What really caused the fall in AI stocks yesterday?

A headline about a "$20 billion hole" in OpenAI's revenue turned out to be an accounting error, but it was enough to drag AI stocks down.

By the SpyStocks desk · · 16h ago · 5 min read

Exclusive information from our research desk©

What really caused the fall in AI stocks yesterday? 🚨

Spoiler: A headline about a "$20 billion hole" in OpenAI's revenue, which turned out to be a mistake. 📡

What happened behind the scenes? We checked:

We checked the articles, the data, and the order of publications, and here's the truth:

1. The headline 📺

Thursday, October 8 -

The Financial Times reports that OpenAI, the company behind ChatGPT, recently informed investors that its annual revenue stands at approximately $50 billion. 💵

This is $20 billion less than the $70 billion reported just a few days earlier by several media outlets.

The market interpreted this in one way:

If OpenAI is generating less revenue, perhaps demand for AI is weaker than thought, and perhaps there's no one to cover the enormous investments in server farms, chips, and electricity,

Sounds logical. 💡

And the reaction came quickly:

The Nasdaq 100 index plunged by over 300 points within about half an hour, and Nvidia, Oracle, and other data center infrastructure companies were dragged down with it. ⚠️

2. The answer from within:

On that very same day, a source presenting data from OpenAI's investor presentation gave CNBC a completely different picture:

- Total revenue grew by 77% in the third quarter. 🔼

- Revenue from business customers grew by 107% during the same period. 🔼

And now, let's go back to the beginning: where did the $70 billion forecasts come from in the first place?❓

Here begins the real story... 📡

3. The $70 billion figure was not born on Thursday:

On September 29, several media outlets, including the Financial Times itself, reported that OpenAI's annual revenue was approaching $70 billion, after growing by more than 70% since July. 📈

In other words, this was not an official OpenAI forecast. The newspaper that published the "hole" was among those that, nine days earlier, published the very number from which the hole was measured!

On that very same day, September 29:

Steven Roskowski, a director at Oracle, one of OpenAI's largest cloud providers,

made an insider purchase of 25,000 shares at a price of $139.35, for approximately $3,483,800 of his personal funds. 🛒

Oracle "happens" to be building OpenAI's infrastructure... 📱

4. Gross versus net: the difference the market missed:

Here's the funny part: according to the Financial Times, the $70 billion figure was constructed by OpenAI's own investors in an attempt to directly compare it to Anthropic, its major competitor. ⚠️

But here's the problem:

The two companies count revenue using different methods.💡

Anthropic counts gross revenue, including sales that go through its cloud partners, such as Amazon and Google. 🧮

OpenAI counts net revenue, only what remains with it after the share of partners like Microsoft. 💡

Here's a simple example for easy understanding:

A customer pays $100 for a model through a partner's cloud. In Anthropic's method, all $100 are counted. In OpenAI's method, only the portion that ultimately reaches it is counted. 💡

To compare the two, investors converted OpenAI's figure to Anthropic's method and arrived at approximately $70 billion. 💡

The $50 billion is the exact same revenue, just counted as net.💵

5. The update:

After the publication (and after the market reacted with a collapse), the Financial Times updated the article and divided it into two figures:

Net revenue of approximately $50 billion in September, and gross revenue of approximately $70 billion, exactly the number previously reported. ✔️

But the stocks had already fallen. The market saw a headline and fled, even though not a single dollar disappeared...💵

And this aligns with what infrastructure companies have been saying for months:

The limitation is how much can be supplied, not how much people want to buy. ✔️

And now, to the number that truly matters: the growth rate:

How much the company is earning today, if the pace continues for a full year:

And here, no matter which method is used to count, this pace is only climbing.🔼

Look at the steps, according to what was published:

2024 - approximately $6 billion. 💵

2025 - over $20 billion, more than 3 times in one year. 💵

August 2026 - over $40 billion. 💵

End of September 2026 - approximately $50 billion net, and $70 billion gross. 💵

In percentages:

June: +131% July: +150% August: +182% September: +229% (approx. 3.3 times)

In just 4 months, the annual growth rate significantly increased by 100%.

And on the private user side, according to reports, OpenAI added more revenue this quarter than it added in all of 2025. 📈

And a small reminder: OpenAI's CFO directly linked growth to computing capacity, from 0.6 gigawatts in 2024 to 1.9 gigawatts in 2025,

The more servers there are, the more revenue. 📊

A company that doubles its revenue from business customers in one quarter does not look like a company with a demand problem... 💡

The question is how long it will take the market to understand that the headline was accounting-related, not business-related... 🚨

According to this morning's futures, is it possible the market already understands? 📡

Good intel travels.

Share this story

Bank of America: special report on Penguin Solutions, $100 target

PremiumAnalystsPENG11h ago2 min read

Unusual move at LB Pharmaceuticals: two analysts initiate coverage with a buy recommendation

PremiumAnalystsLBRX16h ago3 min read

Unusual options trade in GME

PremiumOption WhalesGME1d ago1 min read