【TMGM Financial Recap】Openai's Annualized Revenue Dropped By $20 Billion, AI Stocks Plunged Collectively

The $70 billion revenue reported at the end of last month was not OpenAI's official figure. The estimate at the time was: restoring July annualized revenue to about $40 billion on a "total basis" basis, then applying over 70% growth, the figure was close to $70 billion by the end of September. The problem lies in that $40 billion: it includes revenue OpenAI previously earned from selling technology to Microsoft to Azure customers, but after the business agreement was updated in April this year, this revenue is no longer counted under OpenAI. OpenAI's latest investor materials show July annualized revenue approaching $30 billion, so by the end of September it was about $50 billion. What changed this 20 billion was not how much OpenAI earned, but the denominator used to value it.

Previously, The Market Priced Orders Based On Unfulfilled Orders

The pricing logic of this industry chain is built on the revenues of several upstream model companies: only when model companies' revenues grow rapidly can there be sustained computing power procurement, and only then can the long-term demand for chips and data centers be met. If the first link is revised downward, the assumptions for the next two stages must be recalculated.

OpenAI has committed hundreds of billions of dollars in computing power and infrastructure, expecting to consume nearly $280 billion in cash by 2030, and is currently negotiating a new round of financing valued at about $1.4 trillion. Based on $50 billion in annualized revenue, its $852 billion historical financing valuation corresponds to a price-to-sales ratio of about 17 times; At a $1.4 trillion valuation, this multiple rises to about 28 times. With the revenue base reduced by about 30%, the same valuation corresponds to a higher multiple, and the entire chain where "you only collect money later" must be revalued.

The Biggest Drop Is In The Segment Farthest From Income

The segment furthest from final revenue saw the biggest declines: Applied Optoelectronics fell 13.58%, Coherent fell 9.63%, CoreWeave fell 7.77%; At the chip tier, the Philadelphia Semiconductor Index fell 3.39%, Intel dropped 5.34%, and Micron fell 4.79%.

Oracle fell 5.48%, marking its largest single-day drop since July 16. On the same day, there were other reports that due to New Mexico's two vetoes of pipeline lines, the commissioning date for its 2.45GW data center project, codenamed "Jupiter," has been postponed from summer this year to next year. The company is considering transporting compressed natural gas to the park to speed up the schedule, with costs about four times that of pipeline gas prices. For a 100-megawatt capacity, a large trailer can only support for about 40 minutes.

On the other hand, it was barely affected: Apple rose 1.11%, the only one among the Big Seven to rise; Palantir rose 2.35%, and Goldman Sachs upgraded its rating from neutral to buy that day. The selling segment was the segment that had not yet generated cash, while the segment that had income on the books was retained.

The Real Downside Amplifier Is Position Versus Interest Rate

Attributing a single day's decline to a single investor material misses two backgrounds. Two trading days earlier, the S&P 500 had just closed at a record high of 7,818.93 points on October 6, and the Nasdaq 100 hit a new high of 31,224 points on the same day. Over the past six months, the S&P 500 has risen 18.3% cumulatively, while excluding AI components, the Goldman Sachs index rose only 6.7% over the same period. This high point was supported by a few stocks, and the higher the index goes, the greater the impact of the same document's price.

Position size and interest rates determine the scale of the sell-off. According to Goldman Sachs trading desk data, the net exposure of the seven giants accounts for about 22% of total U.S. stock exposure, the highest since early 2022, with semiconductor and equipment exposure doubling compared to the start of the year. During the same period, the 10-year U.S. Treasury yield was 5.28%, and the 30-year yield was 5.66%. Since the end of June, the Financial Conditions Index has tightened by about 40 basis points, almost all driven by rising long-term interest rates.

The bullish side has not exited: Bank of America Wealth Management believes the market is still consolidating, and investors are waiting for the Q3 earnings season to begin next week; According to FactSet's forecast, S&P 500 Q3 earnings grew 29.5% year-on-year.

There will be two key points next week. October 14 (Wednesday) 20:30: US September CPI release; At the same time, the Q3 earnings period is intensive, and guidance from computing power chains and cloud providers will test whether orders can be converted into revenue.