Investigation — Aschenbrenner essay vs 13F scorecard — what the Situational Awareness predictions have actually done, and the new megacap put overlay

Investigation Ticker Tape

Question: Aschenbrenner essay vs 13F scorecard — what the Situational Awareness predictions have actually done, and the new megacap put overlay Verdict: open

Living scorecard. Started 2026-06-05 when the Situational Awareness essay (June 2024) was archived as primary research. Tracks the essay's predictions and Aschenbrenner's 13F trades against what the tape has actually done. Next update: Q2 2026 13F (quarter ends 2026-06-30, files by ~2026-08-15) — that filing tells us whether he held, added to, or covered the megacap put book after it ran against him this quarter.

Now Part 1 of the Aschenbrenner research dossier (Parts 2–5: essay-mined lanes, the put book, track-record calibration, cross-manager convergence).

What we're asking

The essay is the thesis; the 13F is the trade; the tape is the verdict. Now that the 2024 primary source is on file, score it three ways:

  1. Essay → reality — which 2024 predictions have come to fruition?
  2. 13F track record — what he bought/sold, and did it work?
  3. The new put overlay — the user's prompt: "he's shorting a lot of the tickers now." Is that true, on what, and has it worked?

Price truth throughout: the scan summaries, latest 2026-06-04. 2y% ≈ since the essay (June 2024); ~3m% brackets his Q1-2026 (put-entry) quarter-end. 13F structure from the corrected all-action 13F read (raw EDGAR salp13fq1xml.xml, data as of 2026-03-31).

The one-line answer

The essay's core call — power/infra is the bottleneck, not silicon — has been massively vindicated, and his long book is the proof. But he has now flipped from the essay's implicit compute-bullishness to an $8.46B notional put book against the consensus AI mega-caps — and so far that short/hedge leg is dead wrong: every name he put is up double-to-triple digits since he entered, most near 52-week highs. He is long the bottlenecks and short the picks everyone else owns. Leg one paid. Leg two hasn't — yet.

Scorecard A — essay predictions vs reality (as of 2026-06-04)

Essay (June 2024) claim Status Evidence
Power, not silicon, is the binding constraint on AI capex come to fruition His entire infra long book re-rated on exactly this — BE +1863% 2y, the miner-pivot cohort +184–894% 2y (table below). This is the prediction that built the ai-power watchlist.
"Trillion-dollar clusters… every six months another zero" / $1T capex annually by ~2027 🟡 on track Hyperscaler capex still ramping into 2026; not yet falsified. ~2027/2030 horizon — pending, not provable now.
"By 2025/26 these machines will outpace college graduates" 🟡 mixed / contested Capability gains real; "outpace college grads" is not cleanly settled. Track, don't score.
100 GW / >20% of US electricity / $1T single cluster by ~2030 pending A 2030 end-state. Still the north star of the lane; unfalsified.
"Nvidia analysts still think 2024 might be close to the peak" (implicitly NVDA-bullish) ⚠️ he has since reversed NVDA is +78% 2y / +17% ~3m — the 2024 bull call aged fine, but by Q1 2026 he is short NVDA (see Scorecard C). The essay's compute-bull framing is the leg he himself walked away from.

Scorecard B — his 13F book vs the tape

Longs (infra equity — ~91% of his Q1 equity book): the thesis paid

Ticker Price RSI ~3m% 1y% 2y%≈essay from 52wH Read
BE $291.26 56 +76.8 +1237 +1863 −9.8% The thesis stock. His #1 ($879M, 23% of equity book).
SNDK $1831.50 74 +205.7 +4578 n/a* −1.6% His #2 ($724M) + a $389M call on top. Memory leg ripping.
IREN $65.48 61 +49.4 +572 +522 −14.8% Miner-pivot, added +$72M Q1.
CORZ $28.93 73 +82.6 +137 +272 −5.0% Miner-pivot core, kept.
APLD $44.71 55 +56.1 +223 +894 −11.9% Added +$42M Q1.
RIOT $27.65 69 +67.3 +181 +184 −4.5% Added +$64M Q1.
CLSK $17.61 67 +65.2 +80 +13 −25.4% Added +$88M Q1; weakest 2y, strong recent.
CRWV $110.93 51 +39.5 −21 n/a* −40.7% The lone laggard. Rotated calls→equity Q1; still −41% off high.

*SNDK/CRWV lack a clean 2y (recent spin/IPO). Verdict: 7 of 8 longs strong-up, most multi-bagger since the essay. CRWV is the one soft spot (−21% 1y, deep below its high) — the levered-to-equity rotation he made there hasn't paid yet.

Exits (100% liquidated Q1): he sold winners early

Ticker ~3m% after exit Read
LITE +37.8 His largest single exit ($479M, optical). Kept running; +1948% 2y. Early.
COHR +51.9 Second optical exit ($89M). Ran after he left. Early.
CIFR +63.6 Full miner exit ($155M) — then ran +64%. Looks like a mistake so far.
HUT +142.2 Exited ($40M); Loeb bought it same quarter. HUT then ran +142%. Loeb right, Aschenbrenner wrong (so far).
TSEM +116.9 Small-cap semi exit ($85M); ran +117%. Early.
EQT −10.9 The one clean exit. Gas pure-play, now down/−20% off high. Correct — consistent with the thesis evolving from "gas-as-fuel" to "compute-as-bottleneck."

Pattern: 5 of 6 exits kept rising 38–142% after he left; only EQT was right. His Q1 housecleaning trimmed names that kept working — a recurring failure mode worth flagging when reading his next filing.

Scorecard C — "he's shorting a lot of the tickers now": TRUE, with two big caveats

Yes. The Q1 2026 13F added 11 put rows, $8.46B underlying notional — 2.19× his equity book (R19). This was NEW in Q1; his Q4 2025 book was essentially unhedged long. But two things the raw "he's shorting" framing misses:

Caveat 1 — he's shorting the mega-caps, not his own cohort. Every put is on a consensus AI name or the semi ETF — not on BE/SNDK/IREN/etc. It's a pair trade: long the bottlenecks, short the picks everyone owns. This is not a retreat from the power thesis; it's the same thesis pushed to its logical short side ("the mega-caps are over-valued relative to the infra they depend on").

Caveat 2 — so far the short leg is dead wrong. Every put underlying is up since he entered, most near 52-week highs:

Put (notional) ~3m% since entry 30d% from 52wH Working?
SMH $2.04B (biggest) +59.8 +25.9 −0.8% (at highs)
NVDA $1.57B +17.3 +8.2 −9.2%
ORCL $1.07B +51.7 +27.8 −33.4% ❌ (but −33% off its high — only put-name well below highs)
AVGO $1.0B +51.2 +15.1 −3.2%
AMD $969M +168.5 +58.9 −0.7% ❌❌
MU $584M (also a $422M call) +153.9 +76.5 −6.6% ❌ (but vol/spread, not clean short)
TSM $535M +22.5 +8.7 −3.0%
ASML $494M +23.6 +24.5 −1.0%
INTC $159M (Q4 call→Q1 put flip) +147.3 +17.7 −15.1% ❌❌
GLW $21M +38.8 +25.7 −5.2%

The hardest tell: INTC. Q4 2025 he held an INTC call ($747M, bullish); Q1 2026 he flipped to an INTC put + ran straight into +147%. A clean, badly-timed reversal.

The honest qualifier (from R19, do not drop it): the 13F gives no strike, expiry, or premium, and the snapshot is now ~65 days stale. If these are deep-OTM tail hedges, the capital loss is small (cheap insurance that didn't pay) even though every underlying ran. If they're at-the-money directional shorts, the book is deeply underwater. We cannot tell P&L from a 13F — only that the underlyings moved hard against the put direction. Treat "hasn't worked" as a statement about the tape, not a measured loss.

What this means for our perspectives (no key_ticker changes — observation only)

  • ai-power-bottleneck — the long book is the perspective spine and it's validated; nothing to change. New texture: the operator of the thesis now expresses it as a pair trade, hedging/​shorting the consensus beta. That's the R19 "AI mega-cap mean-reversion" candidate, still unspawned — this scorecard is more evidence for it, not a decision to open it.
  • optical-supercycle — his LITE + COHR exits both kept running (+38% / +52%). Confirms R19's "smart-money split, not converging" read: he left optical right as it kept working. A negative positioning vote that the tape has so far rejected.
  • memory-supercycleSNDK +206% ~3m behind his #2-position-plus-call is the cleanest win in the book; hardens the memory leg. MU is a put+call vol structure, not a clean vote.

Verdict + reasoning

Verdict: open (living). The scorecard's first reading:

  • Essay core thesis: vindicated. Power-as-bottleneck built the lane and paid.
  • Long book: working (7/8 strong-up, CRWV the laggard).
  • Exits: poorly timed (5/6 kept running; only EQT right).
  • New short/put book: the open question. Directionally it's the essay's thesis taken to its short side; on the tape it's been wrong every name; on P&L it's unknowable from the 13F. This is the single most important thing to resolve, and only the Q2 2026 13F (~Aug 15) can — did he hold conviction, add, or cover?

Reasoning stays inside the producer lane: this is positioning + tape observation, not a buy/sell call and not a key_ticker mutation. We are not trading off a 65-day-stale options snapshot.

Caveats / what would change this

  • 13F is a 45+ day delayed, partial snapshot (no cash, swaps, futures, non-US, no option Greeks). Positioning context, not live exposure.
  • The put book's P&L is indeterminate without strike/expiry/premium (R19 "Unresolved").
  • If Q2 2026 shows the puts covered, the "dead wrong short" reading flips to "tactical hedge, closed" — a very different signal. If added into the loss, it's conviction.
  • ORCL ($1.07B put): +51.7% ~3m / +28% 30d but −33% off its 52-week high (only put-name materially below highs) — the one name where a mean-reversion put has a partial directional case, though the recent tape still ran against it.

Price-truth note

All % and price figures are from the scan summaries, latest 2026-06-04; 2y≈essay window, ~3m≈since Q1-2026 close. 13F dollar magnitudes are filing-reported underlying notional as of 2026-03-31, not live-price claims. No return/P&L was computed in prose — figures are read directly from the code-computed summaries.