Escape Velocity | Spending Is Not the Same as Accelerating
The mass of what holds you back—and your distance from it—determines the velocity needed to escape.
Photo: Getty Images
Escape velocity is the speed an object must reach to break free of a gravitational field without further propulsion. We use it here as a working definition of what we are looking for in this cycle: companies whose own cash flow will eventually generate enough return to sustain their growth without continued reliance on external financing.
Excerpt from Client Publication (June 2026)
Executive Summary
The artificial intelligence narrative is shifting from a story of unprecedented infrastructure spending to one of measurable economic productivity. While the initial wave of capital has heavily rewarded the technology companies building the foundation of AI, the next phase of market leadership will be defined by the businesses that successfully harness these tools to expand margins, increase cash flow, and build durable competitive advantages.
This transition, which we term "Escape Velocity," carries profound implications for global markets and long term asset allocation. The sheer scale of the current AI infrastructure buildout—now approaching three quarters of a trillion dollars annually—brings substantial financing risks and structural dependencies that we are monitoring closely. Furthermore, the physical constraints of AI deployment, specifically regarding power generation and the control of critical minerals, have elevated technological development into a primary arena of geopolitical competition.
Our portfolio strategy is designed to navigate this inflection point with a disciplined focus on high-quality cash flow and capital protection. Over the near term, we are becoming increasingly selective regarding the earliest AI infrastructure beneficiaries, where current valuations often reflect exceptionally optimistic expectations. Moving forward, we are systematically positioning portfolios to capture the broader, second order effects of this cycle. We are targeting investments across vital energy infrastructure, resilient global supply chains, and an emerging class of "productivity compounders"—well-managed companies across diverse sectors that are utilizing AI to fundamentally improve their operational economics. Ultimately, our focus remains on identifying durable businesses capable of self-funded, sustainable growth through changing macroeconomic conditions.
Introduction
This paper is about productivity: whether the current investment cycle produces a durable, economy-wide acceleration in output per unit of input — of the kind last seen with electrification and the early internet — or simply a very large and well-financed technology buildout. AI is the catalyst, not the subject. The scale of that catalyst is nonetheless part of the evidence: AI infrastructure spending has been revised upward three times in under a year — from roughly $600 billion at the end of 2025, to $700 billion at the beginning of 2026, to a current range of $725 785 billion (BNP Paribas; Morgan Stanley puts the 2026 figure closer to $765–805 billion). The amount of capital is substantial, as is the optimism reflected in the rate of change. A separate question, and in our view a more important one, is whether the investment will increase productivity and with it the potential for higher long term non inflationary growth across the economy.
On the spending/investment sustainability side of the equation, free cash flow across the handful of companies funding this buildout (hyperscalers) — Amazon, Microsoft, Alphabet, Meta, Oracle, Alibaba, and Tencent — peaked near $240 billion in 2024 and is compressing sharply as capital spending continues to climb; one widely cited analysis of SEC filings (Epoch AI) finds aggregate hyperscaler capex growing roughly 70% a year against operating cash flow growth of about 23%, putting the group's combined free cash flow on a path toward zero around the third quarter of 2026. The gap between the money being invested and the cash flow AI is currently generating in revenue is now roughly $600 billion a year, per Sequoia Capital's widely cited estimate (David Cahn). This shortfall is increasingly being financed through debt — Morgan Stanley estimates hyperscaler borrowing could exceed $400 billion in 2026, more than double 2025's roughly $165 billion — vendor financing, and related-party arrangements that have not yet been tested through a full credit cycle. We think this financing dynamic has received less attention than it deserves, relative to the more familiar debate about whether AI will eventually be productive.
Physical constraints are also becoming binding. Interconnection queues for new grid connections now commonly run several years in many regions, and the International Energy Agency estimates roughly one-fifth of planned data center projects face delays tied to grid constraints. Electricity availability, rather than chip supply or capital, is increasingly the constraint on how quickly committed capacity can be deployed.
At the same time, the first measurable productivity gains are appearing. U.S. nonfarm business labor productivity rose roughly 2.5–2.9% year-over-year through late 2025 and into early 2026, above the sector's trailing five-year average of roughly 1.5–2% (U.S. Bureau of Labor Statistics, Productivity and Costs), and firms that have integrated AI into their operations are showing meaningfully wider margins than firms that have not. The evidence is early, and the question of how much reflects genuine efficiency gain versus capital deepening and headcount reduction remains open. But there is now real data to evaluate, rather than a narrative to accept or reject on faith.
A central question we return to throughout this paper is which companies can convert this investment cycle into durable, self-generated cash flow before today's favorable financing conditions change. We expect the next 18–24 months of evidence — on financing sustainability as much as on productivity — to separate companies still dependent on external capital from those that have reached a point of self-sustaining growth. We also expect the geopolitics of this cycle — who controls the power, the semiconductors, the critical minerals — to matter as much as the balance sheets, and we devote a section of this paper to that question directly.
This paper proceeds in six parts: what has actually changed in the past six to twelve months, what is at stake economically, what is at stake geopolitically, how the buildout is being financed, where returns are already showing up, and — the part we think matters most — a framework for telling durable compounders from companies that are simply along for the ride. Every section closes with its portfolio implications.
Please log in to read the full research paper.
PRIVATE CLIENT ACCESS
ABOUT THE AUTHORs
Keith Savard has more than 40 years of economic/finance research and investment experience in the United States and overseas. He has worked as a staff member at the Board of Governors of the Federal Reserve System and as an international economist at the U.S. Department of State. A solid background in macroeconomic analysis and financial regulatory and monetary policy issues, combined with sovereign and credit risk skills, has enabled Keith to offer investors actionable top-down investment strategies and expertise facilitating a broad-range of asset allocation decisions. His deep knowledge and understanding of emerging markets, gained through extensive travel and meetings with cabinet-level officials, company CEOs and local investors, affords opportunities to invest confidently beyond the U.S.
Susan Dahl is a seasoned executive, industry leader, and dedicated client advisor, with over thirty years of international and domestic investment experience working for both large and small organizations. Susan is known for her ability to unravel complex questions, and her steadfast commitment to well-designed process. This background has translated directly into her work on investment process design for private wealth clients, as well as the LEVATUS Integrated Wealth Service model. Susan's investment background spans asset classes and geographies — a combination that is increasingly rare and directly relevant to the complexity her clients navigate.
October's earnings season and the market divergence that accompanied it expose a growing tension: technology giants are spending billions on AI that could disrupt some of their established revenue streams, while other companies may already be benefiting from the capabilities those investments create.