🤖 How did Midjourney start
Inside Midjourney's Growth: Key Early Moves
Hi, I’m Anthony! 👋 Every Week, I look at the 10% of startups that did not end up in the startup graveyard. Most analyses are done on mature companies which is not actionable for budding entrepreneurs. I want to understand what the founders did at the very beginning. What did they do before their startup became a household name?
If you haven’t subscribed yet, join the other folks interested in starting your own company by subscribing here:
Actionable Insights
I think that Midjourney is still early in its founding and it might be premature to call them a success. But, I was curious how a bootstrapped AI company managed to lead the pack, hence this article!
Here are key takeaways I have learned from the decisions that Midjourney has made at the beginning of its journey.
💡 Focus on Core Resources Over Capital: Recognize what your key resource is. Holz recognized early on that the key resource for Midjourney wasn't money but computing power. This strategic focus demonstrates the importance of identifying and leveraging the most crucial resources for your business, beyond just financial capital.
🌐 Leveraging Community and User Feedback: Midjourney's integration with Discord showcases the power of community engagement and rapid user feedback in refining a product. This approach led to a vibrant, self-growing user base and a robust product-market fit. Holz's strategy underscores the value of building a product in close collaboration with its user community, utilizing their feedback for continuous improvement.
🤖 Harness AI for Disruption: Midjourney demonstrates the potential of AI to create groundbreaking solutions and disrupt markets. Embrace AI to innovate and differentiate your startup in competitive landscapes.Background of Midjourney
Midjourney is one of the leading image-generative artificial intelligence companies in the world. It calls itself an "An independent research lab exploring new mediums of thought and expanding the imaginative powers of the human species".
Within a mere two years since its inception, Midjourney has not only revolutionized the field of AI but also captured the public's imagination. This is evidenced by its prominent mentions, like John Green's engaging cat Tweet and a feature in a John Oliver segment, highlighting its widespread cultural impact.
As we step into 2024, a year anticipated to further elevate AI's prominence, let's start by spotlighting Midjourney and its remarkable journey in transforming both technology and imagination.
David Holz: The Genius Founder?
David Holz is the founder of Midjourney and he is a smart guy with experience in launching startups.
His journey began in Florida, where he ran a design business during his high school years, simultaneously delving into math and physics. Holz's academic pursuits led him to earn a Ph.D. in Applied Mathematics from the University of North Carolina. He further honed his research skills at the Max Planck Institute and NASA Langley Research Center, focusing on LiDAR, Mars missions, and atmospheric science.
Transitioning from academia, Holz founded Leap Motion, a venture capital-backed company pioneering in AR/VR technology. Leap Motion developed innovative hardware for hand motion tracking and device interaction, securing over $100 million in funding. Despite its initial promise, Leap Motion faced challenges in achieving a market fit and was eventually sold for $30 million.
After a 12-year tenure at Leap Motion, the demands of working in a high-stakes venture-backed environment weighed heavily on Holz, leading him to seek a new direction. It was during this period of reflection that he encountered a pivotal research paper from Cornell University on 'Denoising Diffusion Probabilistic Models'. With his deep technical understanding, Holz realized the potential of these models to revolutionize AI-generated imagery. This insight planted the seed for what would become Midjourney, marking the start of an exciting new chapter in his entrepreneurial journey.
Core Team
David Holz embarked on his new venture, Midjourney, in August 2021, after his tenure at Leap Motion. Starting with a core team of 10 skilled engineers, the company has maintained a lean team structure despite its rapid growth. Midjourney's hiring strategy emphasizes assembling a team with robust technical skills, and strong work ethics, and just trusting them to get on with it.
The effectiveness of Midjourney's culture is evident in its impressive financial performance. Boasting a staggering $5 million revenue per employee, the company outshines established SaaS giants like Dropbox, Slack, Squarespace, and Palantir. This remarkable figure not only underscores Midjourney's efficiency but also highlights its profitability, powered by a capable and dedicated team.
Midjourney's streamlined team size also allows for a flat organizational hierarchy. With minimal managerial layers, no board of directors, and small, autonomous teams, the company fosters an environment of independence and agility.
Additionally, Midjourney's unique incentive model offers employees profit-sharing instead of traditional stock options, aligning their interests directly with the company's success. This approach not only motivates the staff but also provides immediate financial benefits, setting Midjourney apart in both its operational structure and its approach to employee rewards.
Birth of Midjourney
With the team in place, the private demo of Midjourney was launched in September 2021. A few months later in Feb 2022, it was released to the public via Discord.
Midjourney's unique integration with Discord positions it in a dynamic social setting.
New users access Midjourney through Discord's server, where they can use text prompts, similar to ChatGPT, to generate images. However, Midjourney's platform adds a communal dimension. Users interact in various newbie channels, witnessing and drawing inspiration from each other's prompts and the resulting images. This collaborative environment fosters a sense of community and creativity.
Adding to its user-friendly interface, Midjourney introduces the ability for users to enhance their prompts with their images or a combination of multiple uploads. The AI processes these inputs, and within about a minute, presents four generated images for all the users in that channel. The user will then select the best image from these options.
Customer Acquisition
From the onset, David Holz identified Midjourney's primary user groups:
Advertisers: They use Midjourney to generate customizable images, offering an efficient alternative to traditional stock photos or hiring photographers.
Artists: These users leverage Midjourney for exploratory and conceptual work, rapidly visualizing ideas for clients before commencing actual projects.
To attract these key users, Midjourney recognized the need for virality, which is driven by a high volume of image generation.
Power of Discord
For Midjourney to quickly reach that stage, Discord helps in 3 ways:
Community-Ready Platform:
Discord, popular among gamers, content creators, and streamers, aligns with Midjourney's target audience. This synergy allows Midjourney to harness the social nature of Discord, encouraging users to share their creations and organically grow the user base. As a result, Midjourney's marketing expenses are significantly reduced.
Given the controversial nature of AI-generated images, creating a place where all users of Midjourney gather also makes it easier when decisions have to be made.
Generating Volume:
The vibrant community on Discord actively creates and shares images. This visibility encourages more people to join, creating a strong viral loop.
In just a year, Discord's user base surged from two million to an impressive 14.8 million, reaching 18 million by 2023.
Source: Photutorial
Rapid User Feedback:
For an AI-centric product like Midjourney, user data, and feedback are vital for refining the model. On Discord, Midjourney engages users to choose from four generated images, aiding the AI in learning and enhancing future outputs. This quick feedback mechanism is instrumental in evolving the AI's capabilities.
Product Market Fit
At the heart of Midjourney's success is a decisive bias for action, a trait embodied by David Holz and his team. This proactive approach is complemented by the robust feedback loop from Discord, enabling continuous enhancement of Midjourney's offerings. While each release may not be flawless, this iterative process has rapidly evolved Midjourney to V5, with V6 already on the horizon, all within just 1.5 years.
Midjourney's strength lies in its vibrant community, which consistently generates high-quality images. This, coupled with the platform's ongoing improvements, has solidified Midjourney as a robust product in the market.
For new users, the journey to recognizing Midjourney's value is remarkably short. They simply have to:
Be a Discord member
Type in a prompt (or copy others in the group)
Get 4 amazing images
This ease of use and immediate gratification lead to quick 'Aha!' moments for users.
User-driven growth is central to Midjourney's strategy. The platform's users, predominantly artists, and advertisers, are the main drivers of its expansion. This organic, product-led growth model allows Midjourney to minimize marketing expenditures while effectively reaching and retaining its target audience.
This synergy of a strong product, engaged community, and efficient market penetration showed Midjourney's clear product-market fit.
The Business Model
"It's a pretty simple business model, which is, do people enjoy using it? Then if they do, they have to pay the cost of using it because the raw cost is actually quite expensive. And then we add a percentage on top of that, which is hopefully enough to feed and house us. And so that's what we're doing."
—David Holz, Founder Midjourney
Revenue
Midjourney's revenue model is primarily subscription-based, with plans ranging from $10 to $120 per month. Additionally, the company offers the option to rent extra GPU time, allowing users to extend their sessions seamlessly.
The higher tier plans allow users a private and faster experience. This contrasts with the public channel's bustling activity, where interactions from numerous users flood the screen. As of November 2023, Midjourney reported an impressive Annual Recurring Revenue (ARR) of $200 million.
Expenses
Their expenses are likely to fall into 3 categories:
Data Collection: This involves gathering the necessary data to feed and train the AI models.
Data Cleaning and Training: These are critical steps for ensuring the quality and effectiveness of the AI, likely representing significant fixed costs.
Active Server Costs: These costs are variable, scaling directly with Midjourney's usage. They are tied to the number of prompts processed.
Among these expenses, the data collection cleaning and training of the model are likely to be fixed costs. In contrast, the active server costs fluctuate based on user activity. For Midjourney, and AI companies in general, the ability to secure GPUs and manage compute costs is crucial.
Efficient resource management directly impacts profitability and sustainability in the competitive AI landscape. Keeping server or compute costs low while handling increasing user demands is a key challenge that Midjourney must navigate to maintain its growth trajectory and market position.
Funding
Midjourney is completely bootstrapped with no outside money.
"We're self-funded. We have no investors. We're not really financially motivated. We're just sort of here to work on things we're passionate about and have fun. And we were working on a lot of different projects."
—David Holz
To be fair, such words can only be spoken by an entrepreneur that has money to spare and hence is "not financially motivated". But there are upsides for not raising a single cent from VCs:
Holz retains the power in decision-making
There are many unresolved issues with AI companies. Copyrights of the images used in the training for the bot, controlling the outputs of the bots, etc, the list goes on. From the view of the company, having a single voice with no outside influence might be critical for survival. No potential for a coup when opinions do not align (👀 OpenAI).
Unreliable VC capital
David Holz has voiced concerns regarding the sustainability of companies excessively dependent on venture capital. He points out the problematic cycle of burning through funds and perpetually seeking more investments, a scenario Midjourney has consciously steered clear of.
Additionally, Holz's skepticism extends to the reliability of venture capitalists. He observes that they invest in a hype cycle and exit once the bubble bursts. This perspective is likely colored by his own challenging experiences at Leap Motion, his previous startup. Despite raising over $100 million in venture capital, Leap Motion faced a disappointing outcome, selling for only $30 million in 2019.
Key resource
Holz understood from the beginning that what was needed wasn't money but computing power. By securing that, he can keep expenses low and also improve user experience.
From the beginning, he leveraged his network and did many things to secure and lock down access to GPUs. Today, Midjourney actively searches for GPUs in parts of the world where it is nighttime to improve the time taken to generate images for users.
Projects such as the Andromeda Cluster founded by Daniel Gross and Nat Friedman which provides computing power to startups show an interesting shift from providing capital to compute.
Summing up what Midjourney is all about with this quote:
"No funding. No management layers. No B2B. Just discord and the excited public!”