Interest rate hikes are not the tech killer; EPS is: the strategy of removing the weak and retaining the strong after the big drop in AI mainline.

Author: BIT

Investment Summary

My conclusion is simple: the true end of the tech rally isn't the Fed raising rates by another 25 basis points, but industry involution and EPS falsification; before these two signals appear, a sharp decline like June 5th is more like “reversing to pick up people,” not “car wreck and loss of life.” This sentence is the main thread of this report and also my position principle for handling this round of rate hike panic. U.S. non-farm payrolls added 172k jobs in May, significantly higher than the market expectation of 88k, pushing the probability of rate hikes this year up to 63%, and approaching 100% before January next year. The Philadelphia Semiconductor Index fell over 10% that day, and the Nasdaq dropped 4.18%. But I won't abandon the tech mainline just because of a macro data day, because historically, what truly determines whether tech stocks can withstand rate disturbances is never the rate itself, but whether EPS continues to be upwardly revised. [1] [2]

My judgment is that AI trading has now shifted from “broad rally narrative” to “shrinking circle validation.” This is not a stage where you can continue indiscriminately buying all high-beta tech stocks, but it’s also not the end of the AI mainline. The core holdings should be large-cap assets with high order visibility, stable gross margins, strong cash flow quality, and EPS still being upwardly revised by analysts; for quantum, aerospace, and some small chip stories lacking profit cycle closure, it’s better to reduce positions during rebounds or hedge portfolio volatility with options structures.

1. Factual Judgment: Rate Hike Panic is a Trigger, Not the Main Cause

The market reaction on June 5 was very intense, but the trigger chain was not complicated. Official BLS data show that U.S. non-farm employment increased by 172k in May, with the unemployment rate holding at 4.3%, and employment for March and April revised upward by 93k; strong employment data reinforced market concerns about inflation stickiness and further rate hikes. [3] Reuters and market reports indicated that the Nasdaq fell 4.18% that day, the Philadelphia Semiconductor Index (SOX) dropped over 10%, and investors quickly re-priced risk assets to scenarios of “higher and longer-lasting interest rates.” [1] [2]

Variables

Changes around June 5

My interpretation

U.S. May non-farm employment

172k, significantly above the 88k expectation

Interest rates re-priced in the short term, but employment structure still depends on wages and inflation transmission

Unemployment rate

4.3%, unchanged from previous

Labor market shows no signs of recessionary deterioration, even suppressing rate cut expectations

Probability of rate hikes this year

Market raised to 63%

Creates valuation pressure on long-duration assets, first hitting high-valuation tech

SOX Index

Drop over 10% in a single day

Semiconductors are a crowded trade core, first impacted by liquidity shocks

Nasdaq

Down 4.18% in a single day

Market panic evident at index level, but not equivalent to EPS being falsified for AI

I define this decline as a concentrated release of rate shock combined with crowded trading. It will wipe out some over-expanding valuations and force funds to withdraw from weak logic, highly elastic, and low-profitability certainty assets; but if AI infrastructure orders, cloud vendor capital expenditures, GPU/light module/PCB demand do not substantially downgrade, the tech mainline has not ended because of this day’s decline.

2. Historical Review: The Lesson from 1999 is Not “Chasing the Bubble,” but “Watching EPS”

The 1999 dot-com bubble period is often cited to warn about today’s tech stocks, but I believe this analogy should not focus solely on valuation; profitability must also be considered. At that time, the Fed was in a continuous rate hike cycle, the Dow was mostly sideways, but the Nasdaq continued to rise sharply until March 2000 peak. An institutional study shared by Moomoo notes that in 1999, the Nasdaq 100 EPS grew about 60%, while the Dow’s EPS growth lagged significantly; by Q1 2026, Nasdaq 100 EPS grew about 36%, the Dow about 4%, showing a renewed profit divergence. [2]

Research by Nasdaq Investment Intelligence on rate hikes over the past 30+ years also supports this conclusion. During 13 rate hike phases lasting at least 6 months from 1985 to 2021, Nasdaq-100’s average total return was 22.6%, higher than S&P 500’s 11.3% and Dow’s 12.7%; during the 10-year yield rising phase from October 1998 to January 2000, with about a 2.2 percentage point increase, Nasdaq-100 gained 165.3%, significantly outperforming the S&P 500 and Dow in the same period. [4]

This history teaches me that the lesson isn’t “high valuation can always rise,” but that rising rates are not sufficient reasons to sell tech stocks. The real dangers are twofold: first, stock prices rely solely on PE expansion without EPS growth; second, industry competitive dynamics worsen, with leading companies’ gross margins and cash flows turning downward first. If neither occurs, rate hikes are more about adjusting pace rather than declaring the mainline dead.

3. Valuation Framework: Short-term View of Prosperity, Long-term View of Margin of Safety

I oppose judging AI leaders’ bubbles solely by a single PE or PB percentile. In the short-term (one year), stock prices are mainly driven by revenue growth, ROE change rate, and EPS revision direction; over three to five years, PB, free cash flow yield, and capital return cycles will truly determine long-term returns. Pacer ETFs’ Nasdaq-100 study shows that at the end of 1999, Nasdaq-100 was about 73x earnings with a free cash flow yield of only 0.76%; by the end of 2023, it was about 31x earnings with a free cash flow yield of 2.68%. The current sales, profit, and free cash flow scales of leading companies are also incomparable to those of internet bubble companies in 1999. [5]

Valuation misconceptions

Incorrect view

My judgment method

Is high PB necessarily a bubble?

Any high PB equals bubble

Focus on ROE, technological barriers, capital expenditure returns, and EPS revision trends

Is high PE necessarily a sell?

High PE percentile means reduce positions

If EPS keeps upward revisions, high PE can be absorbed by profits; if EPS stagnates, high PE is dangerous

Are AI leaders too crowded?

Crowded equals top out

Crowded more resembles a shrinking circle signal, with funds moving from weak assets to strong assets

How to handle rising rates?

Sell all tech

Reduce leverage, cut weak assets, keep bottom holdings with strongest EPS evidence

Therefore, I categorize core AI assets into two types. The first are “toll station assets” with real orders, real gross margins, and real cash flows, including AI server chains, advanced packaging, optical modules, PCBs, and key cloud infrastructure suppliers. The second are high-beta assets with only long-term stories and fuzzy profit paths, such as some quantum, aerospace, conceptual chips, and software stocks lacking order verification. The former should be observed for accumulation opportunities during sharp declines; the latter should be reduced during rebounds to lower risk exposure.

4. Crowding Level: This is a Shrinking Circle, Not the First Top

Currently, funds are clustering around core AI assets, siphoning liquidity from dividends, small caps, and non-mainline assets, which must be acknowledged. But crowding itself does not equal a top. True tops usually require three conditions: first, industry capital expenditure shows marginal slowdown; second, leading companies’ competitive landscape deteriorates, with price wars or gross margin declines beginning; third, EPS upward revision trend halts or turns downward. So far, this correction aligns more with “left-side phase of high peaks and lows” and “mainline shrinking circle,” rather than confirming the first mid-term top of AI. [2]

I view the late June to July earnings season as the real validation window. A-shares interim reports, U.S. tech companies’ Q2 guidance, cloud vendor capital expenditure data, and semiconductor supply chain order visibility will jointly determine whether this correction is a healthy rotation or the mainline starting to falsify profits.

5. My Investment Interpretation: Keep the Tech Mainline, but Focus on the Strong

My operational principle: keep core holdings with EPS evidence, and no longer waste risk budget on high-beta story stocks. In AI infrastructure, I prefer companies with high order visibility, stable gross margins, good cash flow, and in the critical client capital expenditure cycle. Optical modules, PCBs, AI servers, advanced packaging, cloud infrastructure, and software platforms with bargaining power are my preferred directions to withstand volatility.

Asset categories

Current actions

Core reasons

Risk control

AI infrastructure leaders

Maintain core holdings, observe during sharp declines

EPS and order evidence still present, short-term rate shocks do not alter industry trend

Avoid chasing highs, wait for key nodes to materialize

Optical modules / PCBs / advanced packaging

Maintain core focus

Most direct transmission of AI server capital expenditure to hardware chain

Reduce immediately if margins or orders downgrade

Cloud and platform software

Select strong ecosystem moat players

AI application entry points and enterprise refresh cycles still hold long-term value

Prevent valuation from over-accelerating

Quantum / aerospace / some conceptual chips

Reduce during rebounds

Narratives are strong, EPS weak, most vulnerable to valuation cuts during rate hikes

Use options to hedge high-beta exposure

Dividend and cash assets

As portfolio stabilizers

Hedge macro uncertainty

Do not treat defensive holdings as long-term mainline

This is not blind optimism. On the contrary, I believe the next month requires stricter monitoring of four key nodes: June 10 CPI release, if core inflation driven by oil prices exceeds expectations, reduce leverage; oil prices and US-Iran tensions, if oil remains high long-term, inflation will be sticky; mid-June ECB and BOJ meetings, which will influence global liquidity; June 18 Wosh’s statement, if hawkish, will reshape rate path pricing. Macro nodes determine the rhythm, EPS determines the direction.

6. Conclusion: Reversing to Pick Up People, Only Picking Those with Earnings

I won’t abandon the tech mainline because of the single-day plunge on June 5, but I will upgrade the portfolio from “buying AI stories” to “buying AI profit statements.” If a company can prove continuous order flow, gross margins, cash flow, and EPS, its decline during rate shocks is more like an opportunity; if a company only has concepts without profit paths, it should be cut during rebounds.

The final conclusion remains the same as at the start: the end of the tech rally is caused by industry involution and EPS falsification, not the Fed raising rates by 25bp. The current adjustment is “reversing to pick up people,” not “car wreck and loss of life.” Keep positions with actual performance, wait for the four key nodes to land.

This report is prepared by a designated analyst. The views expressed herein are solely those of the author and do not represent BIT platform’s opinions. This material is for reference only and does not constitute investment advice.