In response to a post by the founder of Keras, Elon Musk warned that the AI industry may face a serious shortage of funds next year; if the time from paying computing power costs to the time when the equipment can stably generate revenue is delayed by several months, hundreds of billions of dollars in working capital alone may be needed. (Preliminary summary: Former SEC Chairman Gary Gensler warned: The AI bubble may trigger an economic recession before the 2028 election) (Background supplement: Bain reckoning: The AI industry must reach an annual revenue of US$6 trillion by 2031 to support data center investment) It is extremely difficult to power up, cool down and operate stably for a long time such a large-scale computing equipment. This was the first sentence Tesla CEO Elon Musk pointed out in his reply: The AI industry may face a severe funding shortage next year. He went on to deduce that if the entire industry is delayed by several months from paying the cost of computing power to the time when the equipment can stably generate revenue, hundreds of billions of dollars in operating capital alone may be needed, and he asks where this money will come from. The background of these remarks is the alternative offered by François Chollet, the founder of the deep learning framework Keras. His statement is very straightforward: Either you think that the AI infrastructure construction has come to an end, AI is just a short-lived bubble, and the demand for computing power has peaked; or you have to admit that the current wave of corrections is a rare buying opportunity. The bill for the computing power race. Musk is not actually talking about whether AI is useful, but a more pragmatic financial issue: time difference. Capital expenditures for data centers are almost upfront and one-time. From land, electricity, cooling to GPU cabinets, everything must be paid up front; but for these equipment to be truly converted into stable revenue, there is a long chain of model training, productization and customer payment. The risk he pointed out is that as long as this gap is delayed for several months at the overall industry level, the cash flow gap will expand by hundreds of billions of dollars. What’s even more troublesome is that the scale of this round of computing power investment is too large to be absorbed by a single company’s balance sheet. In the past two years, hyperscale cloud companies, AI start-ups, and data center operators have each raised debt, increased capital, and purchased power and computer rooms as an arms race; now the market is beginning to ask the same question: when will these equipments be able to support themselves? In an environment where capital costs are still relatively high, if we want to squeeze out hundreds of billions of dollars to invest in an asset with an ever-shifting payback period, the money can only come from existing cash flow, debt market or equity market. This is the sharp point of Musk's rhetorical question. Bubble theory and bargain hunting theory Cholet’s “bubble or buying point” and Musk’s “shortage of funds” are not inconsistent with each other. Both agree that the supply and demand of computing power has reached a stage where financial discipline must be used: if the commercialization speed of AI cannot catch up with depreciation and interest, valuation revisions will only be the result, not the cause. The real difference is time. The bullish side bets that the demand for inference will turn computing power into a long-term charging basis similar to electricity; the bearish side believes that when the depreciation cycle runs faster than revenue growth, this round of expansion will turn into overcapacity. This tension has been reflected in market sentiment in recent months. Whenever cloud companies raise capital expenditure guidance, the reaction of stock prices is no longer purely bullish, but also faced with pressure on free cash flow and depreciation and amortization; conversely, every time AI concept stocks retreat, there will be a debate about "whether this is another opportunity to get on board." Related reports Oracle rarely revealed that its data center "may not be able to recover its capital", Oracle's stock price plummeted 40% in June. AI capital expenditures will reach 3.2% of GDP in 2027, exceeding the US defense budget for the first time. Meta's free cash flow of $780 million hit a four-year low, AI bills began to eat up the industry, and fell 7% after the market. Musk warned: AI The industry may be severely short of funds next year, requiring hundreds of billions of dollars in additional investment> This article was first published on BlockTempo "Dongqu Dongzhi - the most influential blockchain news media".
Musk warns: The AI industry may face a severe funding shortage next year and require hundreds of billions of dollars in additional investment
In response to a post by the founder of Keras, Elon Musk warned that the AI industry may face a serious shortage of funds next year; if the payment […] Musk warned: the AI industry may face a serious shortage of funds next year, requiring hundreds of billions of dollars in additional investment> This article was first published in BlockTempo "Dongqu Dongzhi - the most influential blockchain news media".

This article may contain an automated translation. The original source is linked for verification.



