IA : Fin de bulle et début du krach ?

The semiconductor industry, particularly that linked to AI, is experiencing a boom, leading many to cry "bubble." Last year, some stocks saw a threefold increase, and there's been over a 60% rise in large-cap stocks. Memory-related values have seen two, three, or even fourfold increases in mere months. The question arises: is this a bubble, and will an AI crash occur? What could end this exponential growth? This growth might only be the beginning, as money from tech giants and hyperscalers is flowing into AI-related companies for chips, printed circuits, raw materials, energy, and memory. Spending is projected to double from 2025 to 2026, and continue growing into 2027, albeit at a slower pace. While some valuations seem high, with certain companies trading at 20-30 times their revenue without making significant profits, a closer look at large-cap companies reveals that future profits, given current exponential revenue growth, are not being overvalued. For instance, Micron, a memory-related company, has exploded but is still considered undervalued relative to expected profits. There's a bottleneck, or bullwhip effect, where extreme demand for a few years drives up prices, leading to significant profits. These profits are then reinvested into building factories to meet demand. However, if demand isn't as strong as anticipated, overcapacity and price deflation could lead to a "mini-cataclysm." Currently, some companies are highly valued based on future profits, but this often reflects a "narrative" or government intervention, such as the desire to diversify away from a single supplier like TSMC in Taiwan. This premium is paid for companies like Intel, as governments are heavily investing to de-risk the supply chain. While some niche areas like photonics might be overvalued, more liquid, larger capitalization stocks are often considered reasonably priced, or simply reflect future profits already integrated into current prices. The real question is what's already priced in. For many stocks, the "best-case scenario" is already factored into prices. Even with record profits expected, the level of integration of next year's or 2027's profits into today's prices is key. This "bubble," if it is one, might just be starting. Token usage is increasing, requiring more computing power, but engineering advancements are leading to economies of scale, making it cheaper. AI is already generating substantial revenue and profits for some companies. The critical factor is whether these enormous investments will yield a return. Historically, stock market profits have never been stronger and are expected to continue rising. This suggests that while the market might be overvalued, the "winner-take-all" dynamic, concentration, and ETFs lead investors to accept this premium. The influx of money, partly due to leverage, and increasing retail investment in stocks (especially from younger generations due to expensive real estate) further fuels this. Investment by large tech companies is projected to grow and match their cash flow by 2026-2027, indicating continued investment capacity. However, several factors could disrupt this trend. Market concentration means performance depends on a few stocks and sectors. Interest rates are a major concern. Persistent inflation could prevent central banks from lowering rates, or even push them to raise rates. Higher rates increase the cost of money and bonds, impacting AI companies that rely on the bond market for financing. Continued inflation of raw materials and prices would make building and electrifying data centers more expensive, potentially affecting margins and future profits, even with economies of scale. Beyond industrial ramp-up, misallocation of capital, delays, or unused data centers due to component shortages could arise. Even with increasing AI adoption and token demand, industrial ramp-up might not keep pace. The market currently prices in a "best-of-all-worlds" scenario. However, issues with ramp-up or a miscalculation of demand could lead to oversupply. The electric car market serves as a cautionary tale: initial overestimation of demand led to significant revaluations. Historically, stock market crashes often begin with interest rate hikes, which can be triggered by rising energy prices. While geopolitical events might temporarily impact oil prices, the long-term effect could be a cascade of price increases. However, it's important to adjust for inflation; today's $100 oil is comparable to $50 oil twenty years ago. Also, less oil is used now, and alternative energies have a mitigating effect. While pump prices have risen, the shock is less severe than previous crises. Significant oil price increases often precede recessions. The US economy, with its concentrated valuations and numerous American investors and consumers with stock portfolios, is closely tied to the stock market. A major market crash could impact affluent consumers, leading to reduced spending, stock sales, and ultimately a recession. This could lead to lower demand, overcapacity from past investments, and ultimately deflation. These are sequential sub-waves to consider. Historically, markets don't drop during wars or embargoes, but typically six months later. Using past events like 1970s oil crises as scientific predictors for inflation and oil crises is problematic due to limited occurrences. Today, the world is less dependent on oil. Renewable energy, after an initial boom and overestimation post-COVID, is seeing renewed interest as part of a diversified energy mix, alongside oil, gas, and nuclear. The global middle class is growing, especially in emerging economies, increasing energy demand. The market operates on three pillars: exploding profits for companies benefiting from large capital expenditure investments by major players, abundant capital driving up stock prices due to buybacks and reduced float, and a divide between thriving sectors and those struggling with tariffs, high energy costs, and reduced consumer spending. Mergers and acquisitions are expected to continue stimulating the market. While speculative investments are being made based on the AI narrative, it's crucial to acknowledge that much of this narrative was already present a year ago. The key is understanding what's already priced in. The market is bound to become more volatile. In this "bubble" scenario, the market shifts from upstream to downstream. Software companies, currently struggling, might be the first to benefit in a later cycle, similar to how Google and Meta leveraged Cisco and Intel's internet investments. AI will drive innovation in areas like robotaxis, drones, weaponry, autonomous cars, and robotics, leading to a re-industrialization cycle. However, a major AI stock crash would likely stem from an oversupply relative to demand, or a reduction in demand, coupled with rising interest rates. Higher rates increase the cost of capital, limiting credit and the ability to invest in projects that may no longer be profitable at those rates. This would constrain investment and the ability to deploy new offerings.