Roko's Basilisk
Posts
Stability’s Open-Source Trap

Stability’s Open-Source Trap

Plus: NVIDIA's uncanny DLSS, Altman’s coder warning, and Mistral Forge.

Roko's Basilisk
March 20, 2026

Here’s what’s on our plate today:

🧪 Stability AI, open models, and the brutal math of monetizing “free.”
🧠 Headlines: NVIDIA’s uncanny DLSS, Altman’s coder warning, Mistral Forge.
🧰 Weekend To-Do: Stability AI, Hugging Face Spaces, and ComfyUI.
🗳️ Friday poll on what actually works as an open-source AI business model.

Let’s dive in. No floaties needed…

In partnership with

Here’s how I use Attio to run my day.

Attio is the AI CRM with conversational AI built directly into your workspace. Every morning, Ask Attio handles my prep:

Surfaces insights from calls and conversations across my entire CRM
Update records and create tasks without manual entry
Answers questions about deals, accounts, and customer signals that used to take hours to find

All in seconds. No searching, no switching tabs, no manual updates.

Ready to scale faster?

Start now →

_{*This is sponsored content}

The Laboratory

How Stability AI Changed The Face Of Generative Open-Source AI

TL;DR

Open source ignited the image AI boom: Stable Diffusion launched in August 2022 as a fully open model anyone could download, modify, and build on. It racked up over 300M downloads and dominated AI image generation, but the company behind it was burning $8M/month while pulling in under $5M per quarter.
Free models, brutal economics: Unlike traditional open-source software, where companies monetize support and deployment, open-source AI hands over the entire product when it releases model weights.
Big Tech made it worse: Meta’s Llama family and similar releases from trillion-dollar companies turned open-weight models into a strategic commodity play, not a revenue line.
The enterprise pivot is the real test: New CEO Prem Akkaraju cleared $100M+ in debt, brought in James Cameron and Sean Parker, and reoriented toward paid creative tools, enterprise licensing, and partnerships with EA, Universal Music, and WPP.
The stakes are industry-wide: If Stability AI’s licensing and enterprise model works, it validates a path for independent open-source AI companies. If it doesn’t, the market structure consolidates around a handful of cloud and platform giants who can afford to give models away forever.

Stability AI helped spark the generative AI boom by releasing a free model anyone could use, but the same openness that drove massive adoption also exposed the harsh economics of building a business around open-source AI. Photo Credit: The Information.

Among the many uses of artificial intelligence, generative AI is considered one of the biggest technological shifts since smartphones and cloud computing, as it enables computers to act creatively rather than just logically. Gen AI lets models create new content such as text, images, video, audio, or code by learning patterns from large datasets.

Unlike traditional AI, which mainly analyzes or predicts, generative AI produces original outputs using deep learning models such as transformers and diffusion networks. And with tools such as ChatGPT, Stable Diffusion, and Midjourney, the technology has become mainstream, enabling individuals and companies to generate content with simple prompts and accelerating its adoption across media, software, design, research, and business.

While it differs in how the technology works, it faces the same problems that frontier AI labs do when it comes to monetizing the technology and building sustainable business models around it. To understand the challenges facing Gen AI, one company stands out for democratizing the technology in its early phases and for continuing to push the boundaries of how open-source Gen AI models are monetized.

Breaking open the walled garden

Before Stability AI, generative AI mostly existed inside closed ecosystems. OpenAI had DALL-E, Google had Imagen, and although both were powerful, access was limited. In 2022, generating images from text often meant joining waitlists, using tightly controlled APIs, and working within rules set by a few large companies.

All this changed in August 2022 when a London-based startup released Stable Diffusion. The model was openly available for anyone to download, run on their own hardware, modify, fine-tune, and build products on top of, all without gatekeepers, strict API limits, or special permission.

Massive adoption, minimal revenue

Within weeks, developers began building open-source interfaces, artists began experimenting at an unprecedented scale, and communities released thousands of fine-tuned versions of the model. New commercial tools ranging from marketing software to game asset generators quickly appeared. Within a year of the model’s launch, Stable Diffusion accounted for a large majority of AI-generated images online, with total downloads eventually exceeding 300M.

However, despite the immense success of Stability AI, the company behind Stable Diffusion, was generating under $5M in quarterly revenue while burning roughly $8M per month.

The company owed nearly $100M to creditors, and the researchers who built Stable Diffusion had started leaving the company. Key investors also resigned from the board, and by March 2024, founder Emad Mostaque had stepped down as CEO under pressure, telling the world on X that “the concentration of power in AI is bad for us all.”

The open-source AI business problem

The tough times for Stability AI were not due to a lack of technology or user adoption, but to the difficult economics of managing an open-source AI business. In traditional open-source software, companies can give away code while still earning revenue from enterprise support, security, and deployment services. In open-source AI, however, that gap is far smaller. Once model weights are released, developers can fine-tune and deploy them independently with minimal vendor support. The model itself is the product, and it can be copied freely, even though the original company paid the high cost of training it.

As a result, much of the commercial value flows to the community rather than to the creator, a structural challenge that Stability AI brought to light.

Competing against Big Tech’s free models

Adding to the burden of monetizing open-sourced AI models was the increasingly competitive gen AI landscape, which now included companies like Meta releasing its Llama family of open-weight models, including increasingly capable image and multimodal capabilities, for free.

It is important to note here that for Meta, the motivations for sharing open-source models are fundamentally different from those of Stability AI.

Meta does not need Llama to generate revenue directly. It is a strategic distribution play: by commoditizing the model layer, Meta drives developer adoption of its ecosystem, strengthens its advertising infrastructure, and denies rivals a competitive advantage. With $70B+ in annual AI infrastructure spending and nearly $20B in quarterly profit from its core apps business, Meta can fund open-weight AI development indefinitely without ever needing to sell a model license.

On the other hand, for Stability, the model itself is the product. But it has to compete with companies that treat similar models as side projects funded by much larger businesses, such as advertising or cloud services.

This reality put pressure not just on Stability AI but on the entire open-source AI ecosystem. If trillion-dollar tech companies can release powerful models for free as a strategy rather than a business, it raises a difficult question: how can an independent company charge for the same technology?

New leadership, new playbook

Stability AI’s response to the crisis has been a major repositioning rather than small changes. After Prem Akkaraju, former CEO of Weta Digital, took over in June 2024, the company raised about $80M in new funding and had more than $100M in supplier debt forgiven.

The leadership team expanded with high-profile names, including Napster co-founder Sean Parker as Executive Chairman, filmmaker James Cameron on the board, and Oscar-winning visual effects veteran Robert Legato in a senior technical role. At the same time, the company shifted its focus to enterprise customers rather than the broader open-source community.

Licensing, partnerships, and paid tiers

The product strategy moved toward commercial creative tools and in 2025, the company introduced Stability AI Solutions, a set of products aimed at marketing, advertising, and design workflows, followed by partnerships with companies such as Electronic Arts for game asset creation, Universal Music Group and Warner Music Group for licensed AI music tools, and WPP to integrate Stability AI models into large scale advertising production systems.

Alongside this, the company introduced a licensing model that allows free use for smaller developers while requiring paid enterprise licenses for larger organizations, an approach meant to keep the ecosystem open while still generating revenue from commercial users.

By late 2024, Akkaraju said the company had cleared its debt and was growing quickly, although the figures are not public. What is clear is that Stability AI is no longer trying to survive solely as an open-source model provider and is instead attempting to become an enterprise technology company selling production-grade creative AI tools, with its future depending on whether the reputation built through open access can translate into sustainable business at scale.

What Stability AI’s fate means for the industry

What makes Stability AI’s story special is that it is not just about one startup, but about a cycle the entire AI industry is now going through. Every company building open-source AI models faces the same tension: how do you run a business when the core technology is free, when large platform companies can give away competing models as part of bigger businesses, and when AI does not require the kind of complex deployment support that once helped open-source software companies make money?

Stability AI’s efforts to solve this through enterprise licensing, paid tiers, partnerships, and production tools are being closely watched because they may reveal whether open-source companies can survive on their own or whether a few large cloud and platform players will eventually dominate the industry.

This question matters to customers and investors as much as to developers because when companies choose AI tools, they are comparing features and the financial stability of the vendor behind them. Meanwhile, investors are also becoming more cautious, shifting money toward companies that can show a clear path from usage to profit, as adoption alone is no longer enough to justify funding.

With this backdrop, Stability AI’s biggest impact may be that it exposed the central contradiction of open-source generative AI. By making powerful models freely available, it helped define the modern AI boom. Still, it also created a market where the company that opened the technology to everyone struggles to capture the value it created.

Now, through its licensing experiments, enterprise deals, and efforts to become financially sustainable, Stability AI is forcing the industry to confront a problem it has long avoided. The question of how to build a lasting business around free models remains unresolved, and the answer will shape the future structure of the generative AI market, no matter what happens to Stability AI itself.

Framer for Start Ups

First impressions matter. With Framer, early-stage founders can launch a beautiful, production-ready site in hours. No dev team, no hassle. Join hundreds of YC-backed startups that launched here and never looked back.

Key value props:

One year free: Save $360 with a full year of Framer Pro, free for early-stage startups.
No code, no delays: Launch a polished site in hours, not weeks, without hiring developers.
Built to grow: Scale your site from MVP to full product with CMS, analytics, and AI localization.
Join YC-backed founders: Hundreds of top startups are already building on Framer.

Eligibility: Pre-seed and seed-stage startups, new to Framer.