Even Snap Can't Afford AI Video: What the Dotmo Spin-Off Reveals About Generative Video in 2026

What Happened: Snap Is Spinning Out an AI Video Team Called Dotmo

Snapchat's parent company, Snap, is doing something unusual: it's spinning an internal generative AI video team out into a brand-new standalone company. The venture is called Dotmo, and according to TechCrunch, it will focus on developing AI models that can create interactive gaming and entertainment experiences.

The details are telling. Dotmo won't be funded directly by Snap, but Snap's cofounder and CTO, Bobby Murphy, will serve as lead investor and take a significant personal stake — while continuing to work full-time at Snap and lead its GenAI research. In exchange for handing over both the talent and a technology license, Snap receives a large equity stake in Dotmo, keeping exposure to any upside if the new company prospers. The initial Dotmo team is made up of current Snap staff who are leaving to launch it, and Dotmo may eventually raise outside funding.

It's a clever deal for Snap. But the most important word in the entire announcement was the one Snap used to explain why it was doing this at all: "due to costs."

"Due to Costs" Is the Tell

Spin-offs happen for many reasons — to showcase an asset, attract investors, or give a team more freedom. But Snap explicitly cited the high costs of conducting this work internally as a driving factor. When a company worth tens of billions decides it would rather spin a promising AI capability into a separate entity than keep paying the compute bill itself, that's a loud signal about where generative video economics actually sit.

This is the same Snap that, earlier in 2026, cut roughly 1,000 jobs and spun off its Specs smart-glasses line into a separate company (an unveiling that went poorly — Snap's stock tanked after the ~$2,200 price tag drew criticism). Dotmo is its second major spin-off of the year. Read together, the pattern is clear: Snap is aggressively shedding expensive, speculative bets so it can keep its core business lean — and the costliest of those bets is generative AI.

Why Generative AI Video Is Brutally Expensive

If you've only dabbled with text chatbots, it's easy to underestimate what generative video actually costs to run. A few seconds of AI video requires a model to predict, frame by frame, millions of pixels across time — keeping characters consistent, motion coherent, and physics plausible. Compared with generating text or even a still AI image, the compute demands are on a different planet.

That cost shows up everywhere:

• Training. Building a competitive video model means training on enormous datasets with oceans of GPU hours — a fixed cost only deep-pocketed labs can stomach.
• Inference. Every clip a user generates burns serious compute. As we explored in our piece on the AI inference gold rush, the meter is always running — and for video it runs far faster than for text.
• Scale. Generous free tiers at consumer volume can crater margins overnight, which is why so many video tools quietly slash credits or raise prices.

In other words, Snap didn't spin out Dotmo because the technology wasn't working. It spun it out because the technology works — and the bill to keep it running at scale is staggering.

The Wider Pattern: Big Tech Is Offloading AI Risk

Dotmo isn't an isolated story. Across 2026, the companies that can afford frontier AI have been busy not absorbing all of its risk. The smarter structural move is to let specialized startups burn the capital, chase the breakthroughs, and bear the margin pain — while the incumbents retain equity, licensing rights, and an option to re-acquire the winners later. Snap's "we license the tech, fund the team via our CTO, and keep a big equity stake" arrangement is textbook risk-transfer.

The implication for the broader market is that pure-play generative video remains a capital-intensive, low-margin grind. The survivors will be a small number of extremely well-funded players (think the Sora-class efforts inside major labs and a handful of well-backed independents), a layer of open-source options, and a long tail of smaller tools that live or die on their funding runway.

What This Means for Anyone Choosing AI Video Tools

You don't need to be Snap to feel the consequences. If you're picking AI video tools for marketing, content, or product, the Dotmo story is a reminder to shop defensively:

• Prices and credits will keep moving. Because the underlying compute is so costly, expect volatility. A tool that's cheap today may slash its free tier tomorrow. Budget for the real per-clip cost, not the launch promo.
• Survival matters. A video tool's roadmap is only as durable as its funding. Favor vendors with clear revenue, strong backing, or an open-source fallback you can self-host.
• The model layer is consolidating. As the frontier concentrates, the smartest bet is often a tool that's model-agnostic — one that can route to whichever engine is best and cheapest at any moment, rather than welding you to a single provider that could vanish or get spun off.

How to Pick an AI Video Tool That Won't Disappear

For teams building a video workflow that needs to last more than a single funding cycle, here's a pragmatic playbook:

Frequently Asked Questions