

• ▶ 0:00 Introduction and David Holz’s Early Life (0:00 - 3:16) - AI images and Midjourney lead into David Holz’s childhood, education, research, and early restlessness.
• ▶ 3:16 Leap Motion: Vision, Setbacks, and Acquisition (3:16 - 11:56) - Leap Motion’s founding, user-education focus, funding, Apple acquisition attempts, VR pivot, failure, and sale to Ultra Haptics.
• ▶ 11:56 From Diffusion to Midjourney’s Launch (11:56 - 15:55) - The diffusion paper, DALL·E’s turning point, Midjourney’s open beta, Colorado State Fair win, and the artist backlash.
• ▶ 15:55 AI Scrutiny and Midjourney’s Four Strategies (15:55 - 22:26) - Managing public scrutiny with Discord growth, product-market fit, bias to action, product focus, need, and learning.
• ▶ 22:26 David’s Ongoing Learning and Water Analogy (22:26 - 23:55) - David’s continued community engagement and his water analogy for AI.