Institutional Filers Stopped Moving Through the Mega-Caps Together

A 45-day 13F window shows near-unanimous accumulation in Apple and Nvidia alongside contested two-way flow in Microsoft and Amazon — the block is decomposing into separate names.

Sterling scan summary

Institutional filers have stopped moving through the mega-caps as one block.

6
Checks run
3
Flagged
2
Convergent signals
Developing
Confidence
✗Insider buys✗Insider sells✓Congressional✗Sector price✓13F✓News

Why Sterling flagged it

  • 13F rotation flagged 14 names with six or more filers moving the same direction over the 45-day window; among the six largest by reported value the splits are not uniform — AAPL 10 adders / 1 reducer ($90.2bn reported), NVDA 12 / 2 ($30.9bn), TSLA 7 / 1 ($17.9bn) against MSFT 9 / 6 ($29.9bn) and AMZN 9 / 7 ($19.8bn).
  • The dispersion runs a clean gradient: adders are 91% of moving filers at Apple and 86% at Nvidia, but 60% at Microsoft and 56% at Amazon — Amazon is close to a two-way market among the filers that moved at all.
  • News pressure concentration flagged 215 tickers clearing six articles in 48 hours, but attention is lopsided: NVDA drew 208 articles, more than three times the next ticker (GTLL, 62) and over four times WMT and STLA (47 each). GOOG drew 44; MSFT, AMZN and AAPL cleared no top-sample slot.
  • Congressional disclosures flagged two qualifying clusters, both buys, and neither in mega-cap tech — JPM (3 members, $48,501.50 combined mid) and ADBE (3 members, $24,001.50). Political disclosure is not tracking the same names.
  • No corporate-insider cluster corroborates any of this: the 7-day Form 4 window produced zero qualifying buy clusters and zero sell clusters, and the sector-breadth check was not evaluated on a historical scan — so the price dimension is unsourced here.

How Sterling got here

  1. Observation

    The 13F check flagged 14 names with at least six filers moving the same direction inside 45 days. The six largest by reported value do not share a common posture: adders account for 91% of moving filers at Apple (10 of 11) and 86% at Nvidia (12 of 14), but only 60% at Microsoft (9 of 15) and 56% at Amazon (9 of 16).

  2. Cross-check

    News pressure gives an independent read on where the narrative sits. Of 215 tickers clearing six articles in 48 hours, NVDA alone accounts for 208 articles — more than triple the second-place ticker. The bundle's own market wrap notes a blowout Nvidia quarter alongside rising investor questions about valuations across semiconductors and AI infrastructure. Attention and filer conviction are concentrated in the same name.

  3. Independent confirmation

    Disclosure data complicates rather than confirms. The two qualifying congressional clusters are both buys and both outside the complex — JPM at $48,501.50 combined mid across three members and ADBE at $24,001.50 across three. Corporate insiders add nothing: zero qualifying Form 4 clusters in either direction over seven days.

  4. Inference

    Taken together, the evidence would be consistent with the mega-cap block being repriced name by name rather than as a single exposure, with the AI-capex payers drawing the contested flow and the AI-capex recipients drawing the one-sided flow. It is not yet evidence of net selling anywhere: every one of the six sampled names has more adders than reducers.

  5. Status

    What is not corroborated: no insider activity, no evaluated sector-breadth data, and no dollar-weighting behind the filer counts. The primary reading rests on two flagged sources, one of which is structurally lagged. This is a configuration to monitor, not a conclusion.

What data Sterling analyzed

Six deterministic checks ran against the 22 May window; three flagged.

CheckSourcesResultDetail
13F institutional rotation13F filings, 45-day windowFLAGGED14 names with ≥6 filers moving the same direction; top-sample splits range from AAPL 10:1 to AMZN 9:7.
News pressure concentrationScored news feed, 48-hour windowFLAGGED215 tickers with ≥6 articles in 48h; NVDA at 208 versus 62 for the next-highest.
Congressional trade clusteringSTOCK Act disclosures, 14-day windowFLAGGED2 symbol/direction pairs traded by ≥3 members — JPM buy ($48,501.50 mid) and ADBE buy ($24,001.50 mid).
Insider buy clusteringForm 4 filings, 7-day windowCLEARNo qualifying buy cluster.
Insider sell clusteringForm 4 filings, 7-day windowCLEARNo qualifying sell cluster.
Unusual sector moveDaily sector breadth (live sessions only)CLEARNot evaluated for historical scans.

What would confirm

  • 27–28 May AI-infrastructure prints diverge — MRVL (27 May, $0.798 est.), SNOW (27 May, $0.3193) and DELL (28 May, $2.96) reacting differently from one another rather than moving as a bloc would corroborate that the complex is being priced name by name.
  • Next 13F window widens the gap — reducer counts rising above adders at Microsoft or Amazon while Apple and Nvidia stay above 85% adders would escalate this from dispersion to genuine divergence.
  • Attention broadens off Nvidia — the news check showing a second mega-cap approaching NVDA's 208-article magnitude would indicate the narrative is splitting alongside the flow.
  • Disclosure follows the split — a congressional or insider cluster appearing in any single mega-cap name, in either direction, would add the third independent source the scan currently lacks.

What would invalidate

  • Filer counts converge — the next 13F window showing adder shares clustering within a narrow band across all six names would reduce this to sampling noise in small filer counts.
  • Dollar-weighted flow proves synchronized — evidence that Microsoft's and Amazon's six and seven reducers are immaterial by position size would support the mechanical explanation over the thematic one.
  • Macro dominates the tape — a Core PCE surprise on 28 May (0.3% est.) or a CB Consumer Confidence miss on 26 May (91.9 est. vs 93.8 prior) moving the complex uniformly would show the block still trades as one exposure.
  • 27 May CRM print resets the group together — a single result ($3.13 est.) that re-synchronizes software and infrastructure sentiment would dissolve the separation this scan detects.

Portfolio impact map

Typically supported
    Typically pressured

      Typically supported
        Typically pressured
          Exposure sensitivities describe how asset classes have typically behaved under these conditions — they are not recommendations.
          Developingsignal strength — Two independent flagged checks point the same way — 13F dispersion and single-name attention concentration — but the 13F source is lagged and filer counts are small.
          This configuration is under observation. A full Market Intelligence report will follow if additional confirmation emerges.
          Wall St. Intel Research is published for informational purposes only and is not investment advice, an offer, or a solicitation. Research is produced by Sterling, an AI system, and reviewed by Wall St. Intel before publication. Data as of the dates indicated. Investing involves risk, including loss of principal.