From STAR 50 Rebound to STAR 50 Giveback, the Market Changed Its Mind in a Week

A Sharp Reversal Inside One Week: From Up to Down

On Monday, May 25, the STAR 50 closed strongly up 5.878%, reaching 1896.04. The story that day was clear: Nvidia (NVDA) had beaten on its 5/20 AI-chip earnings, resonating with the relay through the A-share AI-semiconductor chain that week. The market’s read was uniform - AI is not dead, the boom is intact.

By Friday, May 29, the STAR 50 closed at 1751.32, down about 7.6% from the 5/25 close, off 5.039% on the day, with turnover of 182.1 billion yuan. Over the same period:

One story split into two: call the first “AI leadership revival,” call the second “high-valuation rollback.” Which is right is not something this week’s data can decide.

This is the index-level view. From the same week and the same data, there was a separate pulse reading a single-session high-versus-low-beta split; here the scale shifts - the lens pulls back to watch one index, the STAR 50, travel from the top of a rally to the top of a selloff in a single week.

What the Other Side Sees

First voice: the 5/25 rebound was excessive to begin with.

After falling 3.7% on 5/21, the STAR 50 ran up 10.9% in a week (5/18-5/25). That kind of pace usually signals an oversold bounce rather than a confirmed trend. The four-day straight line from 1709 to 1896 was elastic enough to attract profit-taking. The giveback starting 5/26 was simply support failing at the highs - Nvidia’s earnings heat faded, chasing positions found no relay, and the decline naturally accelerated.

Second voice: this is not an AI problem, it is a broad high-valuation rollback.

The STAR 50’s 1500-plus at the start of the month was already the platform for this leg up; today’s 1750 is down 7.6% from 5/25’s 1896, yet still about +3% above the mid-May low (5/15 closed at 1696). ChiNext fell 2.112% while the SSE 50 closed +0.2365% - but read that contrast carefully: +0.24% is a noise-level move, hardly “catching inflows,” and more likely just the mechanical result of high beta falling while low beta does not. Any risk-off day produces this kind of split, and on its own it cannot prove “growth is being swapped for value.” I have no fund-flow data that confirms a style switch here; it may simply be the arithmetic of a beta difference.

Third voice: synchronization with US tech.

QQQ is up 11.6% on the month and still sits high even after a +3.33% over the past five days. But QQQ’s gain is distributed (mega-cap AI plus a compute-architecture mix), while the A-share gain is concentrated (a single AI-chip line). Once the A-share base consensus - “AI chips are the tightest and most expensive thing” - is falsified or shown to be merely a phase, the giveback runs faster.

Two Numbers Best Capture the Week

1. The turnover contrast. The STAR 50 traded 222.0 billion yuan on 5/25 and only 182.1 billion on 5/29, an 18% drop in single-day volume. People show up on the rebound and leave at the highs - a classic distribution-at-the-top signal.

2. The advance-decline split. On 5/25 the STAR 50 had 34 advancers to 16 decliners, a clear bullish edge; on 5/29 it flipped, just 8 up against 41 down, with most constituents falling with the index. But the STAR 50 has only about 50 constituents, so a broad decline when the index drops 5% on the day is a definitional outcome - that ratio on its own is not an exit signal independent of the index.

The Macro Backdrop Did Not Change; the Market’s Reading of It Did

The 10-year Treasury yield is at 445bp (as of 5/28), up just 9bp on the month, showing no clear pressure. The VIX is at 15.32, an extreme low. CPI is running about 3.8% year over year (still elevated but not out of control), and the Fed’s patient posture remains.

On the surface the macro mood should support growth stocks. But the fact is that what supports growth stocks is not the macro - it is their own valuation level and market sentiment. When valuation moves from “reasonable” to “high” and the market shifts from “chasing” to “taking profit,” favorable or neutral macro cannot rescue it. That is exactly what happened across the four days from 5/25 to 5/29.

The Takeaway

No forecast. Just both sides of one turning point on the record:

One side is the resilience of the rebound (5/25’s +5.878% did happen, and it did reflect AI-earnings resonance); the other is its fragility (four days later, the great majority of positions began to leave, with no relay at the highs).

May’s final trading days will decide whether this is “a short-term giveback inside a rebound” or “a turn at the tail of a rebound.” Breadth is the key tell - if the SSE 50 follows down, this is not a “growth versus value” rotation but a broad retreat in market risk appetite.

What to Watch

The following is an observation framework, not a trading signal.

• Whether the STAR 50 can hold 1700-1750. A further test down toward 1650 would say the week’s rebound was just “a rebound within a rebound,” with no base-building; stabilizing above 1700 may mean it is testing support.
• Whether the SSE 50 follows down. If value blue chips begin to fall in sync, this is no style switch but a market-wide release of high-valuation pressure - at +0.24% it is still inside the noise and gives no direction.
• Whether turnover repairs. If CSI 300 daily turnover stays below 900 billion yuan, overall market vitality is fading, not just a structural issue.
• The direction of US mega-cap tech (NVDA in particular). If QQQ or NVDA pull back in sync, the A-share STAR giveback upgrades from “valuation adjustment” to “shared pressure on global growth.”
• When the BSE 50 stabilizes. Small caps have fallen since 5/25 (-6.2% on the month); with no sign of stabilizing yet, that likely means money is still concentrating into STAR and squeezing out small caps - though no flow data can confirm it.

Data sources: Eastmoney Choice (A-share index quotes, query date 2026-05-31, covering the 5/25-5/29 range); Yahoo Finance daily bars (US QQQ, NVDA, query date 2026-05-30); FRED / BLS (10Y Treasury yield DGS10 as of 2026-05-28, CPI year-over-year on an April 2026 basis). This is a personal observation, not investment advice, and sets no price targets.