OpenAI pivots from consumer hype to business reality
Add Axios as your preferred source to
see more of our stories on Google.
Illustration: Brendan Lynch/Axios
OpenAI's decision to shut down Sora — its AI video app — is the biggest signal yet of the company's shift away from splashy consumer plays meant to dazzle the masses and toward a more practical, business-grounded strategy.
Why it matters: Compute resources to power AI are in higher demand than ever, and OpenAI's flashier, energy-hungry, unprofitable offerings were bound to be the first heads to roll ahead of a potential IPO.
Catch up quick: OpenAI announced Tuesday it would shut down Sora, the video-generation app that went viral after its September launch.
- The company also says it won't support video generation in the ChatGPT app.
- OpenAI is retreating on multiple consumer fronts at once — on Tuesday it announced that it's changing its consumer shopping strategy toward product browsing and away from handling purchases.
Zoom in: Fidji Simo, OpenAI's CEO of Applications, told employees last week the company is "orienting aggressively" toward high-productivity use cases and pausing all "side quests" to focus on coding and business users, according to CNBC.
- The company is still testing ads that will live alongside conversations in free and cheap versions of ChatGPT, a move designed to impress potential investors, not the consumer.
Between the lines: The consumer products weren't just power hungry and unprofitable. They were also generating legal and reputational risk.
- OpenAI faces lawsuits from families who say their loved ones harmed themselves after extended ChatGPT conversations, and from parents accusing the company of failing to safeguard teens.
- Sora specifically drew fire from deepfake experts and drew complaints from celebrity estates after users generated unauthorized videos of public figures.
Follow the money: Shuttering Sora eliminates an entire category of copyright and safety exposure ahead of an initial public offering.
Yes, but: The pivot to enterprise and agentic AI carries its own risks.
- Agents that take autonomous action on behalf of users will create new liability questions that haven't been tested yet.
- OpenAI is also under increased pressure from Anthropic, whose tight focus on AI coding tools first impressed developers and is slowly winning everyone else with its agentic work task tools thoughtfully designed for the rest of us.
- OpenAI's hiring of OpenClaw creator Peter Steinberger in February signaled the turn toward agents, but the company is playing catch-up.
What we're watching: The over-hyped Jony Ive hardware device, originally announced in early 2025 is set to be revealed this year, but won't ship before February 2027, according to court filings.
- Hardware is always hard, so a delay wouldn't be a shocker.
The bottom line: AI isn't like social media, where scaling is cheap. You can't do everything all at once when the infrastructure powering the technology is so expensive.
- OpenAI is betting that by trading consumer spectacle for business substance, it can impress investors and save itself some compute.
