As "The Big Short" predicts a crash, a well-known investor presents an "AI bull market timetable": the bubble could expand for another year.

As the Nasdaq index continues to climb under the push of the AI wave, concerns about the overvaluation of tech stocks in the market are also heating up. However, according to Dan Niles, founder and portfolio manager of Niles Investment Management, this AI-driven uptrend is far from over - at least it will continue for another year before reaching its true peak. The Debate between Bubble Theory and Fundamentals Niles' views come at a sensitive moment when market sentiment is highly polarized. Just on May 11, Michael Burry, the prototype of the movie "The Big Short" and a well-known investor, issued a stern warning again: after experiencing a "parabolic" surge, the Nasdaq 100 index is heading for a violent reversal. Burry pointed out that the actual P/E ratio of the Nasdaq 100 index is as high as 43 times, far exceeding the reasonable implicit level of about 30 times. "Wall Street may have overestimated the profitability of the fastest growing, highest valued batch of leading companies by more than 50%," he said. Since the end of March, the Philadelphia Semiconductor Index has soared by nearly 70%, with deviations from the 200-day moving average seen only twice in history - in July 1995 and March 2000, the latter happening just before the collapse of the internet bubble. But Niles provided a completely different interpretation framework. He compared today's AI-driven market to the internet bubble era, but emphasized a key difference - this time, there is strong fundamental support behind the growth. Data confirms this judgment: CICC research shows that information technology and communication services contributed 77% of the S&P 500's increase and 67% of first-quarter profit growth, also contributing to 55% of actual GDP growth in the quarter. The first-quarter profit growth of the S&P 500 reached 28%, a new high since the fourth quarter of 2021, with the highest growth rate in the semiconductor and equipment industry reaching 99%. "The bubble may already exist, but before it bursts, it is still possible to further inflate," explained Niles. He compared a set of historical data: in the three years after the release of ChatGPT at the end of 2022, the Nasdaq index rose by 122%, while during the same period of the internet construction era, the increase was 109% - this set of data reveals the astonishing extent of the surge, and also implies the intensity of capital investment behind the market. How Deep is the Moat of Fundamentals? The core argument supporting Niles' optimistic assessment is that global tech giants are still increasing their capital expenditures. According to TrendForce's report released on May 6, influenced by significant increases in capital expenditures guidance from North American cloud service providers, the combined capital expenditures estimate for the nine major cloud vendors - Google, AWS, Meta, Microsoft, Oracle, ByteDance, Tencent, Alibaba, Baidu - has been raised to about $830 billion by 2026, with an annual growth rate from 61% to 79%. Amazon alone plans to invest around $200 billion in data center expansion and AI chip deployment, while Microsoft has raised its capital expenditures to $190 billion, a growth of about 130% year-on-year. These funds are being converted into revenue and profits at a visible speed. The latest financial report season data provides strong evidence: Google Cloud's first-quarter revenue grew by 63% year-on-year to $20 billion, with cloud backlog orders nearly doubling to $462 billion and an operating profit margin of 32.9%. Alphabet's overall revenue increased by 22% to $109.9 billion, Microsoft grew by 18% to $82.9 billion, and Meta grew by 33% to $56.3 billion. Microsoft's AI business has an annual revenue run rate of over $37 billion, a 123% year-on-year growth. A recent report by Morgan Stanley further shows that the net profit estimate for Mag 7 (Apple, Nvidia, Amazon, Google, Microsoft, Meta, Tesla) in 2026 is expected to increase by 25%, more than twice the 11% increase for the rest of the 493 stocks in the S&P 500 after excluding Mag 7. Goldman Sachs has also raised its year-end target for the S&P 500 to 7600 points, as AI investments are expected to contribute about 40% of profit growth for 2026 to 2027. LSEG data shows that the earnings forecast for the S&P technology sector is expected to grow by over 31% this year. At the same time, the valuation of the AI infrastructure sector has fallen to its most attractive level in nearly seven years after a deep correction at the beginning of the year. Morningstar research points out that the discount rate of the AI theme sector has reached its largest since 2019, with Chief Equity Strategist Michael Field stating, "AI is not a bubble about to burst, its underlying fundamentals are very robust, semiconductor demand continues to exceed expectations, and core growth drivers such as data centers are still in good shape." Paradigm Shift: From Chat to Action In Niles' analysis framework, January 30 is a key inflection point that the market has underestimated. On that day, the final naming of OpenClaw marked the official start of the Agentic AI (Agentic AI) era. Unlike traditional chat-based AI, intelligent agent systems can autonomously perform complex multi-step tasks, requiring computing power that is 10 to 100 times that of traditional AI. This judgment has been fully validated by industrial data. According to Huafu Securities research data, traditional Chatbot consumes a fixed token amount for each question and answer, while AI Agents require autonomous planning, multi-round chained reasoning, and cross-tool invocation, increasing the unit task token consumption by 20 to 30 times. Agent systems are also driving the architectural shift in computing power from pursuing GPU density to heterogeneous computing collaboration - CPUs handle multi-agent coordination control, while GPUs handle inference calculations. Intel CEO Chen Liwu pointed out that the ratio of AI data center CPU to GPU has tightened from a previous 1:8 to about 1:4, with the potential to trend towards 1:1 in the future. The explosive growth in token consumption further confirms this paradigm shift. Data cited by Niles shows that in the two months before the final confirmation of OpenClaw, the growth in word generation was about 20%, while in the two months after, this growth accelerated to over 120%. "At least until early next year, you should see strong growth," Niles predicts, which will then face more severe year-on-year base effects. In fact, CICC positioned the current stage as "more like the internet construction period from 1996 to 1998" in its research at the end of last year, rather than the peak of the bubble from 1999 to 2000. Midpoint of the Bull Market? Niles is not turning a blind eye to risks. In an interview with The Market NZZ in March, he admitted that the AI arms race is unsustainable, and the massive AI computing infrastructure construction may prove to be a misallocation of capital. But in a more recent interview, he showed a more short-term optimistic side: he issued a research report on March 31 stating that "history will not simply repeat, but often has a rhythm," comparing the current AI infrastructure construction on a large scale to the internet infrastructure period from 1997 to 1998, the current AI infrastructure construction is in its fourth year, and after the macro impact, the market will still benefit from the inertia of underlying industry investments. He pointed out that by 2026, the market generally expects cloud vendors' capital expenditures to grow by about 30%, this number first rose to 60% after the first-quarter financial reports, and then further increased to 70%, providing solid fuel for the market's uptrend. Previously, Paul Tudor Jones, one of the most watched macro investors globally and founder and chief investment officer of Tudor Investment, also provided a time schedule for the AI bull market that is both optimistic and cautionary: drawing from history, the current AI-driven US stock market bull market still has about one to two years of upside potential, he himself has recently increased his holdings of AI-related stocks through a "basket". But if the market rises by about 40%, the valuation expansion will create conditions for a "suffocating" pullback. This legendary hedge fund manager gained fame for successfully predicting and profiting from the crash of "Black Monday" in 1987. The truly influential aspect of Jones' remarks in stirring the market is not the conclusion itself, but the precise year-specific time axis he provided for this AI bull market: 50% to 60% complete, with one to two years remaining. This clear positioning - whether ultimately proven to be prophetic insight or precise backtesting error - will have a profound impact on institutional investors' risk exposure adjustments. Goldman Sachs has provided a more cautious framework, suggesting investors pay attention to five major "bubble burst" signals: peak investment spending, declining corporate profits, rapidly increasing corporate debt, Fed rate cuts, and widening credit spreads. Currently, investment spending is accelerating rather than peaking, profit growth remains strong, and although credit spreads have recently widened, they are still at historic lows - the conditions for a bubble burst are not yet ripe, but risks are accumulating. In this game of bulls and bears, Niles stands in the middle ground: he acknowledges the existence of a bubble, but believes there is still room for further expansion of the bubble. This judgment may reveal a more fundamental reality - in the early stages of the AI industry transformation, market optimism and real industry progress are intertwined in unprecedented ways, simply labeling the market as "bubble" or "non-bubble" may miss the core features of the current market. For investors, Niles' views provide a clear roadmap: at least for the remaining time of 2026 and early 2027, the profit growth driven by AI and the expansion of CKH HOLDINGS capital expenditures will continue to support the market. The real test will come after the year-on-year growth rate begins to slow down and the penetration rate of intelligent agent AI surpasses the early explosive stage.