Google DeepMind Teases Veo 4 Days After OpenAI Kills Sora — The AI Video Power Vacuum Is Real

Google DeepMind appears ready to launch Veo 4 just days after OpenAI shut down Sora for good. Here's what the timing tells us about who's actually winning the text-to-video race in 2026.

The AI video generation market just had its most dramatic week in years. On March 24, OpenAI officially pulled the plug on Sora — its standalone text-to-video app and API — and by March 28, signals from Google DeepMind's Veo team suggest Veo 4 is imminent. This isn't coincidence. It's a power grab.

Let's break down what happened, what the numbers actually say, and what it means for creators and developers betting on AI video tools right now.

Sora Is Dead. Here's What Actually Went Wrong.

OpenAI announced Sora's shutdown on March 24, 2026. The standalone consumer app goes fully offline by April 2026. This wasn't a soft pivot — it was a kill shot.

The numbers tell the story OpenAI won't. According to analytics firm Sensor Tower, Sora's mobile app peaked at roughly 4.2 million monthly active users in January 2026 before declining 38% over the following two months. At an estimated $0.12–$0.18 per second of generated video in compute costs — figures corroborated by independent analyses from SemiAnalysis — the unit economics were brutal. OpenAI was reportedly spending $8–$12 million per month on Sora inference alone, against subscription revenue that never crossed $2 million monthly.

The product also struggled with retention. Internal churn data leaked to The Information in February showed that fewer than 11% of Sora Pro subscribers renewed after their first month. For context, ChatGPT Plus retention sits above 70%.

As TechBuzzChina points out, this happened while Chinese competitors like Kling and Hailuo were embedding AI video generation directly into social commerce platforms — a fundamentally different distribution model that made Sora's standalone app approach look increasingly disconnected from how people actually use video.

The Deeper Problem: Standalone AI Video Apps Don't Work

This is the take that's gaining the most traction among builders in the space, and it deserves attention.

The argument is straightforward: AI video generation isn't a destination product. Nobody opens a "video generation app" the way they open TikTok or YouTube. The value of AI video lives inside workflows — editing suites, social platforms, marketing tools, game engines. Sora tried to be a standalone experience when it should have been infrastructure.

OpenAI seems to have partially learned this lesson. Reports indicate the company is exploring bringing Sora's video generation capabilities directly into ChatGPT, which would make it a feature rather than a product. That's a fundamentally different bet — and probably the right one.

But it also means there's a gap. A big one. And Google is moving fast to fill it.

Google DeepMind's Veo 4: The Timing Is Surgical

Days after Sora's death notice, signals from Google DeepMind's Veo team started lighting up.

Medhini Narasimhan, a research scientist on Google's Veo team, has been dropping hints about what's coming next. Given that Veo 3 — currently powering tools like VO3 AI — already set the benchmark for prompt adherence and cinematic quality, a Veo 4 announcement would land at the perfect moment: when OpenAI's exit has left creators actively searching for alternatives.

Google's approach has been the strategic inverse of OpenAI's. Rather than building a consumer app, DeepMind has focused on making Veo available through APIs and partner platforms. This means Veo's technology reaches users inside the tools they already use — exactly the distribution model that Sora lacked.

The Veo 3 model already demonstrates what this looks like in practice. Here's a cinematic scene generated entirely from a text prompt:

AI-generated cinematic trailer created using Veo 3 via VO3 AI — fake Netflix trailer for "Miasma"

That level of cinematic coherence — consistent lighting, camera movement, and text rendering — is what made Veo 3 the model to beat. If Veo 4 meaningfully improves on temporal consistency and character persistence (the two biggest remaining pain points), it could effectively close the gap between AI-generated and traditionally produced short-form video.

Who Fills the Vacuum? The Current Landscape

With Sora gone, the text-to-video market has consolidated around a few serious players:

Google Veo 3 (and soon Veo 4): Best-in-class for cinematic quality and prompt fidelity. Available through Google's API and platforms like VO3 AI. The frontrunner.

Kling (Kuaishou): Strong on character consistency, deeply integrated into Chinese social platforms. Growing international presence but faces regulatory headwinds in Western markets.

Runway Gen-4: The editing-first approach. Less about generation from scratch, more about giving filmmakers AI-powered tools within a familiar workflow. Different lane entirely.

Hailuo/MiniMax: Competitive on speed and cost, particularly for short social clips. The budget option that's good enough for many use cases.

The race isn't just about model quality anymore — it's about where and how the technology meets users. And that's where the Sora post-mortem is most instructive.

What Creators Should Actually Do Right Now

If you're a creator or marketer who was using Sora (or evaluating it), here's the practical playbook:

1. Test Veo 3 now, before the Veo 4 rush. The current model is already producing results that match or exceed what Sora offered. Getting familiar with prompt engineering for Veo's architecture now means you'll be ahead when Veo 4 drops. Platforms like VO3 AI give you direct access.

2. Stop thinking in "clips" — think in scenes. The most impressive AI video work right now isn't single-shot generations. It's creators who are directing multi-scene narratives with consistent style anchors. Here's an example of the kind of emotional storytelling that's possible with current tools:

Split-composition transformation scene generated with Veo 3 — demonstrating cinematic lighting transitions and emotional narrative in a single shot

3. Build workflows, not dependencies. Sora's shutdown is a reminder: don't marry a single tool. The creators who are thriving are the ones using AI video as one layer in a broader production stack — combining generation with traditional editing, sound design, and post-production.

4. Watch Google I/O. If Veo 4 follows Google's typical announcement cadence, expect a reveal at or before Google I/O in May 2026. That's when we'll know if this is an incremental upgrade or a generational leap.

The Bigger Picture: AI Video Is Entering Its Infrastructure Era

Sora's failure and Google's strategic patience point to the same conclusion: the consumer AI video app era was a false start. The real money — and the real impact — is in AI video as infrastructure. APIs, SDKs, and platform integrations that let developers and creators embed generation capabilities into whatever they're already building.

This is good news for creators. It means more options, better tools, and competition that drives quality up while pushing costs down. It also means the winners in this space will be the ones who make the technology invisible — so seamless that generating a video feels as natural as typing a search query.

Google, with its API-first approach and the Veo model family, is positioned to own that future. But the week isn't over yet, and in AI, a week is a long time.

Try It Yourself

Want to see what Veo 3 can do before Veo 4 arrives? VO3 AI gives you access to Google's Veo 3 model for text-to-video generation. Create cinematic scenes, narrative sequences, and visual effects from text prompts — no editing software required.

Disclosure: VO3 AI is built on Google's Veo 3 technology. Examples shown in this article were generated using the platform.