Jordi Visser — AI Crash Over / What Comes Next Investment Thesis

Source: The AI Crash Is Over. Here's What Comes Next, Jordi Visser (YouTube), August 9, 2026.

The Framework: AI Compresses Time — Hedge Crowding, Own Scarcity

Visser's post–July speed-crash playbook: the positioning crash is over, but the structure that produced it is permanent. AI compresses economic time (digital employees replace hiring) and market time (crowded factors crash faster). Index puts fail when factor vol decouples from index vol. The investable response is mechanical: stay long the AI thematic / scarcity stack, cut when rate-of-change and breadth warn, and hold scarce hedges that absorb both the government and AI "printing presses."

Signal	What It Means	Action
Factor vol >> index vol	Gross leverage / long-short stress, not S&P crash	Do not hedge AI with index puts
50-day ROC on 100-name AI index > ~20%	Crowding risk rising (July peak ~50%)	Tighten risk; watch breadth
Breadth (% above 20-day) rolling over	Speed-crash warning inside the bull	Rotate to hedges; keep powder for the puke
Compute / memory scarcity data accelerating	Demand thesis intact after flush	Re-own infra; do not fade the bounce

"I believe the market is changing structurally, and this is never going back to the way it was before, because AI is compressing economic time."

Investment Thesis #1: The July Speed Crash Was Positioning — The Bottom Is In

Argument The four-day QQQ surge (~10%) sits with post-GFC bottom markers (COVID, Liberation Day). S&P four-day rate of change hit 6% — biggest in a year — and the rally exceeded the entire May–July pennant range. Confirmation: IWM, NYSE Composite, equal-weight S&P, DAX, and FTSE all printed new highs; KOSPI +61% YTD (tech +116%). Factor vol collapsed after Situational Awareness Day; VIX at lows since January; CDX tracked VIX — no credit contagion. Thematic portfolio +6.5% for the week (second-biggest of the year). Earnings growth near ~50%, margins 13% → 17%, revenue growth ~15%, PMI near 60.

"That is not something to fade."

Contrarian element Podcast bears call a false breakout into September seasonality. Visser: hedged pods, summer volume, and vol-target constraints force a chase while earnings revisions and global breadth stay constructive.

Trigger Hold above the July lows; factor vol stays subdued while 200-day slope remains upward for 80%+ of thematic names.

Names Broad AI thematic (Morgan Stanley basket + his 100-name / 10-name / 25-name global concentrated books); SK Hynix–linked Korean machinery/construction as long/short confirmation when memory dips.

Investment Thesis #2: Acute Compute Shortage — NVIDIA After Multiple Compression

Argument Gavin Baker's Silicon Valley audit (must-listen): no negative demand metric — GPU availability, rental pricing, DRAM spot, token growth all accelerating. Same B200 cluster that rented ~mid-$2/GPU-hour seven months ago now ~just under $4. Nvidia at lowest forward PE in 10 years; Visser bought a "decent amount" over three weeks. Vera Rubin + token-per-watt cadence keep the cycle intact.

"There will never be enough compute, ever. We will continually be building compute."

Baker: ~250k–500k people using generative/agentic AI vs 8 billion humans — adoption still early; agents are tireless once deployed.

Contrarian element July bears treated open-source token shifts and hyperscaler CDS widening as demand death. Visser: open-source usage still consumes flops/memory/watts; CDS widening reflects leverage risk, not ROIC failure so far.

Trigger Baker-style metrics (GPU rental, DRAM spot, hyperscaler commentary) stay accelerating; any sustained deceleration would invalidate.

Names NVIDIA (NVDA) — "safer" post-compression name; not a near-term double, but asymmetric after the flush. Eaton (ETN) and Schneider Electric as electric-side / Vera Rubin participants in his earnings mosaic.

Investment Thesis #3: Memory Remains the Binding Constraint

Argument June paper ("AI = Memory") still governs. Elon Musk on SpaceX earnings: limiting factor is memory — supply ~+20%/year, demand ~+200%/year (or higher). More GPUs without more memory do not clear the bottleneck; consumer agents raise memory intensity further.

"If you've got demand increasing faster than supply, economics 101 would suggest the price increases."

Contrarian element He sold all Micron into the July ROC/breadth warning and rotated — early on price, not a thesis kill. Memory is still the watch variable for whether scarcity catches up to compute.

Trigger Memory prices / shortage language stay elevated; demand/supply gap does not close toward 20% growth.

Names Micron (MU) as the name he exited for crowding management; SK Hynix as the Korea-tape memory anchor; Entegris (ENTG) as a personally owned materials name in the thematic book.

Investment Thesis #4: Index Puts Are Broken — Hedge With ROC, Breadth, and Debasement Assets

Argument Peak gross leverage is in; factor vol (equal-weight momentum/value/quality/beta) should stay structurally higher (~25–30 rest of year vs old lows). Index puts failed the July panic. Replacement: 50-day ROC on the diversified 100-name AI index (chemicals, power, industrials — not just tech) plus internal breadth. At ~50% ROC, "you're going to fall"; risk grows from ~20%. July rotation that worked: silver, Bitcoin, Eli Lilly while semis puked. Same week: US–Japan yen intervention (first coordinated move since 1998) — Bessent protecting Treasuries via BOJ; gold +7.2% (third-biggest week in 16 years), reclaiming the 50-day; silver and platinum closing/breaking above 50-day.

"The breakaway from the S&P... means using index puts is not going to work the way that it used to."

Contrarian element Consensus still hedges AI with SPX puts. Visser hedges the two printing presses — fiscal debasement and AI dilution of corporate durability — with scarce assets.

Trigger ROC >20% with breadth rolling from highs (~90% above 20-day); gold/silver/platinum/BTC reclaiming trend after debasement signals (yen intervention, Warsh/rate-hike odds).

Names Silver (SLV — AI + debasement; "good-sized" position), gold, platinum, Bitcoin (still bear-market until breakout, but ecosystem/Clarity Act tape resilient), Eli Lilly (LLY).

Investment Thesis #5: Agentic Adoption — Palantir as the Public Proxy

Argument Cannot see Anthropic/OpenAI run-rate cleanly; Palantir commercial is the public adoption mosaic. Commercial growth 54% / 64% YoY prints, one metric 149%; Q2 revenue $306M → $764M YoY. Closed 220 deals ≥$1M, 98 ≥$5M, 78 ≥$10M. Only software name he owns personally; Cadence (CDNS) and Synopsys (SNPS) sit in the thematic book as design software. AI-native one-employee million-dollar firms (post Opus 4.5) are the long-run disruptors of Fortune 500 profits.

"The agentic side is here and if you're not using it, you're falling way behind."

Contrarian element Bear porn on Palantir valuation ignores the commercial acceleration and deal size mix.

Trigger Next PLTR commercial prints keep compounding; AI-native business-formation charts stay parabolic.

Names Palantir (PLTR); Cadence (CDNS), Synopsys (SNPS); Fluence (FLNC) and Camurus as owned thematic laggards he flags for subscribers.

The Ecosystem Map: Where Visser Is Positioning

NVIDIA — Accumulated over ~3 weeks into decade-low forward PE; Vera Rubin setup
Palantir — Personal holding; only software name owned; commercial adoption proxy
Memory — Thesis intact; Micron sold into July crowding; watch for re-entry after puke, not abandon scarcity
Thematic software/design — Cadence, Synopsys in the model book
Owned thematic laggards — Camurus, Fluence, Entegris (plus PLTR strength)
Electric / Vera Rubin stack — Eaton, Schneider Electric earnings mosaic
Hedge / rotation sleeve — Silver, Bitcoin, Eli Lilly (worked in July); gold/platinum reclaiming 50-day
Relative preference — Infrastructure trade over hyperscalers (MSFT/META/AMZN); cloud alone insufficient long-term → utility/real-estate multiple risk
Value chain — Most durable alpha in physically scarce compute/memory/power vs seat-software abundance and bureaucratic public megacaps by 2030

Key Risks

Hyperscaler equity / credit — CapEx stays high while FCF stays negative as a group for a long time; eventually priced like utilities if products do not show; CDS is his preferred bear expression, not equity shorts
Google talent / bureaucracy — Jeff Dean departure and Demis Hassabis role shift as canaries; "negative on all public companies by the time we get to 2030"
More speed crashes — Factor vol structurally higher; next ROC/breadth warning will force rotation again
September Fed hike odds (~45%) — Still too high in his view; a hawkish surprise pressures debasement assets and risk
Bitcoin still below breakout — Explicitly still a bear market until the level clears; Clarity Act odds collapsed (near 50% → ~17% this year) and cold-storage exploit headlines
Demand deceleration — If Baker's GPU/DRAM/token metrics flip from acceleration, the scarcity re-buy thesis weakens

Investment Opportunities at a Glance

Tier	Name / Category	Core Thesis	Conviction Signal
1	NVIDIA (NVDA)	Post-crash multiple compression + acute compute shortage	Bought over 3 weeks; decade-low forward PE; Baker demand audit
1	Palantir (PLTR)	Public proxy for agentic/enterprise adoption	Only personal software holding; $306M→$764M Q2; deal-size ladder
1	Memory stack (MU / SK Hynix)	20% supply vs ~200% demand; Elon/SpaceX confirmation	"AI = Memory" paper; watch for post-puke re-entry
2	Eaton (ETN) / Schneider Electric	Electric infra in Vera Rubin ecosystem	Named in weekly earnings mosaic
2	Cadence (CDNS) / Synopsys (SNPS)	Design-software sleeve of thematic book	Explicitly in model portfolio
2	Entegris (ENTG) / Fluence (FLNC)	Owned thematic names still digesting the crash	Personal ownership called out
3	Silver (SLV)	AI materials + debasement hedge that worked in July	"Good-sized" position; closed above 50-day
3	Gold (GLD) / platinum	Debasement hedge after US–Japan yen intervention	Gold +7.2% week; 50-day reclaim
3	Bitcoin (BTC)	Same two-printing-press hedge; agent rails later	Held through Clarity/Saylor noise; waiting on breakout
3	Eli Lilly (LLY)	July rotation destination when semis crowded	Explicit sell-Micron → LLY path
4	Camurus	Owned laggard still not bounced	Subscriber Q&A ownership mention

Monitoring Checklist

Bottom Line

Do not fade the post-crash AI bounce — historic four-day QQQ/S&P ROC, global ATH breadth, and collapsing factor vol say the July speed crash was deleveraging, not cycle death.
Buy NVIDIA into decade-low forward PE while Baker's data shows zero deceleration — B200 rentals mid-$2 → ~$4 in seven months; there will never be enough compute.
Memory is still the binding constraint — ~20% supply growth vs ~200% demand; selling Micron was crowding management, not a thesis flip.
Stop hedging AI with index puts — peak gross leverage + structurally higher factor vol; use ROC/breadth and own gold, silver, platinum, Bitcoin, and Lilly as the shock absorber.
Palantir commercial is the agentic tell — hundreds of thousands of users vs eight billion people; public-company bureaucracy loses to AI-native firms over the decade.