AI’s Possible “Y2K Liquidity Moment” – and Why Earnings Still Matter More

THE FOUR MAJOR POINTS

1. The Scorecard	2. Earnings > Headlines	3. AI Capex vs. Dot-Com	4. A Y2K-Style Liquidity Fuse?
Oak Harvest stayed constructive through 2H25, forecast a harder and more volatile 2026, identified the late-1Q “7th inning stretch,” and expected a strong 3Q/summer rally.	Q2 earnings and revenues are materially stronger than expected. The market is expensive, but current prices are being supported by rapidly rising EPS rather than P/E expansion alone.	Both cycles are infrastructure booms led by semiconductors. The key difference: today’s leaders have real profits and cash flow, while much of the speculative frontier remains private.	Late-1999 Y2K liquidity helped extend the dot-com melt-up. July 31’s U.S.-Japan yen intervention is not the same policy tool, but could it become a short-term global-liquidity catalyst?

OPEN – WE HAVE SEEN THIS MOVIE BEFORE

Investors, for over 15 months, long before many others, the Oak Harvest investment team had been asking whether the AI cycle is rhyming with the dot-com boom. Recall, we have said this from a positive messaging standpoint.  If it’s a bubble, and it’s early in the cycle, you want to be long equities.  This summer, that comparison got more interesting – not less.

We have extremely strong earnings. We have an AI infrastructure boom and have started a semiconductor capex upturn. We have a stock market making new highs with expanding market breadth. And now we have something we have not had in decades: coordinated U.S.-Japan intervention to support the yen. Everyone the last time that happened was? Specifically on June 17, 1998, during the Asian financial crisis, mid Dot.com.

So today I want to ask a question that hit me this weekend: did Kevin Warsh and global policymakers just create an AI-cycle version of the late-1999 Y2K liquidity window – the kind of liquidity event that helped fuel the final dot-com sprint into March 2000?

I am not saying they did. I am saying the setup is important enough that investors should understand the analogy – and the differences.

POINT ONE – THE OAK HARVEST SCORECARD: PROCESS OVER PREDICTION

Before we look forward, let’s score the roadmap we have been using. Back in our 2025 work, we kept coming back to three variables: earnings, interest rates – including both real rates and inflation expectations – and the U.S. dollar. We stayed constructive on the bull market rather than treating every headline as the start of a recession. Even with tariffs. Even with the war.

Then in our first-half 2026 outlook we said the year could be a harder bull ride: weakness early in the year, followed by summer strength, and a dot-com-style investment cycle analog worth watching. In front of the late 1q26 sell, we described the AI market as being in a “7th inning stretch” – not early, not late, just resting and not finished.

In our June second-half 2026 outlook, the working thesis was simple: strong third quarter, more difficult fourth quarter. So far, the summer rally is doing what we thought it could do. On August 7, the S&P 500 closed at a record 7,757, up 3.6 percent for the week and about 13 percent year-to-date, while the Nasdaq gained more than 5 percent for the week.

That does not mean the next call will be right. It means the framework has been useful: follow earnings first, then interest rates and their two components, liquidity, the dollar and energy – not the daily emotion that the financial networks try to generate.

POINT TWO – EARNINGS ARE THE ENGINE, AND THEY HAVE BEEN STRONG, ACCELERATING and GETTING STRONGER

This is why I keep saying: keep your eyes on the prize. The prize is earnings.

FactSet’s August 7 report is extraordinary. With 88 percent of S&P 500 companies reported, 86 percent beat earnings estimates and 76 percent beat revenue estimates. Blended second-quarter earnings growth is now 50.4 percent year over year, versus only 23.1 percent expected at June 30.  Here’s the link for those who want the detail. FactSet Earnings Insight

Yes, there is an important footnote. Alphabet and Amazon had unusually large investment-related gains. They marked up their venture capital portfolios. The market won’t reward those onetime gains. If you take those two companies out, S&P 500 earnings growth drops from 50.4 percent to 32 percent. That is still incredible and an acceleration. Revenue growth is 15 percent – the strongest since late 2021.

And this is not just one quarter. FactSet currently expects earnings growth of 27.4 percent in Q3, 25.2 percent in Q4, 30 percent for calendar 2026 and another 13.6 percent in 2027. Bottom-up EPS is about $359 for 2026 and $405 for 2027 and moving higher.

Valuation is not cheap. But valuation is a horrible market timing tool. The S&P 500 trades around 20 times forward earnings, slightly above its 10-year average of 19. But here is the distinction from 1999, for 6 consecutive quarters the PE has declined. Earnings have been doing more work than the multiple. The S&P500 is pretty efficient though, as of this writing, the 10-year treasury was 4.649%, old rule of thumb, 1/.04649= PE, right now that’s 21.51x$359/s=7722.1, and the S&P500 closed at? 7757 last Friday, within less than ½ of 1% of the calculation.

This difference matters enormously for retirees. A high multiple with falling earnings is dangerous. A high multiple with rapidly rising earnings can stay high much longer than skeptics expect.

POINT THREE – THE AI CAPEX BOOM: SEMICONDUCTORS ARE THE TOLL ROAD

The third point is where the dot-com comparison remains remarkably useful: both cycles required a huge physical buildout. The internet needed fiber, routers, servers, switches and semiconductor fabrication. AI needs GPUs, high-bandwidth memory, advanced packaging, networking, optical links, power semiconductors, cooling and enormous data-center electrical capacity.

And FactSet is showing the earnings translation in real time. Information Technology revenue grew 35.9 percent in Q2. Semiconductor and semiconductor-equipment revenue grew 77 percent. Semiconductor-industry earnings grew 135 percent. Remove semiconductors from the technology sector and its Q2 earnings growth falls from 70.4 percent to 34.3 percent.

That is why semiconductors remain the toll road underneath AI, the backbone of AI Compute.

But this is also where the risk lives. Capex has exploded. The hyperscalers are becoming more asset-heavy. The key question is shifting from “Is AI real?” to “What return will investors earn on the next dollar of AI capex?”Recently, MSFT, AMZN and GOOGL provided a positive glimpse of the answer to that question.

The dot-com boom broke partly because companies built far ahead of monetizable demand. Today, the demand is already here – agentic AI is increasing query volume and compute intensity, parabolically. Cloudflare said on their quarterly earnings call that agentic traffic exceeded human traffic for the first time ever, and that will continue exponentially for a while.  The biggest public beneficiaries produce real profits and cash flow. OpenAI and Anthropic also remain private, keeping part of the speculation outside public equities. So yes, this cycle can still be a bubble in places. Yes, it rhymes, even down to the earnings growth rates, but it is not a carbon copy of 1999.

POINT FOUR – DID THE JULY 31 YEN INTERVENTION LIGHT A LIQUIDITY FUSE?

Now we get to the most interesting comparison.

The dot-com bubble did not peak immediately after LTCM. In fact, after the 1998 Asian currency crisis and the October LTCM blowup, the Fed restored liquidity. Then in 1999, even while the Fed was raising interest rates, it created special liquidity facilities to protect the financial system from Y2K funding stress. The Special Liquidity Facility opened October 1, 1999 and ran into April 2000.

That did not cause the dot-com bubble by itself. But it removed a major tail risk at exactly the moment investors were willing to take more risk. From October 21, 1999 to the Nasdaq peak on March 10, 2000, the Nasdaq Composite rose about 85 percent.

Fast forward to July 31, 2026. The United States and Japan jointly intervened to support the yen after it fell to multi-decade lows. The New York Fed executed transactions on behalf of the Treasury. That is a very unusual event. But here is the distinction: this was foreign-exchange intervention, not a new Federal Reserve quantitative-easing program and not the same as the 1999 Y2K liquidity facility. Any Fed backstop would require Federal Reserve decision-making and authorization.

Still, could the intervention have a liquidity effect through global markets? Possibly, If policymakers follow with additional measures, maybe announced at Jackson hole, financial conditions could change quickly.

That is the bullish analogy.  Warsh does reference Greenspan a lot. The bearish difference is that Warsh is not Greenspan in late 1999. Inflation is still above target. The 10-year Treasury ended August 7 around 4.64 percent. Markets are debating the next Fed move is a hike rather than a cut.

Remember our interest-rate framework: the 10-year yield is basically real rates plus expected inflation. If inflation expectations rise, valuations compress. If real rates rise because financial conditions tighten, valuations compress. For growth stocks, either interest route raises the discount rate on future cash flows. Both components rising at the same time would be BAD.

So the question is not, “Did the Fed turn on the money printer?” It did not. The question is: did policymakers just remove a global liquidity risk at a moment when earnings and AI capex are already accelerating – and could that help extend the summer rally into a final, more speculative phase?

VISUAL 4 – LIQUIDITY / MACRO SCORECARD

Variable	Late 1999 / early 2000	August 2026
Liquidity catalyst	Formal Fed Y2K liquidity facility; LTCM aftermath	Treasury-led yen intervention via NY Fed; no broad Fed backstop announced
Policy rate direction	Fed tightening into 2000	Warsh Fed on hold; hike risk remains data-dependent
10-year / discount rate	Rising toward ~6.8%	~4.64% on Aug. 7
Oil / inflation	Oil rebounded sharply from 1998 lows	Oil remains elevated; energy still feeds inflation risk
Valuation / cash flow	Many tech names had little/no earnings; extreme P/Es	S&P ~20x forward; AI leaders highly profitable, but capex consuming more cash

CLOSE – STRONG EARNINGS, LATE-CYCLE LIQUIDITY, RESPECT THE RISK

So what do I think investors should take away?

First, our roadmap remains intact: earnings are strong, the summer rally is underway, and the AI capex cycle has not broken.

Second, the dot-com analog is getting later in the game. The historical comparison says late-cycle rallies can become faster, narrower, very profitable, and much more emotional.

Third, today is fundamentally healthier than 1q2000. Earnings, margins and cash flows are real, and valuations are nowhere near many of the extremes we saw at the dot-com top.

And fourth, keep watching the liquidity story. If the July yen intervention proves to be the beginning of broader policy support, the Y2K analogy gets more interesting. If not, if instead real rates, oil and the dollar rise together, that would be the warning sign that the cost of capital is beginning to overwhelm earnings growth.

For retirees and near-retirees, that is the discipline: participate in the earnings cycle, respect valuation, watch the cost of money, and do not confuse a great technology with a risk-free investment.

History does not repeat. But right now, it is rhyming loudly.

DATA NOTES / SOURCES

• FactSet Earnings Insight, August 7, 2026: Q2 earnings/revenue scorecard, growth, margins, forward estimates, valuation and bottom-up EPS (especially pp. 1, 3, 6-15, 31-34).
• Oak Harvest Stock Talk, April 2026 AI-vs-dot-com scripts: 7th-inning framework, agentic AI, private vs. public capital, capex and valuation discussion.
• Oak Harvest Stock Talk, June 2026 2H Market Outlook Preview: “Strong third quarter. More difficult fourth quarter,” earnings/rates/energy framework and expected volatility.
• Oak Harvest Stock Talk, Dec. 19, 2025 1H26 Outlook: stronger early 1Q, higher 2026 volatility, new-Fed-chair risk and dot-com overlay.
• Federal Reserve, July 20, 1999: Century Date Change Special Liquidity Facility; facility operated Oct. 1, 1999-Apr. 7, 2000.
• Federal Reserve speeches, Sept.-Oct. 1999: additional Y2K liquidity planning, repo term extensions and broader collateral.
• Financial Times / Associated Press, Aug. 2026: coordinated U.S.-Japan yen intervention; New York Fed execution on behalf of U.S. Treasury.
• Associated Press, Aug. 7, 2026 market close: S&P 500 7,757.64, weekly and YTD returns, 10-year Treasury around 4.64%.
• StatMuse historical Nasdaq data: Nasdaq Composite +85.1% from Oct. 21, 1999 through Mar. 10, 2000.