Google's Omni 1.1 Flash Is a Price Cut Disguised as a Model Launch

On August 27, Google's DeepMind unit shipped Gemini Omni 1.1 Flash, an upgrade to the video-generation model the company launched this spring. The headline specs: a ten-second render per run, extensions that build footage up to a forty-second ceiling, and a new ability to read ten seconds of earlier footage while extending — where older models could look back only about one second.

Forty seconds is a different product than ten. OpenAI's Sora 2 Pro, the top of its video line, caps a single clip at 25 seconds. At forty seconds with persistent context, a "clip" becomes a "scene" — long enough for an ad, a product demo, or a training segment. That is the difference between novelty content and billable work.

But the spec is not the news. The price is.

Google listed the model at $0.10 per second for 720p, $0.03 per second at 360p, $0.15 at 1080p, and $0.30 for 4K upscaling. The 360p drafts cost a third as much and render up to 60% faster — Google's way of saying iterate cheap, finish expensive. A forty-second 720p scene runs four dollars. A forty-second 4K scene runs twelve.

Put that against the market. Google's own previous flagship, Veo 3.1, listed $0.40 per second for standard quality and $0.15 for Fast. OpenAI's Sora 2 matched Google's $0.10 at 720p, but Sora 2 Pro scaled to $0.70 per second at 1080p — ten seconds of finished footage cost $7 on Sora against $1.50 on Omni. Google priced a longer, 4K-capable model at or below every comparable tier in the market while the best-known rival charges close to five times more for less.

The timing makes the price meaningful. OpenAI is sunsetting its Sora 2 API on September 24 with no replacement listed, and it shut the standalone Sora app in April. The name most retail investors associate with AI video is leaving the developer business entirely. Google, meanwhile, has been giving the model away since May inside YouTube Shorts and the YouTube Create app — a consumer funnel on the one distribution platform no rival owns.

Strip the marketing and the mechanism is simple. Video generation is the most compute-hungry thing a model can do; every second is billed in silicon. Google makes its own silicon with TPUs, trains its own model, and owns the distribution. When you control all three, you do not defend a high price — you set the low one to manufacture volume, and the volume fills your own chips. That is the abundance-scarcity paradox at work: as video becomes nearly free, the scarce assets are the compute that generates it and the attention that monetizes it. Google sits on both sides.

That is the right frame for Alphabet's central investor question, which is not "can Google make video" but "can Google's $205 billion bet pay off." In the second quarter, Alphabet guided 2026 capital expenditures to $195–$205 billion, up from $180–$190 billion in the spring, and posted negative free cash flow for the quarter — and the stock fell after hours anyway, despite revenue up 24% to $119.8 billion. The demand side of that ledger is real: Google Cloud grew 82% to $24.8 billion, the contracted backlog jumped $50 billion in a single quarter to $514 billion, and CEO Sundar Pichai said earlier this year that Cloud is demand-driven but supply-constrained. The bear case is that the capacity sits idle. Aggressively cheap video generation is a direct answer: it manufactures the single most compute-intense workload that exists, monetizes it through the API and through YouTube — whose advertising pulled in $11.1 billion last quarter — and feeds the machines the market fears are empty.

Now the honest boundary. This is a positioning move, not yet a revenue line; analysts read launches like this through competitive positioning, and the durable signals are API usage and how rivals reply. If $0.03–$0.10 video does not translate into volume, Google has simply donated margin to a market where OpenAI, Meta, and Kuaishou can counter-price. Watch the next few quarters for one thing: whether Gemini video-API usage and cloud backlog convert. A price cut that manufactures demand closes the case for the spending; a price cut that merely gives away capability reopens it.

Alphabet's thesis is legible even before the numbers confirm it: when a company prices a creative good toward zero, it is betting that the value sits on the other side — in the chips that generate the content and the attention that distributes it. The next earnings will show whether that feed is real, or whether cheap video is cheap because it has to be.