The AI Infrastructure Thesis — AI-Innovation Portfolio by Logic Fund Management

The core insight

In a world where the cost of digital labor approaches zero, the scarce physical resources that power all of that labor become the most valuable assets in the economy. Intelligence becomes abundant. Power, copper, fiber, uranium, land, skilled labor — those stay scarce. Constrained by geology, physics, and time.

You cannot code a new copper deposit. You cannot download the inputs for a new data center.

The AI-Innovation Portfolio owns the physical infrastructure of the AI revolution — the companies that control the scarce resources powering limitless AI output. Actively managed. Screened for earnings quality. Rebalanced as themes evolve.

Five frameworks

Framework 1

Abundance vs. Scarcity

Intelligence becomes abundant while physical infrastructure remains constrained. The companies that own the scarce physical resources — power generation, copper mines, fiber optic glass, uranium, data center land — capture the most asymmetric share of growth. On February 5, 2026, three frontier AI models and a free open-source agent launched on the same day. The stocks that rallied were the physical infrastructure companies. The stocks that sold off were enterprise software companies built around human seat-based pricing. The market is telling you what's scarce and what isn't.

Framework 2

Bits vs. Atoms

Software scales infinitely. Hardware doesn't. An AI model can be copied and deployed to a million users for negligible marginal cost. But the GPU cluster that runs it requires copper wiring, fiber optic glass, power generation, cooling systems, and physical land. Invest in the atoms. The companies manufacturing the physical components that AI cannot exist without — Corning's fiber optic glass, Amphenol's active electrical cables, Vistra's power generation — are the bottlenecks that determine how fast AI can scale.

Framework 3

Tools vs. Outcomes

In a gold rush, the safest money is made by the people selling picks and shovels. The AI-Innovation Portfolio has invested for the waves of the technology revolution — positioning early in the companies that are the obvious picks and shovels. ASML's lithography machines, Synopsys's chip design software, ARM's chip architecture, Teradyne's test equipment — these are the tools that every AI chipmaker depends on regardless of which model wins.

Framework 4

Users vs. Usage

AI adoption is widespread. What drives the next wave of value is utilization depth. When AI crossed from thinking to acting on February 5, 2026 — autonomous agents running around the clock, executing tasks, managing workflows, transacting money — that wasn't a bump in inference demand. That was an explosion. Every autonomous agent is an inference engine that never sleeps. Every enterprise deploying agent teams spins up inference at a scale that didn't exist before. Revenue is constrained only by computing and power capacity.

Framework 5

Elon's Limitless Economy

Elon Musk's thesis: when the cost of labor approaches zero, economic output has no ceiling. He's changed Tesla's mission to "sustainable abundance." The logic is simple — humanoid robots and autonomous AI remove any meaningful limit on the size of the economy. This aligns with the 1940s-like growth explosion where the economy averaged 14% real GDP growth over a five-year period. If AI drives double-digit real GDP growth, virtually every asset class appreciates. The question isn't whether your holdings go up. The question is what captures the most asymmetric share of that growth.

The February 5th inflection

On a single day — February 5, 2026 — three of the most important AI developments landed simultaneously. Anthropic released Claude Opus 4.6, with autonomous agent teams that manage entire workflows in parallel. OpenAI released GPT-5.3-Codex, a model that helped build itself. And OpenClaw, an open-source personal AI agent, went viral — with 150,000 people running autonomous AI assistants on their laptops around the clock.

This is the moment AI crossed from "AI can think" to "AI can act." Intelligence is no longer sitting inside a chatbot waiting for a question. It's executing tasks, managing workflows, deploying code, and transacting money — autonomously.

The cost of intelligence — the cost of labor — is heading toward zero. And on February 5th, the market told you what that means: the physical infrastructure stocks rallied, the human-seat-based software stocks sold off. The scarce resources that power limitless AI output are where the value accrues.

The always-on inference phase

Every time someone hits enter, every time a machine pings a model, that's inference. And when inference is running, the meter is running. Revenue is being produced. New data is being created.

When autonomous agents run around the clock — executing tasks, managing workflows, calling other agents — that's not human-initiated query-response. That's always-on inference consuming compute, power, bandwidth, and storage continuously.

Google doubled its expected 2026 capex to $185 billion. Amazon announced $200 billion. Andy Jassy said they're "monetizing capacity as fast as we can install it." Taiwan Semi reported a 20% jump in revenue month-over-month in January — typically a quiet month — and greenlighted a $45 billion capex plan. The signal hasn't changed. It has intensified.

The guardrails thesis

The AI revolution has entered the most profitable phase — and the most dangerous phase. 150,000 people are running autonomous AI agents on personal computers with access to their files, browsers, and systems. Enterprise companies are deploying agent teams with access to real infrastructure. The attack risk for critical systems — power grids, financial networks, air traffic control — expanded dramatically.

The companies that build the guardrails — security, governance — are mandatory infrastructure. Cloudflare sits at the network edge, filtering traffic before it reaches the infrastructure. Palantir's Ontology platform is the governance layer — the system that prevents an autonomous agent from executing an unverified action. In a world of AI agent swarms, somebody has to be the air traffic controller.

How the portfolio is built

The AI-Innovation Portfolio holds 20 actively managed positions organized around these five frameworks — the companies that control the scarce physical resources, the essential tools, the mandatory guardrails, and the platforms that power the AI revolution. Every position maps to a specific framework. Every entry and exit is documented with the thesis, the catalyst, and the reasoning.

Subscribers receive the full portfolio with entry prices, the thesis behind every position, entry and exit alerts in real time, and ongoing research notes as the landscape evolves.

Go deeper on each thesis

The Scarcity Thesis

Investing in the AI Revolution

The Abundance Thesis

Robotics & the Limitless Economy

AI Agents

The Infrastructure Inflection

The Chip Supply Chain

Lithography to Test Equipment

Cloud & Compute

The Capex Arms Race

The Energy Bottleneck

Power for the AI Era

Security & Governance

Guardrails for the Agent Era

The research. The portfolio. The frameworks.

20 actively managed positions. Entry and exit alerts. The thesis behind every move. $297 / quarter.

Subscribe to AI-Innovation Portfolio →

Privacy · Terms · Risk Warning

The returns shown are from the AI-Innovation Portfolio model portfolio — actual recommendations at published prices. Because subscribers manage their own accounts, individual results will vary. All performance figures should be considered hypothetical. Past performance is not indicative of future results.