AI startup valued at $800 billion struggles to monetize, cancels a string of projects: the truth behind the AI hype

An AI startup valued at $800 billion is struggling to monetize, with hundreds of billions of dollars in deals and initiatives still “on paper.”

OpenAI has continued to raise massive capital, including a $122 billion funding round that valued the company at $852 billion ahead of an IPO expected later this year. But several major product launches and partnerships have been scaled back, canceled, or delayed, raising questions about how quickly OpenAI can convert announcements into durable revenue.

Funding surge, but monetization challenges

OpenAI’s strategy has included broad product experimentation across advertising, shopping, healthcare, social media, browsers, physical devices, video creation, and a marketplace similar to the App Store. In January, Fidji Simo, OpenAI’s head of applications, defended the approach as testing multiple directions at once, saying the company did not view these efforts as separate bets.

However, within two months, OpenAI reversed course on some of its most prominent initiatives, including abandoning its Sora video model and canceling a Disney licensing deal that had been expected to include a $1 billion equity investment. The shift has been described as a move toward tighter financial discipline as the company faces pressure to generate revenue ahead of a potential IPO and competition from Anthropic.

Sora and the Disney licensing deal

OpenAI’s Sora had ranked among the top free apps on the App Store. In December, OpenAI announced what it called a landmark deal with Disney: Disney would invest $1 billion and license 200 iconic characters for use in Sora’s video creation app.

Under the agreement, Disney would air Sora-created videos on Disney+ and build new apps using OpenAI’s model. OpenAI CEO Sam Altman previously praised Sora as a key step toward “ChatGPT for creativity,” while Disney CEO Bob Iger also expressed confidence in the investment.

In March, OpenAI shut down Sora and canceled the deal. The article cites that Sora’s compute costs were extremely high, with estimates of roughly $15 million per day at its peak, and that Sensor Tower and Appfigures estimated Sora’s total app revenue to be under $3 million. OpenAI also stopped related AI models tied to Sora.

Instant Checkout for shopping in ChatGPT

In October, Walmart and OpenAI announced a partnership enabling customers to buy 200,000 Walmart products directly in ChatGPT through “Instant Checkout.” The article says user adoption was limited and sales through ChatGPT disappointed.

Walmart executives told Wired that the conversion rate—users who actually purchased products shown in ChatGPT—was significantly lower than when users were redirected to Walmart’s website. The article also notes that other retailers such as Shopify and Etsy listed products in ChatGPT, but only a limited number supported Instant Checkout and item descriptions were not always accurate.

In early March, OpenAI halted Instant Checkout. It said users could instead browse and purchase through dedicated apps within ChatGPT, including Walmart’s Sparky chatbot. OpenAI stated it wanted more flexibility and that the initial Instant Checkout version did not provide the flexibility it sought.

GPT-4o shut down

In February, OpenAI formally shut down the GPT-4o model. The article describes the model as known for a friendly, cheerful tone, while also being criticized as overly pandering and easily agreeable.

Stargate data center project

On the first day of Donald Trump’s second term, OpenAI announced the Stargate project, described as worth $500 billion, in partnership with Oracle and SoftBank to build AI data centers nationwide. The announcement was made at the White House with Altman, Larry Ellison, and Masayoshi Son present.

A year later, the article says the project stalled due to challenges recruiting staff and deploying data centers, largely tied to disagreements among partners over structure, control, and responsibility. After the collaboration collapsed, OpenAI reportedly considered building its own infrastructure but later abandoned it due to fundraising challenges and conflicts with partners. OpenAI declined to comment.

The article also says OpenAI and Oracle did not extend a 1,000-acre data center in Abilene, Texas after financial negotiations failed and OpenAI’s compute needs changed. Oracle said it completed hiring an additional 4.5 gigawatts of AI capacity, while Crusoe said a 1.2 gigawatt project remained on track.

Nvidia investment plan

In September, OpenAI and Nvidia announced Nvidia’s intention to invest up to $100 billion in OpenAI, with no specific timeline. The article says OpenAI could receive around $30 billion if the IPO comes early, according to Nvidia CEO Jensen Huang, and that Nvidia’s financial results emphasized there was no guarantee the deal would materialize.

In the fundraising round of $110 billion (later raised to $122 billion), Nvidia committed $30 billion in cash, described as Nvidia’s largest investment. The article notes it is unclear whether the funds would be delivered in a single tranche or installments.

The investment is described as helping OpenAI purchase Nvidia GPUs, with the article arguing that the structure effectively supports Nvidia’s own product sales.

AMD deal tied to data center milestones

A month later, OpenAI announced a deal with AMD involving 160 million shares (about $30 billion) in exchange for using AMD chips for 6 gigawatts of data centers.

The article says shares would be allocated only when data centers were operational and met technical and commercial milestones. By the end of 2025, it reports that no shares had been allocated.

If the project succeeds, AMD could earn tens of billions of dollars; if it does not, the article says there would be no payment.

Source: Forbes

By Phuong Linh