$15 million per day? OpenAI’s Sora video spending sparks concern

Sora costs spark growing concern

OpenAI’s Sora video tool is drawing huge attention for its viral clips and massive compute bill. Forbes published an analysis estimating Sora’s compute bill could be about $15 million per day, but OpenAI has not confirmed that figure.

OpenAI executives have said the company expects to finish the year at about a $20 billion annualized revenue run rate.

In comparison, filings with Microsoft and financial coverage suggested OpenAI-related losses of about $12 billion in the most recent quarter. Each number is drawn from company commentary and public reporting and should be attributed.

Daily burn reaches extreme levels

Forbes cited analyst modeling that annualizes Sora’s compute bill at about $5 billion. Some analysts, such as those quoted in industry coverage, have modeled a per-ten-second clip GPU cost in the low-dollar range for consumer configurations.

For example, about $1.30, but the assumed GPU minutes per generation vary across models and reports. Treat these numbers as analyst estimates rather than company disclosures.

With millions of clips created daily, the cost multiplies quickly. SemiAnalysis researchers agreed the estimate seemed reasonable based on typical GPU rental rates and model complexity.

Viral growth supercharges expenses

Appfigures estimated about 627,000 iOS installs in Sora’s first week, while OpenAI and other reports said Sora crossed one million downloads in under five days.

Forbes later cited data suggesting roughly four million downloads by Halloween. Different trackers sample different stores and markets, so these numbers are best reported as ranges with attribution.

OpenAI declined to detail usage, leaving analysts to rely on estimates based on GPU needs, user behavior, and model versions. A rough model assumes 4.5 million app users, with a quarter of them posting 10 videos per day. That equals more than eleven million daily videos, which drives costs to eye-watering levels.

Heavy compute drives the bill

Video generation is far more expensive than text because Sora must reason across three spatial dimensions plus time. Each frame requires precise motion and consistent lighting, which demands long GPU sessions. Mathivanan estimated that some videos take eight to ten minutes on four GPUs working together.

Renting those GPUs at just under $2 an hour creates a baseline cost that quickly becomes overwhelming at scale.

OpenAI has begun selling extra generations and is piloting monetization options. Pricing currently varies by plan and generation, and OpenAI has said it will adjust free generation limits to make the economics viable.

Free access fuels nonstop usage

Analysts say OpenAI is following a familiar tech playbook by offering Sora free to build audience and engagement. Millions of users jumped in to make funny short clips, many of which cost more to produce than they return in revenue. The strategy creates rapid visibility but also drains compute budgets at a stunning pace.

OpenAI leaders hinted that this free period will not last much longer. With costs rising fast, the company cannot sustain unlimited free video generation. Sam Altman said simple meme-making among friends cannot be funded by any ad model at this scale, signaling upcoming limits or paid options.

App store success adds momentum

Sora reached number one on Apple’s free app chart in October, showing how intensely users flocked to the tool. By early November, it slipped to fifth, but engagement remained high. Even with that shift, millions continued generating clips each day, creating a massive stream of data and activity.

Every clip feeds OpenAI valuable training material unless the user opts out. Video data paired with text prompts helps sharpen Sora’s understanding of motion and scene consistency. This added value partly explains why OpenAI is willing to endure such high upfront expenses while refining its video models.

Model costs may fall later

Some analysts believe Sora’s price tag today reflects early inefficiencies that will decline over time. Mathivanan projected that video inference could become five times cheaper within a year as research improves. He also estimated a threefold drop again in 2027 as hardware and designs continue to advance.

If those predictions hold, Sora’s heavy compute needs may become manageable. Analysts said early stages always look unsustainable because cost curves fall rapidly as technology matures. The challenge is surviving the gap between viral growth and meaningful efficiency gains that would ease OpenAI’s financial burden.

Market share drive shapes strategy

OpenAI seems more focused on dominating early AI video markets than making Sora profitable today. Analysts likened this approach to classic web era tactics where companies burn cash to lock in user habits. Viral Sora clips keep the app visible while boosting the number of creators who rely on the tool.

Future revenue may come from power users such as filmmakers, advertisers, or studios who value higher quality controls. If Sora reaches professional grade levels, these groups could pay significantly more than casual meme makers.

User data strengthens training edge

Sora generates an enormous library of user videos paired with text prompts. This data is valuable for training because it shows how users expect actions and motions to look across many scenarios. Each rejected draft or unused clip also gives OpenAI signals about where models fail to maintain consistency.

The company can train on this material unless users opt out, creating a growing pool that rivals may not match. These insights could help OpenAI sharpen future versions of Sora and strengthen its competitive position even while current operating costs remain high.

Free period likely ending soon

OpenAI executives have already said the window for unlimited free video creation is closing. The sheer compute load makes it impossible to sustain. Altman noted in an interview that there is no advertising model today that could fund people making small joke clips for private chats.

This suggests new restrictions, paid tiers, or reduced free credits may arrive soon. As OpenAI moves toward sustainable operations, Sora’s most active users may be pushed toward higher-priced versions, while occasional creators keep some form of limited free access.

OpenAI losses amplify pressure

OpenAI is already spending more than twice what it earns, which magnifies the impact of Sora’s heavy compute bill. The company lost more than $12 billion last quarter alone.

Layering billions in video-generation costs on top of those losses intensifies concerns among observers about long-term sustainability.

Despite the financial strain, OpenAI continues to expand aggressively. Leaders appear confident that future price cuts, efficiency gains, and monetization features will eventually offset early losses. This pattern is common among fast-moving tech companies that endure steep early costs for growth.

Why these costs spark alarm

The core tension lies in how long OpenAI can sustain Sora’s runaway usage while spending $15 million a day. Viral videos fuel excitement, but each clip carries a real cost that piles up fast. With so many users generating content for free, the financial imbalance becomes impossible to ignore.

OpenAI hopes efficiency improvements and paid tiers will close the gap soon. Analysts say the economics could look more realistic once better hardware and cheaper inference arrive.

What this means going forward

Sora is shaping the future of AI video by mixing massive user creativity with enormous compute needs. The tool’s rapid rise shows how eager people are to experiment, even if each ten-second clip carries a steep price behind the scenes.

OpenAI must now balance growth with cost control as usage patterns evolve. Readers are watching to see whether Sora becomes a sustainable product or remains an expensive early experiment.

Sora isn’t the only tool changing how creators work. See what Microsoft just unleashed in Microsoft’s new AI video tool, which is free and shockingly easy to master.

Where Sora stands right now

Sora sits at a critical moment. It remains a viral hit with millions of active users, yet it also represents one of the most expensive experiments in OpenAI’s history. The next phase will depend on when costs fall and how quickly paid features replace today’s free firehose of video creation.

OpenAI expects efficiency gains and new revenue streams to make Sora viable over time. Whether that happens soon or much later is what analysts are trying to understand.

To see how it’s already making waves, check out how the new Sora app might just give TikTok a run for its money.

What do you think about Sora’s fast growth and rising costs? Share your thoughts.

Read More From This Brand:

• Sora video generator coming to ChatGPT
• Meet Sora, OpenAI’s visionary video model
• Gemini Live upgrade gives your AI a sharper vision in real time

Don’t forget to follow us for more exclusive content on MSN.

If you liked this story, you’ll LOVE our FREE emails. Join today and be the first to get stories like this one.

This slideshow was made with AI assistance and human editing.