A piece of news about "being short by $20 billion" why could it press down an entire hardware chain?

The negative candle in global tech stocks on October 8 started with a matter of definition. According to the UK's Financial Times, OpenAI disclosed to investors an annualized revenue of about $50 billion as of the end of September, roughly $20 billion less than the approximately $70 billion widely circulated in the media at the end of the previous month. After the news came out, the Nasdaq fell 1.25%, the Philadelphia Semiconductor Index fell 3.39%, NVIDIA Corporation (NVDA.US) fell 2.94%, wiping out more than $168.4 billion in market value overnight, Oracle Corporation (ORCL.US) fell 5.48%, Micron, SK Hynix, and SanDisk fell more than 4%, and in the optical communications sector, Applied Optoelectronics, Inc. fell 13.58% and Coherent fell 9.63%. But the half-sentence in this news that most deserves to be read to the end comes later: the $20 billion mainly comes from a difference in statistical definitions, and does not mean its actual revenue suddenly declined. The difference is that companies like Anthropic include sales revenue from cloud partners such as AWS and Alphabet Inc. Class C cloud in their annualized revenue, while OpenAI does not use this statistical method; from the perspective of the company's own growth rate, OpenAI's overall third-quarter revenue run rate grew 77%, and its enterprise business run rate grew 107%. The same thing, measured with different rulers, can differ by $20 billion and one big negative candle. This is exactly the place in this industry chain that is easiest to misread right nowthe market still has no unified definition for indicators such as "annualized revenue," "backlog orders," and "computing capacity." Take the most direct example. Three overseas pure computing power leasing companies use three sets of terms: CoreWeave (CRWV.US) talks about revenue backlog, equal to remaining performance obligations plus estimated revenue to be recognized, and will also proactively disclose how much of it has begun delivery; Nebius (NBIS.US) talks about RPO (remaining performance obligations, a statutory disclosure indicator under accounting standards ASC606/IFRS15) and contracted backlog, and discloses the amount of each large order, the yield per megawatt, and the proportion of customer prepayments covering capital expenditures; IREN (IREN.US) has the clearest layeringrecognized revenue, annualized operating capacity already in operation (operating ARR), and signed but not yet operational (contract ARR). For the same company, these three layers of figures can differ by more than a hundredfold. Across the three sets of definitions, "signed" and "received" are clearly separated. Domestically, most use the single term "orders in hand" to contain all states. GBA AI COMP (01396) is relatively restrained in its 2026 interim results: intended AI computing power orders in hand exceed RMB 37 billion, of which more than RMB 20 billion has been delivered and is stably billed; more than 95% of newly added orders this year include customer prepayment terms, and more than 95% are long-term contracts of 3 to 5 years. Only when the two figures are given together can readers know how large the reserve and the realization each are. Range Intelligent Computing Technology Group (300442.SZ) discloses a physical definitionAIDC revenue of RMB 1.995 billion, a year-on-year increase of 126.24%; VNET Group, Inc. Sponsored ADR (VNET.US) uses "total operating capacity of base-type IDC of 1,007 megawatts, signing rate of 96.3%, capacity in racks of 744 megawatts, and racking rate of 73.9%"; GDS Holdings Ltd. Sponsored ADR Class A (09698) expresses it using signed capacity and a domestic investment plan of RMB 3050 billion over the next three years. Companies such as Sharetronic Data Technology (300857.SZ) and Jiangsu Lettall Electronic (603629.SH), which have expanded from manufacturing or their main electronics business, talk about yet another setthe proportion of computing power business revenue in total revenue. On October 8, what fell in the A-share and Hong Kong stock markets was precisely the part being led around by this chaotic set of definitions. That day, the Shanghai Composite Index fell 0.79%, the ChiNext Index fell 3.15%, and the STAR 50 fell 4.82%, with turnover expanding to RMB 1.68 trillion; Hong Kong's Hang Seng Tech Index fell 2.89%, hitting a new low since September 2024; CPO, semiconductors, and optical communications led the decline, with Yuanjie Semiconductor Technology (688498.SH) and others hitting the daily limit down directly. And in the same week, this company disclosed that its first-half data center revenue increased by 640% year on year. Stock prices are driven by definitions and expectations, while revenue is driven by delivery. The two rhythms are inherently out of sync. So before comparing valuations of such targets, do one thing first: translate each company's "orders" into the same rulerhow much is already signed, how much is already delivered, and how much money has already been received. After translating, compare again, and you will usually find that the multiple relationships are completely different from what you saw at first glance. This is also why, from the same interim report data, some read "better than expected" while others read "pie in the sky"; the disagreement is not in the data itself, but in which layer of the definition is being read.