Paul Tudor Jones: “We Are in 1999” — Why the AI Rally Has One to Two Years Left

It’s one of the most important market statements of the week, and it comes from a man who predicted the 1987 crash. Paul Tudor Jones, billionaire and hedge fund legend, gave an interview Thursday morning on CNBC Squawk Box. His message: the AI-driven bull market has one to two years left. We are currently in “1999 mode” — about a year before the dotcom stocks peaked in early 2000. When the rally ends, Jones said, the drawdown could be significant. “Just imagine the stock market goes up another 40 percent. The market-cap-to-GDP ratio would be at 300 to 350 percent. You just know there’ll be some kind of trigger.” These words land in a market that just hit fresh record highs — the S&P 500 closed Wednesday at 7,365 above 7,300 for the first time, the Nasdaq above 25,838.

What Jones Really Says

Behind the “1999” metaphor sits a concrete analysis of historical tech booms. Jones references two parallels: Microsoft’s dominance in the 1980s and the commercialization of the internet in the mid-1990s. Both phases brought years of productivity gains and market upside. AI is currently in a similarly early stage — the technology is not yet mature, but the market is already pricing it aggressively. If the AI boom follows the internet boom, 1-2 years of further upside are plausible before valuation reality bites.

This statement has two readings. The bullish one: Jones confirms the trade — anyone holding AI stocks can expect strong performance for another 1-2 years. The bearish one: when an investor of his caliber draws the 1999 comparison, investors should approach position sizing with elevated caution from now on. A drawdown after 1-2 years on a 40 percent appreciation would mathematically be a 30-50 percent decline from the eventually reached peak. That’s not trivial.

The AMD Data Point Carrying the Story

AMD’s Q1 earnings from Tuesday evening delivered an important detail yesterday that supports the bull thesis. AMD inventories of $8.045 billion at quarter-end. Of that, 31.6 percent was finished goods (versus 21.1 percent a year ago). Only 9.3 percent was raw materials (versus 8.7 percent a year ago). This shift shows: AMD produces massively more finished chips than before and plans no slowdown. Lisa Su, AMD’s CEO, would not hold 32 percent of inventory as finished products if hyperscaler demand were softening.

Su revised AMD’s growth expectations upward yesterday: over 35 percent annual growth, an addressable market of $120 billion by end of decade. Goldman Sachs raised the price target to $250. AMD’s stock gained 18.6 percent on Wednesday — the largest single-day gain by a Mag 7 stock in years.

Microsoft Raises Capex by $25 Billion

Hyperscaler capex escalation continued yesterday. Microsoft raised its annual capex forecast by $25 billion. Meta by $10 billion. Both companies cited “higher component costs” as justification — language showing that even hyperscalers can no longer buy as freely as they want. They have to pay more per chip unit. That’s good for AMD and Nvidia, but it also shows the capex wave is escalating from necessity, not strategic choice.

A random-walk observation: Micron is expected by Wall Street to generate 80 cents of gross profit per dollar of revenue in 2026. Semiconductor manufacturers currently have pricing power not seen since the 1990s. This confirms Jones’s 1999 comparison on a technical level: then as now, picks-and-shovels providers (Cisco/Intel then, Nvidia/AMD/Micron now) are the biggest beneficiaries of the early phase of a technology wave.

Iran Peace Plan Comes Closer

Geopolitically, the situation has dramatically eased from Monday. Iran is currently reviewing a 14-point U.S. proposal to end the conflict. Pakistan is mediating between the parties. Donald Trump declared Thursday morning that he had “very good talks” with Iran and that “it’s very possible we’ll make a deal.” Iran will return its response to the 14-point plan via Pakistan today.

The market reaction is unambiguous: Brent fell 3.8 percent to $97.38 — from above $115 at the start of the week. WTI fell to $91. If the conflict actually ends this week, that would be one of the strongest bullish macroeconomic triggers since the COVID reopening. Lower oil prices mean lower inflation, lower Fed rates, higher equity valuations.

Fortinet 22 Percent Higher

Outside the headlines, Fortinet made a story yesterday that has gotten less attention than it deserves. Q1 earnings beating all expectations, OT security billings (Operational Technology) up 70 percent, $1 billion in free cash flow in one quarter. Stock +22 percent in after-hours, further gains on Thursday. Cybersecurity is a sub-story of the AI wave: more AI means more attack surface, more AI-driven attacks, more demand for AI-driven defense. Fortinet, Palo Alto Networks, and CrowdStrike are the primary beneficiaries.

DoorDash also gave strong Q2 guidance and gained 1 percent. Datadog and Vistra also contributed to the broad earnings strength. Whirlpool, however, fell after weak numbers — a hint that consumers are pulling back on big-ticket items. More on that when May consumer data arrives.

What the Jones Truth Means for Retail Investors

Jones’s 1999 comparison is not a sell recommendation — it’s a position-sizing recommendation. Three concrete consequences for retail investors. First: anyone heavily invested in Mag-7-concentrated portfolios should diversify over the next 6-12 months. Picks-and-shovels (Vertiv, Constellation Energy, Quanta Services) offer less valuation risk than Nvidia or Palantir. International markets (Nikkei +5% yesterday) are less expensive than U.S. Mag 7.

Second: don’t stop savings plans. If Jones’s comparison is correct and we have 1-2 years of upside left, the next 12-24 months of savings plan amounts will be on average below the peak once the correction comes. That’s good for long-term investors.

Third: increase cash position. Berkshire holds $400 billion in cash. Smaller investors should hold 10-20 percent of their portfolio in cash or money market ETFs (FTSE Money Market). When the correction comes, you want dry powder.

The Nasdaq Backtest Encourages and Concerns

The Nasdaq-100 has gained 540 percent over the last 10 years — almost twice the S&P 500. The Invesco QQQ Trust is a central investment vehicle. Top holdings — Nvidia, Alphabet, Apple, Microsoft — make up almost half the ETF. Nvidia alone 8.4 percent. This concentration is simultaneously strength and weakness. Strength: in a bull market, QQQ significantly outperforms the broader market. Weakness: in a Mag 7 correction, disproportional loss.

Anyone believing long-term that AI grows further buys QQQ and holds. Anyone believing Jones is right and we see a 30-50 percent drawdown in 1-2 years either buys VOO (S&P 500, less concentrated) or holds more cash until the correction. Both are valid approaches — the worst mistake would be being fully in QQQ and panic-selling in the correction.

Bottom Line

Paul Tudor Jones did not sell down the S&P 500. He drew a historical comparison sending two messages simultaneously: the bull market continues (1-2 years), but the end phase will be brutal. Investors who understand this position differently now than 6 months ago. They diversify, hold more cash, avoid over-concentrated Mag 7 positions. The Iran peace hopes Thursday and Microsoft/Meta capex increases provide short-term tailwinds. But anyone who listened carefully to Jones yesterday knows: this is not the phase to buy aggressively. It’s the phase to hold what you have, position defensively, and protect gains where possible.