Two of the most reliable early-warning gauges in markets. The HY Credit Spread measures how nervous the bond market is — when it spikes, a stock market drop almost always follows within months. WTI Crude measures energy cost to the global economy — too high causes inflation and recession, too low signals demand collapse. When both are elevated at the same time, that's a confirmed risk-off signal. When both are low, conditions are ideal for risk-on positions.

Traditional GDP measures consumption, labor output, and physical goods. The next decade’s economy will be measured in compute capacity, energy throughput, and automation density — none of which current GDP methodology captures cleanly.

The last time electricity demand grew faster than GDP was the 1990s internet build-out. We’re in that pattern again — data center power consumption is growing at 15-20%/year while GDP grows at 2-3%. No official metric tracks this divergence yet. These are the early indicators.

Sector implication: Structural long positions in power infrastructure (VRT, ETN, PWR), uranium (CCJ, NXE), copper (FCX), and AI chips (NVDA, AMD) are direct beneficiaries of a compute-denominated economy. These positions get stronger as this thesis enters mainstream GDP discourse — likely 2028-2032.

Hyperscalers don't pick locations randomly — they follow the lowest cost of electricity, highest fiber density, and most favorable permitting. Electricity cost is the single largest operating expense for an AI data center, accounting for 40–60% of total TCO. Every $0.01/kWh advantage on a 1GW data center campus = $87.6M/year in savings.

Investment edge: Hyperscaler site-selection announcements are public. Tracking which regions win the build cycle lets you front-run the local utilities, grid operators, real estate REITs, and construction names before Wall Street prices in the 10-year demand curve.

Hyperscalers are spending $300B/year building centralized compute infrastructure. DePIN (Decentralized Physical Infrastructure Networks) is the counter-thesis: aggregate idle compute from billions of existing consumer devices — laptops, GPUs, servers, smart TVs, Raspberry Pis, even idle Tesla compute — into a permissionless, tokenized network that competes on price.

The constraint isn't hardware — it's coordination. DePIN protocols solve this with crypto incentives: contribute idle compute, earn tokens. The supply is already deployed globally. It just isn't connected.

Signal to watch: DePIN sector market cap relative to Big 4 capex growth. As centralized costs rise, the economic case for decentralized compute strengthens. When DePIN utilization rates exceed 40% network-wide, it becomes a credible pricing threat to AWS/Azure spot pricing.

Every compute economy technology — AI chips, EV batteries, wind turbines, fusion reactors, electrolyzers — runs on a handful of metals. China controls 60–98% of global processing for most of them. Export restrictions on Gallium (2023) and Germanium (2023) were the first shots. This is the critical infrastructure layer beneath the entire energy + compute transition. Own the supply chain before the thesis goes mainstream.

Wright's Law: every time cumulative production of a technology doubles, cost falls by a fixed percentage. Solar has followed a 20% learning rate for 40 years without exception. The curve predicts when each technology becomes economically inevitable — before the market prices it in. Track the curve, front-run the re-rating.

Quantum computing and humanoid robotics are both following the same S-curve that AI followed from 2017–2022 — years of "it's still early" followed by a rapid re-rating when a clear threshold is crossed. For quantum, the threshold is ~1,000 fault-tolerant logical qubits (currently 1,000+ physical but not yet fault-tolerant). For robotics, it's when a humanoid can perform an unstructured manual labor task reliably at a cost below $30/hr. Track the qubit count and the robot deployment numbers — the supply chain plays re-rate 2–3 years before mainstream adoption.

Getting one kilogram of anything into orbit cost $54,000 in 2010. SpaceX's Falcon 9 brought that to $2,700. Starship targets $100/kg. That single number — cost per kg to orbit — is the unlock for every space business case. At $100/kg, space-based solar power, low-latency data centers in orbit, and asteroid mining all become economically viable. This section tracks where we are on that curve, what asteroid resources are within reach, and the TCO crossover point where building a data center in space becomes cheaper than on Earth.

Water is the hidden constraint on AI data center expansion. A 1 GW data center uses ~5–7 liters of water per kWh for cooling — ~5 billion liters/year per campus. The regions winning the AI build cycle (Arizona, Texas, Virginia) are also experiencing severe aquifer depletion. This tension resolves in the 2030s and it will be the most important siting constraint nobody is talking about yet. CME launched California water futures in 2020 — the first time water traded as a financial instrument.

Every energy breakthrough follows the same ladder: Lab proof → Pilot → Commercial pilot → Grid scale → Cost competitive without subsidy. The supply chain re-rates 5–10 years before the technology reaches grid scale. Track where each technology sits on the ladder and what metric determines when it advances to the next rung.