For two years, the AI video race had a presumptive winner: OpenAI’s Sora, the model whose first demos defined what text-to-video could look like. Then OpenAI walked away — the Sora app was discontinued in April 2026 and its API followed in September. The crown Sora left behind is now being fought over by three very different contenders: Kuaishou’s Kling, Runway, and Google’s Veo.
Each represents a different theory of how AI video gets won — price and volume, professional craft, or distribution at Google’s scale. Here is where the race actually stands.
Kling: the volume play
Kling, built by Chinese platform giant Kuaishou, has become the workhorse of AI video. Its strengths are practical: strong motion realism, generous generation lengths by the standards of the field, and pricing that undercuts Western rivals hard enough to make it the default for high-volume creators.
The business has scaled with the usage. Kling is reported to run at roughly $500 million in annual revenue — an estimate rather than an audited figure, but one that signals something real: creators are not just testing Kling, they are building production routines on it and paying at volume.
Kling’s challenge is the mirror of its strength. Competing on price and throughput invites a margin squeeze, and its home-market roots raise the familiar questions about availability and policy in Western markets. But in a young market, being the tool people actually produce with is the position everyone else wants.
Runway: the professional’s tool
Runway is the category’s veteran, a New York company that was building AI video tools before text-to-video was fashionable. Its Gen series models headline the product, but the real moat is everything around them: an editor, camera controls, and workflows built for people who make video for a living.
The market values that focus. Runway raised a $315 million Series E in February 2026 at a $5.3 billion valuation — serious money for a company whose product is, at heart, a creative tool. Studios and agencies are its natural customers, and it sells to them the way Adobe does: reliability, control and integration, not the cheapest generations.
Veo: the distribution giant
Google’s Veo plays a different game again. Quality is its headline claim — Veo generations are routinely cited for realism and for native audio — but its decisive weapon is placement. Veo is woven into Google’s AI subscriptions, its creation apps and YouTube’s ecosystem, which means hundreds of millions of potential users meet it without anyone having to win them one by one.
ByteDance’s Seedance deserves mention in the same breath: another platform-scale contender from a company that owns the world’s most powerful short-video distribution. The pattern is clear — as the models converge, distribution becomes the argument.
What the post-Sora race is really about
Sora’s exit taught the market something important: the best demo does not win, and even the best model does not win alone. The race is now about the unglamorous layers — price per finished minute, consistency across a project, editing workflow, and how many users a platform can put the tool in front of.
Expect the three-way split to persist: Kling pressing on volume and value, Runway on professional depth, Veo (and Seedance behind it) on reach. The winner of AI video will likely be decided less by whose model tops a benchmark than by whose tool a working creator opens on a deadline — and at the moment, that contest is genuinely open.
What to watch next
Three signals will tell you how the race is moving. First, default placements: every time a phone maker, editing suite or social platform picks a house video model, the market tilts. Second, cost per finished minute: the provider that halves it redraws everyone’s margins, including its own.
Third, watch where the talent goes. Runway’s professional base, Kling’s creator volume and Google’s research depth each attract different kinds of builders, and the third-party tools, templates and habits forming around each platform are the early shape of a moat.
Sora’s disappearance settled one question — no single model gets to be the synonym for AI video, at least not yet. The three contenders left standing are not really selling the same product, which is why all three can plausibly keep winning their own version of the race at the same time.
0 Comments