An impressive new Gemini ‘Omni’ video model just leaked ahead of Google I/O

Google’s Gemini Omni video-generation model appears to be surfacing for some users, with leaked demos showing highly realistic outputs for prompts like two men eating spaghetti and a professor explaining a trigonometric proof. The new model may be an extension of Veo and includes video remixing, chat-based editing, and templates, but it is also resource-intensive, with one user reportedly using 86% of their daily AI Pro allowance after two prompts. The article suggests Google is preparing broader usage limits ahead of I/O 2026.

Analysis

This looks less like a product gimmick and more like a monetization lever for Google’s AI stack. If Gemini can materially improve video workflows inside a chat-first interface, it increases the odds that AI usage shifts from novelty to recurring, high-intensity consumption, which is where Google can justify premium tiers, tighter quotas, and eventually enterprise upsell. The second-order effect is that every successful video interaction expands the economic moat around Google’s multimodal model family, because video is a much stickier use case than text-only prompting.

The market should care less about the demo quality than about inference intensity. Video generation is compute-expensive enough that unit economics matter immediately: if a small number of prompts can consume most of a daily allowance, Google is signaling that this feature will be rationed, not broadly commoditized. That creates a favorable scarcity dynamic for Google’s pricing power, but it also highlights capex intensity across the AI supply chain, especially for accelerators, memory, and networking if rollout scales faster than expected.

The biggest competitive implication is pressure on AI video startups and adjacent creative software vendors. If Gemini Omni becomes a native workflow inside an existing consumer product, standalone tools face distribution disadvantage even if their model quality is comparable. The overhang is execution risk: if latency, quality consistency, or quota constraints make the feature feel fragile, engagement could fall back to experimentation only, which would limit near-term revenue impact and keep this as an optionality story rather than a fundamental driver.

The contrarian view is that the stock may not move much on the announcement itself because investors already assume Google will keep layering multimodal features into Gemini. The underappreciated catalyst is not model launch, but usage metering and packaging changes over the next 1-2 quarters; that is where revenue per user can step up. If rollout is broad and the feature is embedded into paid tiers, this becomes a quiet ARPU tailwind rather than a headline AI event.

Key Decisions for Investors

• Maintain/trim a tactical long GOOGL into any weakness over the next 1-3 weeks; upside comes from pricing power and ARPU expansion if video is bundled into paid tiers, but position size should stay moderate because the market may view this as incremental rather than transformative.
• Add a relative-value long GOOGL / short CRM or ADBE pair over 1-3 months to express the view that native AI video inside a platform ecosystem can pressure standalone creative software vendors on seat growth and pricing.
• For event risk into Google I/O 2026, buy medium-dated GOOGL call spreads 3-6 months out; payoff is strongest if the company pairs the launch with explicit usage limits and monetization details, which would re-rate the AI revenue narrative.
• Watch AI infrastructure names on any confirmation of broader rollout; if Omni drives materially higher inference load, a tactical long on NVDA or AVGO over 1-2 quarters has better torque than chasing GOOGL alone, since capex intensity is the second-order winner.
• Avoid shorting pure-play AI video names immediately; the feature may validate demand faster than it destroys share, so the cleaner trade is to wait for paid-tier packaging evidence before leaning into downside.