Original Author: Long Yue
Original Source: Wall Street News
The current market situation is strikingly similar to 1997-1998—a Wall Street tech veteran is already starting the countdown for the AI bull market.
On May 11, Dan Niles, a well-known chip analyst from the dot-com bubble era and founder of Niles Investment Management, gave an in-depth interview on "The Master Investor Podcast," systematically presenting his judgment on the current AI market trend: the AI bull market is not over yet, but he predicts a major correction could occur around early 2027, and investors should start preparing now.
Meanwhile, a JPMorgan survey of 56 global investors found: 54% expect the U.S. stock market to experience a correction of over 30% this year or next, with 45% believing it will happen in 2027—highly aligning with Dan Niles' assessment.
It's 1997-1998, Not 1999, and Certainly Not 2000
Niles compares the current market to 1997-1998, not the peak bubble years of 1999-2000 that many fear.
The logic is as follows: ChatGPT launched at the end of 2022, and AI infrastructure build-out is now entering its fourth year. During the internet era, the Netscape browser debuted in 1994, and 1997-1998 was also the third and fourth years.
In 1997, the Thai currency crisis erupted, the S&P 500 fell up to 11% intra-year but still closed the year up 31%. In 1998, the Russian bond default and Long-Term Capital Management (LTCM) collapse saw the S&P 500 drop up to 19% intra-year, yet it still gained 27% for the full year.
Niles says: "Back then, the overarching theme of internet infrastructure build-out provided a cushion, so every macro shock became a buying opportunity. It's the same today."
He believes the oil price shock triggered by the Iran war is a "man-made event," easier to resolve than the currency crisis or bond defaults of the past, therefore judging this to be another cyclical low.
Agentic AI: The New Fuel Driving This Bull Market Forward
Niles attributes the core driver of this year's market to one term: Agentic AI.
Simply put, previously you would ask ChatGPT a question, and it would give you an answer. This is "conversational AI."
Agentic AI is fundamentally different. Dan Niles gives an example: "You can tell it, 'This is Wilfred, go to the BBC website and get this data, go to Bloomberg for that data, go to CNBC for something else, then compile it all into an Excel spreadsheet.'" This series of operations requires numerous concurrent API calls, consuming 10 to 100 times more compute tokens than a chat-based AI.
Data already proves this: For the two months before OpenAI's release of [presumably a key Agentic AI model/API] on January 30, 2026, token growth was about 20%; in the two months after release, token growth exceeded 120%.
This has directly boosted capital expenditure expectations for hyperscale cloud providers: at the start of the year, the market expected ~30% capex growth for 2026; after Q1 earnings, this rose to 60%; and after the latest round of earnings, it climbed again to 70%.
Niles' conclusion: This is not a small change; it's an order-of-magnitude leap, sufficient to support further gains in AI-related stocks.
Changing Hardware Landscape: CPU Comeback, GPU Under Pressure
The architectural nature of Agentic AI is quietly rewriting the competitive landscape of AI hardware.
Training large models involves repeating the same task, which GPUs excel at; conversational AI inference is also manageable. But Agentic AI requires simultaneously managing multiple applications and coordinating multi-step tasks, which is essentially "orchestration"—a CPU's strong suit.
Dan Niles says: "The ratio used to be roughly 8 GPUs to 1 CPU. As we shift to Agentic, that ratio will move closer to 1 to 1."
This means: Intel and AMD benefit, while Nvidia is "marginally affected in terms of its stock price performance."
But Semiconductors Are Severely Overbought
Dan Niles shifts gears: short-term risks cannot be ignored. "In the short term, the current overbought level in semiconductors is the most severe since just before the 2000 or 1995 crashes. That's certain."
He specifically notes the semiconductor ETF is up about 70% year-to-date, and even the Iran war shock couldn't push it down.
However, he stresses that short-term overbought conditions do not equate to a breakdown in long-term logic—the demand for compute from Agentic AI is real. He is willing to accept the risk of Intel potentially falling 15-20% in the short term because he believes the stock will be even higher by year-end.
2027: Where Will the 30-50% Correction Come From?
Dan Niles is already mapping out scenarios for the next cycle.
The Agentic AI surge began around January 30, 2026. Based on this, by early 2027, growth will start lapping tough high-base comparisons, and growth rates will naturally slow significantly. At that point, what happens to the market?
"I think, from the highs they are at then, these stocks could drop 30% to 50%," he says.
The reference point is recent: in 2022, the "Magnificent 7" tech giants fell an average of 46%—that was just the aftermath of the pandemic-era tech build-out wave receding, far smaller in scale than the current AI frenzy.
Another potential trigger point is OpenAI. Dan Niles points out that OpenAI and Anthropic combined account for about half of the backlog orders at hyperscale cloud vendors. The two companies went from a combined ~$7 billion in annualized revenue in early 2025 to now approaching $70 billion (Anthropic ~$45B, OpenAI ~$24B)—an astonishing growth, but this money must be squeezed out of other companies' budgets.
"When OpenAI's revenue was still $20 billion at year-end, it publicly claimed a commitment of $1.4 trillion in capital expenditure over the next eight years. Where does that money come from? If OpenAI runs into problems, that would significantly accelerate this process."
He also highlights a structural liquidity pressure: IPOs from companies like OpenAI, SpaceX, and Anthropic are coming in succession, each potentially valued in the trillions of dollars. "That money has to come from somewhere else. Fund managers aren't sitting on piles of idle cash; they have to sell something else."
Three Signals Flashing Simultaneously: Stocks, Oil, Bonds—One Must Be Wrong
The first thing Dan Niles does every morning is check oil prices, bond yields, and the stock market simultaneously.
The current combination makes him uneasy: stocks at all-time highs, oil prices up about 60% year-to-date, and both the U.S. 10-year and 30-year Treasury yields hitting yearly highs.
Historically, in 10 out of the past 12 recessions, a sustained rise in oil prices preceded them. If oil stays around $90 for one or two quarters, inflation will pick up, consumer purchasing power will be eroded, and then a major stock market correction becomes inevitable. McDonald's recent earnings mentioned pressure on low-end consumers, with same-store sales missing expectations—when these signals start appearing, you have to worry.
He also notes that the incoming Fed Chair Kevin Warsh is inclined to cut rates and views AI as a deflationary force, "which is a positive factor pushing the market higher in the short term." But he warns that if 10-year or 30-year yields keep climbing, market valuations will face real pressure. His conclusion:
Among stocks, oil, and bonds, one must be wrong. When one of them reprices, it could trigger significant market turmoil.
His advice is concise: "Hold lots of cash now.—Just at the end of March, I thought it was a good time to be aggressive entering the market. But now, I think you should hold lots of cash and be highly vigilant about the eventual resolution."
JPMorgan: 54% of Institutional Investors Expect a Major Correction in 2027
Dan Niles' warning is not an isolated view.
A roadshow feedback report released on May 12, 2026, by JPMorgan's global market strategy team shows that analyst Eduardo Lecubarri led the team to five cities in Latin America, meeting with 56 institutional investors.
The core data from the report is as follows:
- 92% of surveyed investors believe the stock market return for the full year 2026 will be positive, but not a single one expects gains exceeding 20%;
- 54% of surveyed investors expect a stock market correction of over 30% sometime between 2026 and 2027 (9% expecting it in 2026, 45% in 2027);
- 75% of surveyed investors believe there is more than 20% upside remaining before reaching a tech bubble peak;
- Regarding regional allocation, views on Europe were highly consistent—100% underweight Europe, 100% overweight the U.S.;
- The sectors most favored by investors are, in order: Technology, Utilities, Industrials.
This highly aligns with Dan Niles' "1998 logic": the bull market is far from over, but the timeline for a major correction is already quietly forming within market consensus.
Quantum Computing: Huge Potential, But Don't Rush
At the end of the interview, Dan Niles also discussed quantum computing. He is a firm long-term believer: "I'm a firm believer in quantum; I think we will get there eventually"—but he cites Bill Gates' famous quote: We tend to overestimate technology in the short term and underestimate it in the long term.
"The earliest AI paper was written over 50 years ago. When did ChatGPT appear? End of 2022. Quantum computing likely follows a similar path. The arrival of quantum computing IPOs will bring back market attention, but truly disruptive applications will take longer to arrive than most people imagine."
Big Tech Company Divergence: Google Out in Front
The recent earnings from tech giants have made Dan Niles' judgment clearer.
Google Cloud: Q4 YoY growth 48%, latest quarter accelerated to 63%, a 15 percentage point acceleration.
AWS: Accelerated from 24% to 28%, a 4 percentage point increase—respectable given its size as the largest cloud provider.
Microsoft Cloud: From 38% to 39%, almost flat.
"These numbers tell you who is truly executing and who is gaining market share," Dan Niles says.
He states his conclusion directly: "Who is the best big tech company to own for the next 3 to 5 years? Clearly Google. They have the full technology stack; you should bet on them. Unless something dramatic happens, they will continue to be winners because they have everything and have massive cash flow to support it."
Google's advantages include: its own large language model Gemini, in-house AI chips (over a decade of history, the longest among the three cloud vendors), strong cash flow supported by its ad business, and the Android ecosystem covering over 75% of global smartphones. Microsoft relies on OpenAI and lacks its own foundational model; Amazon's AI products have limited brand recognition.
Meta's situation is relatively concerning. Dan Niles points out Meta has no public cloud, so it can't sell excess compute to external businesses; its in-house ASIC chip development also started relatively late. More importantly, this quarter Meta saw the first-ever quarterly decline in Family of Apps user numbers—between the two growth engines for its ad monetization model (user count and price per user), the former has now turned, "That is something to worry about."
Regarding Apple, Dan Niles believes the AI-powered Siri and the foldable iPhone will drive a replacement cycle—citing the iPhone 6 as an example: screen size increased from 4 to 5.5 inches, pushing Apple's revenue growth from 7% to 28%.
Stay Nimble, Don't Be Greedy
Dan Niles summarizes his market philosophy in one phrase: "Be Nimble. Hold strong convictions, but hold them loosely."
His assessment framework is: short-term momentum is upward, with Agentic AI and expectations of easier monetary policy still being two key engines; but by early 2027, these growth numbers will start lapping tough comparisons, the explosive growth from Agentic AI will enter a more moderate phase, and combined with the potential risks from OpenAI and the liquidity shock from mega-IPOs, "stock prices could drop 30% to 50% from the highs at that time."
What to do now? Hold more cash, keep a close eye on the three coordinates every morning—oil prices, Treasury yields, and the stock market—and be ready to adjust at any moment.
