Aschenbrenner Q1 2026, day 2: the longs broke first

Yesterday's piece covered the Q1 2026 13F filing from Situational Awareness LP — Leopold Aschenbrenner's hedge fund, $13.68B notional, 62% short the chip stack ($8.46B puts) and 28% long the picks-and-shovels ($3.86B cash long, $1.36B calls).

The structural read of that filing: chips are overpriced, infrastructure is underpriced. Short the consensus AI exposure, long the bottlenecks. 48 hours of post-filing tape now lets us check whether the trade is actually moving the way the thesis says it should.

It's not.

Both sides of his book are red. The shorts are working — modestly. The longs are bleeding harder than the shorts. The structural takeaway is real and worth pinning before the narrative gets re-written.

Today's bubble rotation tells the whole story

QuantAbundancia tracks 12 validated bubbles and aggregates daily moves at the cluster level. Today (2026-05-19):

 +3.28%  Enterprise SaaS AI              ← only green
 -0.17%  AI-Capex Hyperscalers
 -1.76%  Cooling / DC Infra
 -1.91%  Quantum
 -1.98%  Datacenter Power
 -2.13%  Semi Equipment / Litho
 -2.47%  AI Compute Accelerators         ← his shorts
 -2.74%  Nuclear / SMR
 -2.97%  Space / Sat Comms
 -3.19%  Networking / Optical            ← his longs (CRWV, optical)
 -3.62%  DRAM / HBM Memory               ← his shorts AND longs (MU + SNDK)

One green bubble, eleven red. And the only green one is the one bubble Aschenbrenner has zero exposure to: AI-enabled software (Enterprise SaaS AI). The capital that exited AI hardware today didn't go to cash — it rotated into ServiceNow, Salesforce, Palantir-style names.

This is a "rotation within AI", not a "AI is over" tape. Which is good for the structural thesis but bad for the specific positioning.

Position scorecard: today + 5-day

The chip-short stack (the $8.5B he's short via puts) — underlying prices today and over the trading week since the 13F dropped:

| Position | Notional | Today | 5-day | Comment | | --- | --- | --- | --- | --- | | $SMH put | $2.04B | -2.04% | -7.3% | The ETF basket short — primary expression | | $NVDA put | $1.57B | -1.55% | -7.8% | Keystone short | | $ORCL put | $1.07B | -1.22% | -6.2% | Hyperscaler software short | | $AVGO put | $1.01B | -1.11% | -5.7% | Custom-ASIC short | | $AMD put | $969M | -2.14% | -7.6% | MI300/MI350 short | | $MU put (straddle leg) | $584M | -3.76% | -14.8% | Working hard on the put side | | $TSM put (straddle leg) | $535M | -1.07% | -6.3% | Quieter than the rest | | $ASML put | $494M | -1.61% | -8.4% | EUV-monopoly short | | $INTC put | $159M | -1.78% | -7.1% | Reversed Q4 long → put |

Chip-short median: -7% over the week. The structure is doing what it's supposed to do. The 13F dropped, the chip stack sold off, the puts gained value.

Now the picks-and-shovels long side (his $3.86B cash long + $1.36B calls):

| Position | Notional | Today | 5-day | Comment | | --- | --- | --- | --- | --- | | $BE | $879M long + $55M calls | -3.66% | -16.9% | The anchor of the book — biggest dollar swing | | $SNDK | $724M long + $389M calls | -1.99% | -5.3% | NAND held best across the book | | $CRWV | $556M long + $141M calls | -5.44% | -13.9% | Neocloud / GPU-as-service hit hard | | $IREN | $401M | -3.85% | -17.7% | Miner-to-HPC anchor — visible drawdown | | $CORZ | $389M (9.4% activist stake) | -2.72% | -9.6% | Held best of the miners (activist lockup) | | $APLD | $320M | -4.18% | -18.8% | Worst of the activist-tier names | | $RIOT | $142M | -2.71% | -7.6% | Standard miner-basket bleed | | $CLSK | $104M | -3.00% | -4.9% | Was the +513% Q1 winner — paying back | | $TE | $44M | -12.11% | +14.4% | New position, post-Friday rally | | $SHAZ | $18M | -7.53% | -6.0% | IPO-cohort bleed | | $HIVE | $6M | -6.42% | +13.5% | Spiked Monday, crashed Tuesday |

Picks-and-shovels median: -10% over the week (worst names -17 to -19%). The longs are bleeding harder than the shorts in % terms.

Why the longs are bleeding faster

This is the structural insight worth pinning. The chip names he's shorting are mostly mega-caps ($1T+ NVDA, $700B AVGO, $1T TSM, $500B ASML, $500B MU). The picks-and-shovels longs are mostly mid + small caps — BE $25B, IREN $16B, CRWV $30B, CORZ $7B, APLD $4B, miners $1-3B each.

When broad AI risk-off hits:

1. Mega-cap chips absorb selling at fractions of a percent of float. $NVDA at -1.55% on $5T+ float = ~$77B of selling. That's a normal day's volume.
2. Mid + small-cap infrastructure has 10-100× higher beta to AI sentiment. $IREN at -3.85% on a much thinner float means much smaller dollars moving the same percentage. And the float-rotation effect compounds.
3. The picks-and-shovels universe is the marginal beneficiary of incremental AI dollars. When dollars come in, they go disproportionately to small-cap infrastructure (the +319% YTD LPKF-style story). When dollars leave, they leave from the same place.

So a paper short on $1.57B of NVDA notional, even fully delta-hedged, doesn't fully offset a paper long on $401M of IREN when IREN moves 3-5× harder on the same risk-off catalyst. The structural pair is set up to capture long-cycle re-rating, not short-term volatility.

That's exactly what the thesis says. The trade is a multi-quarter bet on relative re-rating: chips compress, picks-and-shovels expand. In a single 1-week window where AI broadly de-rates, the picks-and-shovels can drop more than the chips — and that's not a thesis violation. It's the cost of being earlier on the curve.

What the Enterprise SaaS bubble tells you

The +3.28% green-only bubble today (Enterprise SaaS AI: a small cluster of $NOW, $CRM, $PLTR-tier names that benefit from AI deployment downstream) is the real signal.

This isn't a "the AI trade is dead" tape. It's a "the AI hardware build is too crowded, rotate to AI users" tape. Which is exactly the second-derivative of Aschenbrenner's thesis:

• Direct read: chips overpriced, infrastructure underpriced
• Second-derivative read: anything adjacent to AI that has hardware exposure is overpriced; anything that consumes AI downstream is underpriced

His book captures the first read directly (short chips). The picks-and-shovels long is a bet that some hardware (the bottleneck-tier picks-and-shovels) gets re-priced UP, not the consumer-tier hardware. Today's tape is consistent with the first half (chips down) but doesn't show the long side working yet (picks-and-shovels also down, just slower).

Thesis check

Not broken. But the drawdown is the visible cost.

What would break the thesis:

• BE / SNDK / CRWV / IREN underperforming the chips over 90 days. They're underperforming today; not yet over 90 days.
• The MSFT-IREN-Sweetwater capacity ramp slipping materially (would invalidate the picks-and-shovels-have-real-revenue framing).
• Aschenbrenner unwinding the puts or trimming the longs in the Q2 2026 13F (filing ~Aug 14, 2026 — that's the next data point).

What would confirm the thesis:

• Chips continue to compress while the long-side miner basket stabilizes and resumes the uptrend (Sep-Dec 2026 timeframe).
• AI-utility names (Enterprise SaaS AI bubble) re-rate as the output of all this infrastructure becomes monetizable.
• A second activist-tier accumulation move (CORZ was the 13D filed at 9.4%; if a similar move shows up in $IREN or $APLD next quarter, that's confirmation).

What to watch over the next 30 days

The next data points that matter:

1. $IREN Q4 print (already done 2026-05-07) and Sweetwater utilization commentary — operational confirmation that the $10B MSFT contract is generating GPU-hours. Without that, the long thesis is buying time.
2. NVDA earnings (late August) — if guidance compresses on hyperscaler capex digestion, the chip-short thesis gets external confirmation.
3. The QA composite sentiment chip at /today — today registered "Greed" pre-selloff; tonight's recompute likely flips to Neutral or Fear. If it stays in Fear for a sustained period, that's a different regime than the +319% LPKF / +14× IQE narrative-driven market we've been in.
4. Q2 2026 13F filing window (~Aug 14) — the next time we get a fresh snapshot of Situational Awareness LP's book. If chip shorts are reduced and picks-and-shovels longs are doubled, the conviction is even higher than today. If both are trimmed, the bet didn't survive the drawdown.

The structural takeaway

Aschenbrenner's Q1 2026 trade is the most coherent expression of a "rotation within AI" thesis any public fund has put on. 48 hours of post-filing tape is too short to validate or invalidate it — but it's long enough to mark the cost.

Today, the cost is visible: the picks-and-shovels broke first. The shorts are working, just not enough to fully offset. The thesis says that gap closes over 6-12 months. The tape says the gap is widening for now.

Watch the bubble rotation map at /correlation — that's the lens. When Datacenter Power and Memory start outperforming Semiconductors on a 252-day basis, the thesis is working. Until then, it's a drawdown.

Full position-level data refreshed nightly at /portfolios/situational-awareness. Yesterday's full thesis breakdown at /articles/aschenbrenner-q1-2026-chip-short.

---

Disclosure: QuantAbundancia tracks fund portfolios for editorial purposes only. The Aschenbrenner fund and its principals have no relationship with QuantAbundancia. None of the above is investment advice; the 13F documents publicly-disclosed long-side positions only — short and derivative books that aren't in the form of long-leg puts are not reported and may materially change the read.