Industrial: How much of a pullback in tech stocks will the external rental and sales of computing power by Meta (META.US) cause?

Industrial released a research report stating that Meta (META.US) announced the external leasing of computing power, which has raised concerns about a slowdown in computing power demand, leading to a readjustment of global technology stocks. Regarding Meta's external leasing of computing power, it should not be overly pessimistically interpreted as an excess of computing power/Capex slowing down, because: 1) this is not a completely new piece of news, as there were related reports in May of this year; 2) Meta is a special presence among hyperscalers, with its main business being consumer-facing, which means its AI monetization capability mainly relies on advertising, while its cloud business exploration can enhance shareholder returns and cash flow; 3) Meta still has a computing power gap, as it was just reported this week that Alphabet Inc. Class C restricted its access + Meta signed an agreement with Crusoe; 4) hardware demand comes from inference, not from training side inflation. For this round of adjustments in global technology stocks, the core reason still lies in the fragile trading structure + profit void period (compared to hardware, the lack of high-frequency data tracking CSP's Capex plan and ROIC) + liquidity headwind period (tonight's non-farm payroll report), and short-term fluctuations provide another opportunity for positioning (focus on maximum drawdown of 10% at the bottom + implied volatility >30%). In mid to late July, with the reconfirmation of the favorable economic conditions by the US stock earnings season + the confirmation of the peak of inflation data, the technology sector is expected to usher in a new round of uptrend. Industrial's main points are as follows: Meta's announcement of external leasing of computing power has raised concerns about a slowdown in computing power demand, leading to a readjustment of global technology stocks. It needs to be clarified that the Nasdaq and US technology stocks, as the "engine" of this AI boom and the "anchor" of global technology stock pricing, are the key objects of observation and study. Therefore, it is necessary to conduct a thorough review of the internet stock market trends from 1995 to 2000, as well as the multiple corrections in the Nasdaq during the AI market since November 2022, in order to find patterns of adjustments in historical technology market trends. Firstly, during the 6-year period from 1995 to 2000 when the Nasdaq rose more than 7 times, there was actually only one major correction, which was the nearly 30% drop and 80-day adjustment caused by the ASIA FINANCIAL crisis in the mid-1998. Apart from this, the average maximum drawdown in historical corrections was only around 10% and the average adjustment duration was around 30 days. In the current AI market from 2022 ChatGPT to present, only adjustments from DeepSeek in 2025 and the trade war have exceeded 20%, while the average adjustment size of other rounds is also around 10% and lasts for about 30 days. Therefore, the conclusion can be drawn that unless it is a textbook-level systematic risk, a 10% drawdown lasting 30 days is already a relatively sufficient time and space. As for the core reason for the current adjustment in global technology stocks, it still lies in the fragile trading structure + profit void period (compared to hardware, the lack of high-frequency data tracking CSP's Capex plan and ROIC) + liquidity headwind period (tonight's non-farm payroll report), and short-term fluctuations provide another opportunity for positioning (focus on maximum drawdown of 10% at the bottom + implied volatility >30%). In mid to late July, with the reconfirmation of the favorable economic conditions by the US stock earnings season + the confirmation of the peak of inflation data, the technology sector is expected to usher in a new round of uptrend. Regarding Meta's external leasing of computing power, it is not advisable to interpret it too pessimistically as an excess of computing power or a complete slowdown in Capex, because: 1) this is not completely new news, as there were related reports in May of this year; 2) Meta is a special presence among hyperscalers, with its main business being consumer-facing, which means its AI monetization capability mainly relies on advertising, while its exploration in cloud business can enhance shareholder returns and cash flow; 3) Meta still has a computing power gap, as it was just reported this week that Alphabet Inc. Class C restricted its access + Meta signed an agreement with Crusoe; 4) hardware demand comes from inference, not from training side inflation. Additionally, here are a few quotes from overseas investors for reference: 1) Transitioning to Neocloud usually leads to an increase in capital expenditures, as seen in SpaceX; 2) According to DB, Meta is leasing older non-core computing power to generate cash flow, while continuing to retain premium resources like the Rubin series for the long-term strategic development of superintelligence; 3) According to Jefferies, referring to Amazon.com, Inc. in the early days of AWS development, initially building infrastructures to meet the peak demand of its internal retail business, then leasing out excess computing power, currently Meta still has 35% idle computing power, and reselling computing power will provide the company with more cash flow to invest in Capex; 4) According to JPM, there is a large discrepancy in the market's 2027 Capex expectations for Meta, ranging from $175 billion to $275 billion. Hardware analysts believe that Meta will not cut Capex, while internet practitioners seem more inclined to believe that capital expenditures will be reduced; 5) According to JPM, July was the month with the most intense deleveraging by hedge funds. Today's pullback is just cashing in on profits, while the significant rise in crowded short positions in software and others is consistent with the phase of short covering during deleveraging; 6) According to MS, the current model predicts capital expenditures of $175 billion and $205 billion in 2027/2028, mainly to support internal products. If Meta substantially advances its external cloud services, there may be a stronger incentive for an unexpected increase in capital expenditures to meet new demand. Risk warning If the Fed's monetary policy tightening is more than expected, global demand for computing power or the development of AI large models may fall short of expectations.