Here’s what’s on our plate today:

• 🧪 Runway's path from AI video to enterprise software.

• 🗞️ Open source hijack, data center ban, Samsung's AI windfall.

• 🗳️ Poll on where Runway's Characters technology actually ends up.

• 🧰 Three tools worth trying in the AI video space right now.

Let’s dive in. No floaties needed…

The Laboratory

TL;DR

• Sora flopped; the market didn't: OpenAI's video app peaked at 1M users, then cratered to below 500k, burning $1M/day against $2.1M in lifetime revenue. The competition had already moved on.

• Runway's real play is beyond video clips: The March 2026 Characters launch merges video generation with real-time conversational AI avatars, positioning Runway as a potential interface layer for enterprise software, not just a creative tool.

• The revenue gap is the elephant in the room: Runway targeted $300M in annualized revenue for 2025, but likely hit $90M by mid-year. Characters are the clearest path to enterprise scale, but converting a strong API into sticky contracts is a challenge entirely different.

• Legal exposure is compounding: Active copyright suits land in a year when the $1.5B Anthropic settlement reset the bar for statutory damages, and Characters amplify impersonation risks that go well beyond the deepfake debate.

• The $47.8B question: The AI video market is projected to hit $47.8B by 2034, with digital avatars potentially reaching $270B by 2030. Runway has $860M in funding and sits at the intersection of both, but whether it captures that value or becomes infrastructure that others build on top of is still wide open.

How Runway emerged as a key player in the AI video race

Over the past couple of years, enterprises have been racing to harness AI models for commercial applications. The reality, however, has been far more complex than early expectations suggested. The challenge is not just technological, but financial and structural. Building and deploying AI systems at scale is expensive, and the market itself has become increasingly difficult to navigate as hyperscalers use their capital and distribution advantages to move aggressively into adjacent categories.

The pattern has already played out in areas like coding assistants. Startups such as GitHub and Tabnine helped define the space early on, only to face intensified competition as Microsoft embedded Copilot deeply into its software stack and Google rolled out its own AI coding features across developer tools. What began as a distinct startup category quickly became a feature set absorbed into larger platforms.

However, something different is happening in the market for AI video generation tools.

Despite the presence of major players, companies like Runway have managed to carve out meaningful positions, not just by building models, but by shaping how the category itself evolves. Instead of being immediately subsumed into hyperscaler ecosystems, the space remains more open, with independent players still defining workflows, use cases, and even the underlying creative paradigm.

What makes this space even more interesting is that even hyperscalers like OpenAI shut down its dedicated AI video app, launched less than one year earlier, not because of technical issues but because of economic ones.

According to the Wall Street Journal, despite early hype, Sora peaked at around 1M users before dropping below 500K, while burning roughly $1M a day in compute against just $2.1M in lifetime revenue. As Anthropic gains ground in more lucrative enterprise markets, CEO Sam Altman has chosen to reallocate resources.

A market that kept moving without the biggest name

One reason Sora failed to make an impact was that, by the time it launched its standalone app in September 2025, the competitive landscape had already hardened. Runway had shipped Gen-4.5, which topped both Google and OpenAI on the Video Arena leaderboard. Kling AI, developed by Chinese short-video giant Kuaishou, has built a substantial user base on lower pricing and strong physics simulation. Google’s Veo line, now at version 3 with native audio generation, carried distribution advantages through YouTube and Google Workspace that no startup could replicate. Pika, Luma Dream Machine, and a growing roster of open-source alternatives rounded out a market that post-Sora analysis has segmented into distinct tiers: quality-first (Runway), cost-efficiency (Kling), ecosystem-integration (Google Veo), and open-source (Seedance).

What distinguished these survivors from Sora was not just better products but better business models. Runway and Kling moved to per-second-of-output pricing, which scales naturally with production workflows. Synthesia, the dominant player in the adjacent AI avatar market, built deep enterprise integrations with systems like Workday and SAP, making itself sticky with customers who need compliance and workflow continuity rather than viral demos.

YipitData’s B2B spend analysis found that by early 2026, Runway had become the clear leader in AI video generation among mid-market companies and the only platform in its category monetizing at scale, with most customers adopting it as their sole video generation tool rather than mixing providers.

Sora’s exit removed an underpriced, venture-subsidized competitor from the field. For the companies that remained, that changes the dynamics considerably.

What Runway is actually building

Among those companies, Runway occupies a particularly interesting position, one that extends well beyond filling the gap Sora left. In mid-March 2026, just days before OpenAI’s announcement, Runway launched Characters: a real-time video agent API that generates fully conversational AI avatars from a single reference image.

Characters are built on GWM-1, the General World Model family Runway released in December 2025. GWM-1 is an autoregressive model that generates video frame by frame in real time, a fundamentally different architecture from the diffusion models that power conventional video generation. The system is designed to generate interactive digital entities across three modes: Worlds for explorable environments, Avatars for conversational characters, and Robotics for physical interaction. While these models currently exist separately, the company aims to unify them into a single system.

What makes this significant is how it blurs a line the industry has largely kept intact. Video generation has traditionally focused on creating visual content for media, while avatar platforms have focused on communication through digital presenters. Characters bring the two together, combining the visual richness of generated video with the interactivity of conversational AI to create more lifelike, responsive digital characters.

However, despite the success, the challenges ahead for Runway and for the sector are substantial. Characters introduce technology that can generate photorealistic, interactive human likenesses from a single image, raising sharp questions about impersonation, fraud, and manipulation that go beyond the familiar deepfake debate.

Runway itself acknowledged that interactive avatars amplify these risks: unlike a pre-recorded deepfake video, a conversational avatar can respond to questions, adapt to context, and overcome objections in real time, dramatically expanding the potential for social engineering. The company has implemented content and identity restrictions at launch, prohibiting avatars depicting children, public figures without authorization, or protected intellectual property, and has published a responsible development framework alongside the product. But the gap between policy documents and real-world misuse, once the API is in the hands of thousands of developers through the Builders program, remains an open question.

On the legal front, Runway faces active copyright infringement lawsuits, including the Andersen v. Stability AI class action in which artists filed a third amended complaint in February 2026, expanding proposed classes and alleged infringing datasets.

A contested, not empty, market

As enterprises continue to look for ways to generate economic value from AI, it would be a mistake to read Sora’s shutdown as evidence that AI video generation does not work as a business.

In fact, the opposite reading is more accurate. The global AI video generation platform market was valued at $6.2B in 2025 and is projected to reach $47.8B by 2034. The digital avatar market alone could hit $270B by 2030, according to Grand View Research. The demand is not hypothetical.

The absence of a hyperscaler in this space does not signal that the market is unprofitable. It signals that the market is highly contested, and that the economics of video generation reward specialization and discipline rather than sheer scale. OpenAI could not make Sora work because the video was a sideshow for a company whose core business lies elsewhere. For Runway, a company founded in 2018 by three graduates of NYU’s Tisch School of the Arts, video is the entire business.

Beyond their immediate success, companies like Runway are building the foundational models, developer platforms, and ecosystem incentives that will shape how AI video evolves: not just as a creative tool but as an interface layer for enterprise software, customer interaction, and, eventually, immersive digital environments.

Whether Runway itself captures the full value of that vision, or becomes infrastructure that others build on top of, is the kind of question that will take years to resolve. For now, the company has $860M in funding, a product that merges video generation with real-time interactivity, and a market that just lost its most unpredictable competitor. What happens next depends on whether Characters finds genuine enterprise traction, whether the Builders program produces breakout applications, and whether the legal and ethical frameworks around this technology can keep pace with its capabilities. None of those questions has clean answers yet. That is precisely what makes this corner of AI worth watching.

Quick Bits, No Fluff

• North Korea's long con: Hackers spent weeks building fake company profiles and Slack workspaces to earn the trust of Axios's maintainer, then pushed malware into one of the web's most widely used open-source projects.

• Maine says no: The state is set to become the first US state to ban new data center construction, freezing projects that consume 20MW or more until at least November 2027.

• Samsung's AI windfall: The company’s forecast Q1 operating profit of $37.8B, up more than eightfold year over year and nearly three times its previous quarterly record.

3 Things Worth Trying

• Runway: The AI video platform everyone's talking about, now with real-time conversational avatars you can build into products via API.

• Kling AI: Kuaishou's video generator punches above its weight on physics simulation and pricing, a serious alternative if Runway's cost structure isn't working for you.

• Synthesia: The enterprise-grade AI avatar platform with deep integrations into Workday and SAP, built for teams that need compliance and workflow continuity, not just cool demos.

Thursday Poll

The Toolkit

• Leonardo AI: AI image studio for concept art, product visuals, and styled assets at speed.

• Modal: Serverless Python platform to run AI and data workloads without touching infrastructure.

• QuillBot: AI paraphraser that rewrites, tightens, and clarifies text in multiple languages.

Rate This Edition