Jordi Visser — AI Supercycle Investment Thesis

Source: The AI Supercycle with Jordi Visser | Raoul Pal the Journey Man, Raoul Pal The Journey Man, June 11, 2026 (recorded June 4, 2026).

The Framework: Compute vs. Energy and the Bottleneck Supercycle

Visser's organizing lens with Raoul Pal: the old business cycle — labor vs. capital, offices, hiring — is giving way to compute vs. energy. Intelligence is scaling on a double exponential (Reed's law: network effects squared), but physical buildout lags. That mismatch creates bottlenecks, and bottlenecks concentrate capital and attention into whatever is blocking the path to more intelligence per unit of energy.

Old Cycle	New Cycle	Investment Implication
Labor vs. capital	Compute vs. energy	Capex replaces hiring as the growth engine
PMIs and employment surveys	Bottlenecks in chips, power, materials	Shortages slow earnings even when demand is "too big"
Human TAM (7.5B consumers)	Agent TAM → infinity	Billions of digital workers consume only compute
One mega-cap winner (Nvidia)	Rotating bottleneck trades	Power, batteries, silver, semis inputs, applications, crypto

"The new world is if we can't actually make all the chips we need and we can't actually get the power we need... you end up with a little bit of demand versus supply mismatch."

Investment Thesis #1: The AI Supercycle Is Non-Optional — Own the Bottleneck Layer

Argument: Pal and Visser agree this is almost certainly a supercycle, not a normal business cycle. Data centers are roughly 30% built versus stated targets. The US–China race and the danger of any single frontier lab winning force continuous spend. Bottlenecks do not kill the cycle — they require more expenditure to clear the roadblock. When a bottleneck appears, capital floods into that layer until it clears.

"There is almost no way for this not to be a super cycle... what is the largest capex cycle I think humanity will ever see at this rate."

The market still treats this like 1995–2001 productivity growth with normal ups and downs. Visser argues the agentic shift — Opus 4.5 as the "official gun going off" for the agentic world — caught trillions of dollars of PMs off guard through early 2026. They are now fully onboard, which means the next leg is rotation, not a single-name melt-up.

Contrarian element: Consensus assumes bottlenecks cap AI earnings. Visser flips it: demand is so large that supply constraints slow reported earnings even while the secular trend accelerates — a commodity-world framing where volumes matter more than software-style multiples.

Trigger: Hyperscaler capex guides stay elevated while power/transformer/turbine lead times remain extended; semis trade sideways while bottleneck names outperform for 3–6 months.

Names: NVIDIA (NVDA), Micron (MU), Marvell (MRVL), GE Vernova (GEV), Freeport-McMoRan (FCX), silver (SLV).

Investment Thesis #2: Power and Battery Innovation — Silver, Grid Storage, and the Nvidia–Siemens Stack

Argument: Power is now the visible choke point — data center delays, sold-out turbines, transformer shortages. The recent Nvidia and Siemens work with Fluence on batteries (solid-state focus in China) is Visser's proof that the stack is solving around grid limits. Solid-state batteries need far more silver than lithium chemistry; he flags massive silver imports as the tell. If battery innovation unlocks peak-shaving on the existing US grid, today's peak-capacity constraint becomes a storage problem rather than a generation problem — but that path still runs through materials and grid hardware.

"Everyone should go look at how much silver is needed in solid state batteries instead of what lithium takes... if we could store and have it to deal with those peak capacity days, then we wouldn't we'd be able to use the grid almost completely."

Wall Street still treats commodity inflation as transitory. Visser sees structural underinvestment in atoms after a decade of bits-first markets — the same mistake as 2021.

Contrarian element: The copper-shortage trade is crowded and "not the greatest trade on Earth." Elon's Cybertruck move — 12V to 24V, 70% less copper — shows intelligence routing around material limits. Silver for next-gen batteries is the fresher bottleneck expression.

Trigger: Continued data-center delay headlines; solid-state battery pilot announcements; silver import data accelerating; power-equipment backlog guides re-affirmed.

Names: SLV, FCX, GEV, Fluence Energy (FLNC), Siemens Energy (SMNEY / ENR.DE).

Investment Thesis #3: Rotation to Applications — Eli Lilly as "Human Software"

Argument: Jensen Huang's five-layer cake puts energy and chips at the bottom and applications at the top. Institutional meetings still fixate on seat-based SaaS — the wrong mental model. Visser names Eli Lilly as the company already sucking in capital for the application layer: GLP-1 cash flows financing longevity and biological software, with a 1,000-GPU LillyPod data center on campus and a Nvidia partnership in Silicon Valley. Advertising was Google's starter engine; GLP-1s may be Lilly's equivalent for the human-software era Pal has discussed (peptides, genetic science).

"You know who's starting to suck in all of the capital right now for the application layer? It's for human software... I believe we're going to look back and realize that GLP1s were the ability to finance the next stage of the human software."

This is the rotation Visser wants when the capex trade digests — alongside longevity themes and bottleneck commodities.

Contrarian element: AI investors are still mapped to Mag 7 infra names. The application-layer winner may be a drug company with a GPU cluster, not a SaaS ticker.

Trigger: LillyPod / Nvidia partnership milestones; GLP-1 revenue sustaining capex into AI-driven drug discovery; semis consolidating while LLY holds relative strength.

Names: Eli Lilly (LLY), NVIDIA (NVDA) as enabler.

Investment Thesis #4: The IPO Digestion Window — Peak Infrastructure Capex Trade for Now

Argument: The market just finished an "AI capex all you can eat buffet." Google's ~$85 billion public equity raise — larger than the bottom 360 S&P 500 names combined — is a fight for finite capital ahead of massive IPOs (~$4 trillion potential supply including SpaceX, OpenAI, Anthropic). Cerebras already trades ~50% below its IPO — normal for new issues but a warning. Visser expects 3–6 months of digestion: not the end of the market, but a peak in the infrastructure capex trade where capital rotates.

"We just had an AI capex all you can eat buffet and it just it's going to end with the IPOs... we need to digest everything that just happened for 3 to 6 months."

When Mag 7 earnings are this strong, narrative assets lose the attention war — Bitcoin and crypto lag because PMs can justify traditional P/E buys. That is temporary. Once infra names stall, probability of crypto catching up rises faster.

Contrarian element: Strong AI earnings are bearish for crypto only during the gap-higher phase. The setup is a rotation trigger, not a refutation of the supercycle.

Trigger: Major AI IPO pricings; NVDA / semi complex fails to make new highs for 8–12 weeks; capital flows into lagging themes (longevity, bottlenecks, crypto).

Names: Alphabet (GOOG), Bitcoin (BTC), layer-1 blockchains (BTC, ETH — Visser focuses layer 1s, not mid-curve DeFi).

Investment Thesis #5: Crypto's Third Wave — Financial Guardrails for the Agentic Economy

Argument: Visser first outlined this at Pal's Miami event: crypto's third wave requires (1) AI physical infrastructure being built, (2) the agentic trigger firing — now confirmed by Opus 4.5 — and (3) financial guardrails for token-hungry digital employees. Agents need identity, settlement, and machine-readable value transfer. Tokenization unlocks the two-thirds of global assets that are illiquid — real estate, private credit, PE, art — raising velocity and GDP by definition. The invisible agentic economy (API calls, MCPs, vault data) becomes the largest marketplace on Earth, mostly unseen.

"The third wave of crypto... the financial guard rails, the transactions, the velocity of money, all of that will be happening... tokenization needs to come to make them more liquid to where that money can leave this massive $400 trillion and start to enter the crypto world."

He is hyperfocused on layer 1s, explicitly not mid-curve plays like Hyperliquid where investors demand cash-flow buckets they already understand. Crypto is in a painful digestion after ETF + political enthusiasm — lower highs and lower lows — but the underlying trend (banks, stablecoins, tokenization) keeps building.

Contrarian element: Patience is wrong framing; impatience is fine if you read parabolas as sticks, not bubbles. The third wave starts when the capex trade breaks a moving average and capital seeks a new race.

Trigger: NVDA / semi leaders consolidate; BTC breaks a key moving average after capex-trade exhaustion; tokenization pilots at major exchanges accelerate.

Names: Bitcoin (BTC), Ethereum (ETH), layer-1 protocols; tokenization infrastructure (category, not a single ticker named).

The Ecosystem Map

Personal winners cited: Micron (MU) and Marvell (MRVL) — "lucky to catch" prior third-wave-style moves; waiting for the same in crypto
Infra stack: NVIDIA (NVDA) partnerships — Lilly, Siemens/Fluence; DGX setup planned for personal Hermes agents
Hyperscaler capital: Google (GOOG) $85B equity raise as capex war chest; OpenAI / Anthropic debt-constrained → equity IPO path
Research method: "Knowledge brains" — Jensen Huang transcripts in Notebook LM / Claude; "he's basically telling you what companies to buy"
Crypto focus: Layer 1s; stablecoin and bank adoption observed at Consensus Miami (15,000 attendees, no retail)
Macro partners: Raoul Pal's compute-vs-energy framing, invisible economy / token factory essays, GMI intelligence-per-energy index (in progress)
Explicit avoid / not core: Hyperliquid (mid-curve cash-flow bucket); rebuilding SaaS from scratch when agents can plug into existing APIs (Xero example)

Key Risks

Algorithmic efficiency collapses capex: Agentic and algorithmic gains could mean "the capex numbers may never happen to the degree people think" — a bubble on physical buildout while software-side optimization wins.
Bottleneck innovations reduce material trades: Higher voltage wiring, optical fiber (Vera Rubin), and AI-assisted engineering can erode copper/silver shortage narratives faster than expected.
Strong earnings starve narrative assets: When AI earnings justify any P/E, Bitcoin and crypto lose attention and capital — extending crypto's lower-highs pattern.
IPO supply shock: ~$4T potential IPO pipeline plus Google's raise competes for the same liquidity pool — sloppy 3–6 month digestion, not necessarily a bear market but frustrating for momentum traders.
Chimath-style SaaS terminal value: Long-duration software repriced to near-zero terminal value in an agent world — bifurcates markets; legacy SaaS drags indices even as infra wins.
"Smarter is worse" pattern matching: Investors with bloated mental context windows map 2026 to dot-com, 1987, or 1929 on three data points and miss the supercycle.
Semi input bottlenecks (NAPA, helium, petrochemicals): Lithography and fab inputs can throttle volumes even when GPU demand is infinite.

Investment Opportunities at a Glance

Tier	Name / Category	Core Thesis	Conviction Signal
1	NVIDIA (NVDA)	Center of agentic buildout; partnerships (Lilly, Siemens); personal DGX/agent infra	Named across every layer of the stack; "backlog on all the Apple stuff" vs. Nvidia work queue
1	Micron (MU)	Prior third-wave win; memory/HBM scarcity in agentic broadening	"I've been lucky to catch one in Micron" — wants repeat in crypto
2	GE Vernova (GEV)	Power bottleneck — turbines, transformers, data-center delays	Data-center buildout "30% built"; power named as clearest choke point
2	Silver (SLV)	Solid-state batteries need more silver than lithium; import surge	Nvidia–Siemens–Fluence battery push; "huge silver imports"
2	Freeport-McMoRan (FCX)	Copper still structural to electrification despite efficiency hacks	Crowded trade but still in bottleneck rotation set
2	Alphabet (GOOG)	$85B raise funds capex war; ads were starter engine for bigger buildout	Largest public equity raise Visser has seen; fights for finite capital
3	Eli Lilly (LLY)	GLP-1 cash flows finance "human software"; 1,000-GPU LillyPod	First time stated on podcast: Lilly as application-layer capital sink
3	Marvell (MRVL)	Agentic networking / custom silicon; prior Visser winner	Named alongside Micron as caught third-wave move
3	Fluence Energy (FLNC)	Grid storage partner in Nvidia–Siemens battery work	Explicitly named with Siemens on battery/solid-state path
3	Bitcoin (BTC)	Third-wave setup when capex trade digests; financial guardrails for agents	"If Nvidia stops going up, probability of crypto going up goes up much faster"
4	Layer-1 crypto (BTC, ETH)	Machine-readable money for invisible agentic economy	Focus layer 1s; avoid mid-curve DeFi cash-flow proxies

Monitoring Checklist

Bottom Line

This is a supercycle, not a cycle. The largest capex wave Visser thinks humanity will run — and bottlenecks don't stop it, they tax it forward into power, materials, and storage.
The trade rotates every few months. January 2026 agents were misunderstood; by Computex 2026 everyone was onboard. Next leg is not the same Nvidia-only expression — it's silver, power, Lilly-style applications, then crypto.
Google's $85B raise is the tell. Finite capital meets infinite AI ambition; add ~$4T of IPO supply and you get a 3–6 month digestion, not the end of the trend — the handoff to third-wave crypto and bottleneck commodities.
When AI earnings are this good, crypto looks broken — that is the setup. Narrative assets lose the attention war during the gap higher; Visser explicitly links NVDA stalling to faster BTC catch-up.
Lilly is the applications-layer call nobody in AI meetings is making. GLP-1 cash flows plus 1,000 GPUs on campus is "human software" financed like Google's ad engine — the rotation target when infra consolidates.