qinbafrank
qinbafrank|Nov 22, 2026 14:42
Is it 2000 or 2008 now? The recent market crash has reignited this debate. Let's look at some differences between the present and 2000 or 2008. 1. Compared to 2000: Many semiconductor stocks did indeed exhibit a parabolic price pattern in the first half of the year, somewhat similar to 2000. However, the fundamental difference lies in the following: Around 2000, the internet industry hadn't yet found a viable business model. It wasn't until 2000, and even into 2006 and 2007, that the four major business models—gaming, advertising, e-commerce, and value-added services—were truly established. Currently, the infrastructure is highly developed, and information transmission is rapid, allowing for the rapid commercialization of AI. At the beginning of the year, this link (https://(((x.com)))/qinbafrank/status/2023982690144579970?a=FqpjTZVmDX3xU7XU3cA&s=46) discussed how the mobile internet accomplished in less than 10 years what the internet took 20 years to achieve, and AI might accomplish in even less time what the mobile internet took 10 years to accomplish. Currently, the commercialization of AI is progressing quite well. Here are a few examples: 1) This link (https://(((x.com)))/qinbafrank/status/2114901741183619102?a=r_h3DSm04ibzNc_-bGI&s=46) discussed how Anthropic should easily reach $50-60 billion in revenue this year and is highly likely to surpass $100 billion in actual revenue next year, which is a very large scale. 2) Previously, this link (https://(((x.com)))/qinbafrank/status/2117158355034870100?a=VBFbGfOCrR0HlV8ZqF4&s=46) also mentioned that the three major cloud providers had a combined cloud business volume of $110 billion in the third quarter, and their growth rate had been accelerating since the first half of the year, clearly continuing this acceleration trend in the third quarter. This is how internet commercialization works; once validated, penetration and commercialization will steadily increase until saturation. Currently, it seems commercialization hasn't reached saturation yet. 2. Comparison with 2008 US housing prices peaked in the second half of 2006. On April 2, 2007, New Century Financial, the second-largest mortgage lender in the US, filed for bankruptcy. In the second and third quarters of 2007, more than half of the US mortgage lenders went bankrupt. By August 2007, the serious delinquency rate for floating-rate mortgages in the US had reached nearly 20%. In the same month, three funds from BNP Paribas suspended redemptions. In September, Northern Rock went bankrupt, and Citigroup, Credit Suisse, and others disclosed huge losses in subprime mortgages. Buyers of subprime securities were also borrowing short-term funds. Once assets became difficult to value, lenders refused to continue lending, forcing holders to either replenish collateral or sell assets. The shortfall in household mortgage payments was thus passed on to investors who had hoped to redeem their assets at any time. The collapse of underlying asset prices began to cascade upwards, almost unstoppable. In October 2007, the US stock market peaked, and the market began a year-and-a-half-long decline. The key point here is that starting in early 2007, mortgage defaults began on a large scale. The current situation is that the overall scale of corporate debt, procurement commitments, lease agreements, and guarantees is indeed very large; this was also mentioned in the Federal Reserve report on November 13th. However, the cycles are all very long. Procurement commitments, leases, and guarantees typically have cycles starting at five years and lasting over a decade; they are not immediately fulfilled. Here are a few examples: 1) Meta has $278.99 billion in leases that have not yet commenced, with leases starting gradually from the remainder of 2026 to 2036, typically for two to three decades, with a maximum of 30 years. 2) Google disclosed $707 billion in future fixed or guaranteed commitments with terms of one year or more, the vast majority of which are long-term supply agreements. Energy agreements have terms of 2–26 years, extending as far as 2054. 3) Another example is Nvidia. Its Q3 earnings report this week disclosed $518 billion in future commitments, data as of October 25th. The company also listed $32 billion in computing power cloud service agreements and $118 billion in guarantee exposure. It should be noted that the $32 billion purchase only begins to be calculated after certain conditions are met. Of the $390 billion in commitments, $270 billion is for the next year and a half, which seems substantial. However, given Nvidia's current business scale, it genuinely needs such a large volume of purchases. This year, Nvidia has approximately $420 billion in revenue and a 74% gross profit margin, meaning its costs are at least $110 billion – all of which are procurement costs. Nvidia officially projects a 75% growth rate next year; if the gross profit margin remains unchanged, this means procurement costs will reach $190 billion next year. The Ohio project guaranteed for OpenAI/SB Energy will not take effect until 2028. This week, the acting CEO also stated that the guarantee will only take effect when the data center's land, power, and enclosure infrastructure are truly ready. 4) The Federal Reserve report shows that advanced computing-related company debt totals $1.1 trillion, a very large amount, but also with a very long maturity. FTSE Russell's analysis of 172 bonds issued by several major US cloud computing companies, totaling $423.02 billion, shows an average principal recovery period of 17.27 years. Microsoft's longest average is 22.69 years; Oracle's debt, which is of greatest concern, has the shortest average maturity of 15.70 years. What does this tell us? In 2007, widespread mortgage defaults occurred in the US, cascading down to mortgages, subprime mortgages, and various financial derivatives built upon existing loans. Currently, however, we haven't seen any major cloud vendors, new cloud vendors, or large tech companies substantially defaulting. The Federal Reserve's Financial Stability Report is also proactively addressing the potential impact of lower-than-expected AI revenue on financial stability. In fact, the market, like the Federal Reserve, is pricing in potential future risks in advance. Of course, different people will react differently to the same data. Positive reaction: These debts, procurement commitments, lease agreements, and financing guarantees are either long-term or haven't even taken effect yet. Negative reaction: With such a large financing volume and ecosystem guarantees, a major collapse is inevitable if commercialization falls short of expectations. Therefore, overall, while some semiconductor stocks showed a parabolic trend in the first half of the year, the fundamental commercialization trend is far better than in 2000. Although the overall debt and guarantee commitments are massive, the cycles are long, and there have been no substantial defaults. It won't reach the levels of 2008, perhaps only 2005. Moreover, if commercialization continues to accelerate and the growth rate of commercialization exceeds the growth rate of capital expenditure, this potential risk will be largely eliminated. Therefore, the current concern is about debt risk and a 28-year spending cliff, coupled with strong fears of a policy reversal due to dramatic changes in the political landscape after the election, leading to concentrated position liquidation. However, simply copying the patterns of 2000 and 2008 is not accurate. In the short term, we'll look at the extent of position clearing. But let's reiterate the medium- to long-term perspective discussed in the previous post: 1) Oil price/inflation trends will determine the direction of the Fed's interest rates; 2) Under the new political landscape, AI policies will be tighter yet slightly more moderate than the current market expects; 3) Whether AI commercialization continues to accelerate.
+5
Mentioned
Share To

Timeline

HotFlash

APP

X

Telegram

Facebook

Reddit

CopyLink

Hot Reads