Who Owns AI?

Mapping capital and influence across the frontier AI supply chain

Ownership positions across the frontier AI supply chain will determine which countries and companies accrue the economic and geopolitical power that advanced AI systems create. The dollar figures matter less than the identity of who is writing the checks and on what terms.

Download the new report by Nikhil Mulani, Patrick Fitz, and Kristina Fort.

Who Owns AI? maps investor positions across the frontier AI supply chain, analyzes how ownership translates into influence, and recommends how sovereign and private capital can be deployed to strengthen international security and democratic ownership of the AI stack. The report covers venture capital, growth and private equity, corporate strategics, and sovereign wealth funds across the Middle East, East Asia, North America, and Europe, with detailed ownership case studies of OpenAI, Anthropic, and TSMC.

What we found

Ownership is not investment volume. Influence flows through five distinct mechanisms: direct equity governance, informal influence, physical asset control, sovereign leverage, and privileged access to information, infrastructure, and talent. The defining deals of the past eighteen months are hybrid compute-plus-equity arrangements — compute has become equity by other means.

Value capture across the stack is asymmetric and time-varying. The near-term apex sits where moats, margins, and growth align: chip design (Nvidia ~75% gross margin), advanced-logic manufacturing (TSMC ~59%), and high-bandwidth memory (63–67% — now exceeding TSMC foundry margins for the first time since 2018). Data centers and energy sit a tier lower at utility-like returns; model developers run at negative or minimal margins with the widest path-dependent uncertainty. Each layer of the stack now funds the next: chokepoints produce geopolitical leverage, high-margin segments produce national wealth, and the civilian stack produces military capability directly.

Positioning is concentrated and strikingly uneven. A handful of U.S. venture firms, three hyperscalers, and Nvidia dominate global private capital. Middle Eastern and East Asian sovereigns dominate state-affiliated capital. U.S. federal positions are growing fast but remain legally fragile. No European sovereign holds a private stake in OpenAI or Anthropic — European sovereigns are materially under-positioned relative to their fiscal capacity.

The window is closing within roughly five years. AI capability timelines converge on 2027–2030; physical infrastructure locks in; financial consolidation accelerates. Investors who commit during this window will see disproportionate upside. Those who wait will find the table set and the seats taken.

What we recommend

Four principles should guide democratic capital.

• Compute is currency: power-purchase agreements, fab-capacity reservations, and TPU and Trainium offtake are equity by other means; chokepoints (EUV, HBM, CoWoS-L, refined gallium, germanium, dysprosium) do more for security than diversified breadth.
• The 2026–2027 IPO window is the largest near-term entry: OpenAI's targeted ~$1T listing and Anthropic's pre-IPO at >$900B are the last meaningful opportunities to take frontier-lab positions before passive index holders displace strategic ones.
• Anti-fragility, then coordination: sovereign positions resting on a single executive order will not survive an electoral cycle, and only allied pools acting together can counterbalance Gulf and Chinese capital.
• Allocate to high-leverage but neglected segments: frontier security and interpretability, leading-edge fabs in democratic geographies, critical-minerals refining outside Chinese control, strategic data curation, and sovereign-anchored compute capacity.

The report lays out jurisdiction-specific priorities for sovereign investors — codifying CHIPS-derived equity and OSC warrants in the United States, scaling the UK Sovereign AI Unit into a £2–3B revolving facility, converting EU InvestAI pledges into binding EIB instruments, and anchoring AI infrastructure via AP-fund alternatives capacity in Sweden, among others.

For private investors, pre-Series-G venture and growth firms should make model-spec transparency, RSP-equivalent commitments, and by-law-protected safety-committee independence standing follow-on conditions at frontier labs; hyperscalers and chipmakers should tighten compute-for-equity discipline and disclose ownership clearly given SEC, FTC, and Congressional scrutiny of circular structures.

The decisions that sovereigns, investors, and policymakers make over the next eighteen months — about which deals to enter, on what terms, and with which conditions — will be pivotal in determining who benefits most from AI.

Download Who Owns AI?

The frontier AI supply chain is a priority area at Augur. Subscribe to keep up with our future updates on these topics, and drop us a note if you or your organization are interested in working together.