OpenAI pursues $100 billion funding amid $14 billion projected 2026 losses

OpenAI is racing to secure up to $100 billion in fresh capital as the artificial intelligence company grapples with mounting expenses and growing threats from rivals.

People close to the matter say the money could push the firm’s worth to roughly $830 billion, more than what Argentina produces in a year.

Those involved expect talks to gain traction over the next two weeks. The size of the round signals more than ambition. Leaked internal papers obtained by The Information show OpenAI facing a $14 billion shortfall in 2026, with total red ink climbing to $115 billion by 2029. The company does not expect to turn a profit until sometime in the 2030s.

Massive computing project backs funding push

The fundraising plan calls for investments in two separate phases. Microsoft and Nvidia are set to put in money first, with Amazon discussing a commitment that could go as high as $50 billion.

The timing lines up with OpenAI’s Stargate venture, a $500 billion computing project built with SoftBank and Oracle. That effort aims to harness 10 gigawatts of electricity and deploy millions of processors to power what the company hopes will become artificial general intelligence.

User numbers climb despite pressure

Chief executive Sam Altman has worked to keep employees and investors confident even as financial strain builds. He wrote to staff on Friday that ChatGPT is “back to exceeding 10% monthly growth,” according to an internal Slack message that CNBC reviewed. Around 800 million people now tap into the service each week. But rivals are gaining ground. Google Gemini and Anthropic’s Claude have both pulled in users, with Gemini seeing a bump in web visits after Apple added it to its Intelligence features.

OpenAI rolled out GPT-5.3-Codex last week in an attempt to stay ahead. The company bills it as its first “agentic” coding tool, meaning it goes beyond simple suggestions. The software can tackle complex jobs on its own, fixing bugs across entire code libraries and handling software launches without human help.

Engineers at OpenAI used early builds of GPT-5.3-Codex to develop the final version, relying on it to troubleshoot training sessions and review test outcomes. The model also became the first to earn a “High capability” label under the firm’s Preparedness Framework for cybersecurity. That classification forced the company to add extra protections against automated hacking attempts.

Super Bowl sparks public feud

Tensions between OpenAI and Anthropic spilled into public view during Super Bowl LX. OpenAI ran a one-minute commercial aimed at developers, while Anthropic took shots at its competitor’s plan to show ads. One Anthropic spot depicted a “short king” asking for workout tips, only to get hit with promotions for shoe lifts. The ad closed with a pointed message: “Ads are coming to AI. But not to Claude.”

Altman fired back on X, labeling the ads “deceptive” and “dishonest.” Still, OpenAI moved forward with ad testing inside ChatGPT on Monday. The company tried to soften potential blowback by publishing its “Ads Principles.”

Only users on the Free and “Go” tiers, the latter costing $8 each month, will see sponsored content. Those paying for Plus, Pro, Team, or Enterprise subscriptions will not encounter any ads. The company also promised to keep ads away from conversations involving health, mental health, or politics. Sponsored messages will show up in a marked box at the bottom of the screen and are meant to stay separate from the chatbot’s actual responses.

As fundraising discussions pick up speed, OpenAI faces pressure to show it can make money from its massive user base through these new ad options and business tools. At the same time, it needs to justify the enormous spending required to compete in the race toward artificial general intelligence.

Want your project in front of crypto’s top minds? Feature it in our next industry report, where data meets impact.