AI Infrastructure

TXN posted 19% revenue growth — its highest in four years — led by data center revenue surging 90% YoY, while pricing held flat against management's expectation of a decline, signaling demand is running ahead of even industry insider forecasts.

Global AI token usage has surged roughly 250% since early January 2026, from 6.4 trillion to 22.7 trillion tokens per week, pushing compute demand past available supply and creating significant alpha potential in AI infrastructure.

Global AI token usage has surged 250% since January to 22.7 trillion tokens per week, pushing compute demand above available supply and creating significant alpha potential in AI infrastructure.

Morgan Stanley's thematic basket — spanning AI infrastructure, energy security, defense, healthcare, and humanoid robotics as an emerging area — averaged 38% returns in 2025 and remains 12 points ahead of the S&P YTD in 2026.

$156 billion in data center projects were blocked or stalled by local communities in 2025, and total new data center capacity under construction declined for the first time since 2020 — a regulatory headwind the market may be underpricing.

The 'HALO' (Heavy Assets, Low Obsolescence) framework holds that AI commoditizes software businesses by eliminating their competitive moats, while hard-to-replicate physical infrastructure networks — logistics, fulfillment, satellites — attract higher valuation premiums precisely because they can't be disrupted the same way.

At least $156 billion in data center projects have been blocked or stalled by local communities in 2025, data center capacity under construction fell for the first time since 2020, and 14 states have active moratorium or restriction bills — a material headwind for AI buildout.

Cancellation rates for contested data centers have more than doubled year-over-year, with states like Maine banning new builds outright — creating a structural supply bottleneck for AI compute.

Morgan Stanley's Chief US Economist sees AI-driven CapEx spending and productivity gains as largely orthogonal to the oil shock, with positive structural trends expected to persist into 2026.

Morgan Stanley's Chief US Economist sees the oil shock as orthogonal to AI investment — business spending, productivity gains, and AI-related CapEx remain in place barring oil above $150/bbl.

US grid capacity has been flat since the 1990s while data center power demand is surging exponentially. Nuclear is the safest, cleanest baseload source and the only one that can serve always-on industrial and compute workloads at scale.

100-300MW SMRs are the ideal size for dedicated data center power, operating fully islanded from the grid. Hyperscalers already plan to 'bring your own energy' (BYOE), and offering surplus power to local communities resolves permitting friction while adding minimal project cost.

War-damaged Gulf energy infrastructure requires natural gas turbines already identified as among the world's scarcest goods due to AI data center buildout — adding a new demand shock to the power generation equipment industry supply chain.

The massive AI infrastructure buildout is transforming a capital-light tech sector into a capital-intensive one, increasing demand for physical materials at the same time that geopolitical disruption threatens supply — a combination that could trigger a broader commodities super-cycle resembling the 1970s.

Tech's shift from capital-light to capital-intensive is raising physical materials demand at the same time geopolitical pressures strain supply — conditions mirroring the 1970s, when gold and oil sharply outperformed equities.

Just as declining labor share mechanically grew corporate free cash flow since 1980, AI adoption that further reduces labor's cut of output would be 'a plus for stock values' — potentially sustaining today's elevated valuations.

Anthropic's revenue explosion from $1B to $30B annualized run rate in months demonstrates that hyperscaler CapEx spending will generate returns, refuting the AI bubble narrative.

Prediction markets show half of 2026 data centers may be delayed or canceled due to insufficient electricity generation, creating opportunities in the electricity space despite unlimited AI demand.

OpenAI board members and industry analysts warn the massive AI spending is sustained by partners borrowing from each other, creating circularity where someone will eventually have to pay the piper.

An 18-month New Yorker investigation reveals insiders warning the AI infrastructure buildout is sustained by partners borrowing from each other, with a board member saying OpenAI is "levered up in a way that is scary."