06-reference

diamandis innermost loop ai infrastructure thesis

2026-05-12·reference·source: X long-form note (Peter Diamandis personal)·by Peter H. Diamandis
diamandisai-infrastructureinvesting-thesisinnermost-loopchip-stocksdata-centerpower-infrastructurefrontier-labsabundancefounder-capital-deployment

Investing in the Singularity — Peter Diamandis

Why this is in the vault

Founder has capital sitting in cash and wants to develop RDCO's investing capability. Diamandis presents a structured 4-layer thematic-investing framework ("Innermost Loop") with concrete 12-month return data across 17 named tickers and 3 demand-side anchor metrics. Worth filing as the anchor doc for RDCO's investing-thesis surface. Also: I (Ray) initially skimmed this based on a bookmark/like-ratio heuristic that was wrong for a 16-hour-old post; founder corrected me. Pulled via xmcp article.plain_text (the full-body field that bypasses the WebFetch 402 wall on X long-form notes).

The core argument

There is no longer a single AI-tech investment cycle. There are four interlocking supply chains all hitting escape velocity at the same moment:

  1. Frontier labs build models
  2. Chips train and serve them
  3. Data centers house the chips
  4. Power feeds the data centers

Each layer's demand drives the next. The bottleneck has migrated from chips to power. $156B of US data-center projects are currently blocked or stalled by local power-infrastructure opposition (per CNBC / Data Center Watch). That makes the power layer the highest forward-leverage public play because it's the named constraint.

This is structurally different from the dot-com cycle or the smartphone buildout, where only one supply chain was under pressure at a time.

The Innermost Loop framework

Layer 1: Frontier Labs (mostly private)

Lab Valuation Trajectory Funding
OpenAI $80B early-2024 → $852B March 2026 ($122B raise, largest ever) → projected $1T+ IPO Revenue $2B (2023) → $6B (2024) → $20B (2025); 910M WAU
Anthropic $18B early-2024 → $380B Feb 2026 → $850-900B preemptive offers late April → $1T secondary on Forge Global May 2026, surpassed OpenAI 55x in 2 years; IPO projected ~Oct 2026
xAI $230B Jan 2026 Series E ($20B raised) Revenue ~$100M (2024) → $3.8B annualized end-2025 = 38x YoY; $45B total funding
Google DeepMind Not separately valued Alphabet $2.1T → $4.81T (+129%) is the implicit DeepMind premium

Aggregate: frontier lab valuations grew by >$2T in 2 years. Q1 2026 funding to foundational AI startups doubled all of 2025.

Layer 2: Semiconductors — 12-month returns May 2025-May 2026

Ticker Return Note
MU (Micron) +770% HBM memory for AI
INTC (Intel) +483% Foundry turnaround + CHIPS Act
AMD +343% AI accelerators
TSM (TSMC) +133% Foundry leader
AVGO (Broadcom) +107% Custom ASICs
NVDA (Nvidia) +85% GPUs

Average ~320% vs S&P 500 +31% (roughly 10x SPY). Roughly 1/3 of the entire S&P 500 gain came from just these five chip stocks plus Intel.

Layer 3: Data Center Infrastructure — 12-month returns

Ticker Return Function
VRT (Vertiv) +256% Cooling, power distribution, racks
GEV (GE Vernova) +158% Turbines, grid infrastructure, nuclear services
BE (Bloom Energy) +1,647% Fuel cells for onsite DC power
OKLO +178% Advanced nuclear for DCs
FLNC (Fluence Energy) +200% Battery storage, grid-scale
TLN (Talen Energy) +73% Nuclear power for AWS

Average ~419% vs S&P 500 +31% (~13x SPY).

Layer 4: Hyperscalers (already large-cap)

Ticker 2-yr trajectory AI angle
GOOG $2.1T → $4.81T (+129%) DeepMind, Gemini, Cloud
AMZN $1.9T → $2.93T (+54%) AWS AI infra dominance
META $1.2T → $1.57T (+31%) Llama open-source + compute buildout
MSFT $3.07T (flat) OpenAI partnership + Copilot - most "AI-leveraged" large-cap

Collective hyperscaler market cap: >$12.4T. Projected 2026 AI capex: $750B in a single year (larger than Switzerland's GDP).

Three demand-side anchor metrics

  1. $1 trillion in global semiconductor sales projected 2026 (vs $527B for entire year 2023). Q1 2026 alone: ~$300B; March 2026 alone: $99.5B. The chip industry is doubling in 3 years.
  2. $750 billion in 2026 hyperscaler AI capex (CreditSights revised up from $650B). Flows directly into chips, power, cooling, DC construction.
  3. $2.02 trillion projected AI data center market by 2032 (27.5% CAGR from $471B in 2026). Smartphone market took 15 years to reach similar scale; AI infra is on pace to do it in 6.

Mapping against Ray Data Co

(a) Founder's "capital sitting in cash" question

This is a directly actionable thesis for RDCO's investing capability. Four practical observations from the data:

(b) Cross-link to today's Stratechery piece

Today's Stratechery "SpaceX and Anthropic, xAI's Two Companies" filed at 2026-05-12-stratechery-spacex-anthropic-xai-musks-two-companies.md discusses an Anthropic-xAI compute deal. The Diamandis piece names xAI's $230B valuation and Anthropic's $1T secondary; the Stratechery piece names the partnership. Combined: xAI is providing compute to Anthropic, which means xAI is monetizing its compute moat while Anthropic continues to lead on the model layer. The compute-supply concern that justified our prior "stay on Claude" substrate position is materially resolved.

(c) Connection to abundance-thesis / Not Boring synthesis

[[06-reference/2026-04-30-not-boring-scarce-assets-abundance-driven-scarcity]] — Packy's framework predicts the scarce asset migrates when an adjacent input becomes abundant. Diamandis's data shows this happening in real-time: compute (abundance-side) is up massively, but the scarce asset has migrated to power (the named bottleneck). This validates the Packy framework with concrete market evidence.

(d) RDCO investing project folder

Worth standing up ~/rdco-vault/01-projects/investing/ with:

This is a logical extension of RDCO's existing project-folder structure and gives the founder's investing capability the same vault-canonical operating discipline that data + content surfaces have.

Disclaimer

Diamandis explicitly says: "I'm not a financial advisor and this isn't financial advice." This vault note inherits the same disclaimer. Capital deployment decisions sit with the founder + a real CFP / wealth manager / accredited-investor counsel. This file is research synthesis, not a recommendation.

Related