13F Filings, Q2 2026: Alphabet, Amazon and TSM Lead a Four-Quarter Accumulation Screen

Research note. SEC 13F-HR filings for the 18 managers that reported Q2 2026, computed from a frozen receipt on August 22, 2026. The article excludes two stale-quarter managers from every ranking. Source digest d0715be1def0. How we use AI.

Which stocks kept attracting capital quarter after quarter?

A one-quarter 13F screen treats a first purchase and the fourth consecutive increase as the same green dot. They are not the same signal. One says a manager is curious. The other says a manager has held a view for a year and kept paying a higher price to add to it, which is the more expensive thing to be wrong about.

Read the matrix by row rather than by column. A single strong quarter is common and says little. A row that stays green across all four is rare, and the streak column on the right names the managers who did the repeated buying.

What did the panel own and buy this quarter?

The single-quarter picture still matters; it just answers a smaller question than most coverage claims of it. Alphabet is the broadest material holding at 10 of 18 managers. Amazon follows at 6, and Taiwan Semiconductor at 5. The important correction is that consensus and conviction are not separate worlds: GOOGL, AMZN, TSM, SPCX, NVDA, V all appear in both the top-ten holder list and the 10% conviction set.

Where did independent managers both risk real portfolio damage?

Breadth can be shallow: ten managers can each own a token position. Shared conviction is the harder test. Only four companies had at least two current-quarter managers put 10% or more of their reported equity book into the same issuer. Alphabet and Amazon combine this with broad ownership and active buying. GE is different: only two qualifying managers own it, but both made it roughly a third of their portfolios.

Whose bet is actually the loudest?

The biggest percentage is not the boldest decision. A 20% holding is unremarkable in a four-stock fund and remarkable from a manager who normally spreads risk across seventy-five issuers. What separates the two is how far a position sits above the size that manager gives an ordinary idea, so the ranking below measures each holding against its own fund rather than against the others.

By that measure the loudest bet in the panel is Stanley Druckenmiller's Natera position: 16.6% of a book spread across 75 issuers, roughly 3.1× the size of an ordinary holding there and the largest position he reports. He increased it again this quarter.

Which stocks were actually new to the panel?

“New” is the most abused word in 13F coverage. It can mean a manager bought a stock after sitting out a quarter, a small old holding drifting above the reporting floor, or a private stake that became reportable the day it listed. Only the first is a decision; the other two are bookkeeping. A headline that calls all three “new positions” hands you three very different things as though they were one.

Three debuts attracted more than one qualifying manager: SPCX, AMAT, TTMI. By contrast, WBD and HD were opened by multiple managers this quarter but had appeared in earlier filings. Calling all five “new stocks” would erase the most useful distinction.

First appearance anywhere in the five-quarter archive

Re-entries that a one-quarter screen mislabels as debuts

SpaceX is a disclosure event before it is a buying signal

The ticker in these filings is SPCX, not SPXS. SpaceX began trading on June 12, 2026, and four managers reported it at quarter-end: Baron, Gerstner, Loeb, Tepper. Their combined reported weight was 41.6%, dominated by Ron Baron's 37.4% position. The SEC offering terms confirm the symbol and first trade date.

That does not prove they bought those stakes in the public market. A position held privately can become 13(f)-reportable when the security lists, while a manager can also buy more after listing. The filing does not separate those paths. The defensible statement is that four stakes became visible, not that four managers independently initiated fresh positions.

Did the panel really stop selling?

The raw position count says yes. Q2 produced 142 buy actions and 119 sell actions, a net +23 after three negative quarters. But one move in a 244-issuer analyzed book carries the same raw vote as one move in a four-issuer book. Equal-weight the managers and the apparent turn nearly disappears.

Only 6 managers were net buyers; 8 were net sellers and 4 were balanced. The median manager was exactly 0, and the average manager-level directional score remained -0.7%. My read is narrower than “the panel turned bullish”: a few active books bought enough lines to flip the aggregate count, while the typical manager did not.

A hypothetical ten-stock high-conviction portfolio

Everything above is an observation until it commits to something. This list ranks every qualifying holding on four things a reader can check later: how many managers own it, who bought it this quarter, who has kept buying it, and how large a bet it is inside its own fund. A company needs at least two of those working for it. SPCX is left out because the listing hides how much was actually bought, and IEP because Icahn owning his own company is not independent conviction.

The ten are equal-weighted at 10% each, because the evidence supports picking the names and not much more than that. The clock starts the first trading session after this note publishes, August 24, 2026, and the benchmark is SPY. Nothing is backdated to June 30, when none of these filings were public yet.

How we checked it

Frequently asked questions