Investors remain fixated on the coming OpenAI and Anthropic IPOs, yet this focus overlooks a deeper structural change underway in the AI economy. With machine intelligence costs falling fast and “capex vigilantes” now scrutinizing hyperscaler budgets closely, the real momentum is building in the infrastructure supply chain and among enterprise adopters, not the frontier labs.
Quick Take
- The narrative shift: Institutional capital is pivoting away from the frontier model hype cycle to focus on the deeper, unglamorous layers of the AI value chain, including foundries, memory, and cooling.
- The tokenomics effect: Plummeting AI token prices threaten model-builder margins but create massive demand elasticity, shifting value capture to downstream enterprise adopters.
- The capex threat: “Capex vigilantes” are emerging, demanding tangible ROI on the billions hyperscalers are pouring into GPU clusters and data centers.
- The under-reported angle: Trust, security, and workforce adaptability are becoming measurable competitive moats for AI deployment, effectively replacing raw model parameters as the ultimate metric of market durability.
What happened
In recent market briefings, financial strategists from firms like iCapital and J.P. Morgan are aggressively rotating their AI investment thesis. Rather than treating the eventual public debuts of OpenAI and Anthropic as the zenith of the AI trade, analysts are urging markets to look at the broader technology stack.
This updated thesis argues that massive dispersion in stock performance—exacerbated by a higher-for-longer interest rate environment—requires a surgical approach to AI exposure. Capital is shifting toward semiconductor foundries, memory chip manufacturers, and companies across Asia—including China and Vietnam—that serve as the physical backbone of the AI supply chain.
Furthermore, a consensus is building that the ultimate financial winners of the AI boom will not be the “builders” of frontier models, but the “adopters.” Analysts predict that non-tech enterprises using cheap cognitive power to drive unprecedented productivity gains will likely see the greatest margin expansion in the coming decade.
Why it matters now
This capital rotation strikes at the very heart of how large language models (LLMs) are scaled and deployed. Until now, the AI race has been funded by a seemingly infinite pool of capital willing to underwrite massive compute costs. But the emergence of AI “capex vigilantes”—investors actively scrutinizing whether cloud giants are generating actual software revenue from their hardware binges—signals that the blank-check era for massive GPU clusters is ending. If the market demands faster ROI, cloud providers may be forced to alter their infrastructure roadmaps, which could directly bottleneck the training runs for next-generation frontier models.
Simultaneously, the AI industry is undergoing a brutal “tokenomics” stress test. Open-source models and fierce competition between OpenAI, Anthropic, and Google are driving the cost of inference down to the floor. While falling token prices threaten the profit margins of proprietary model labs, they unlock massive demand elasticity. Cheaper AI means intelligence can be embedded into nearly every software layer, shifting the economic center of gravity from the laboratories inventing the intelligence to the corporations applying it.
Who is most affected
| Stakeholder | Impact | Why |
|---|---|---|
| Frontier Model Labs | Margin Compression | Falling token prices and relentless open-source competition commoditize raw intelligence, severely limiting pricing power. |
| Enterprise Adopters | Margin Expansion | Cheaper inference costs allow traditional companies to deploy AI agents and automation, drastically cutting operational costs. |
| Foundries & Memory (e.g., TSMC, Micron) | Sustained Demand | Even if model labs struggle with software profitability, the raw volume of inference and training requires vast memory and silicon output. |
| Hyperscalers | Capex Scrutiny | “Capex vigilantes” will punish cloud giants if their multi-billion dollar GPU investments fail to translate into durable enterprise cloud revenue. |
About the analysis
Researched and written with AI assistance under the editorial responsibility of the i10X News team. The title image is an AI-generated image. Read our AI Use Policy. Spotted an error? Request a correction.
This is an independent, research-based i10x analysis built for founders, CTOs, and investors tracking the evolution of the AI economy. It synthesizes current market strategies, value-chain economics, and shifting institutional perspectives on AI capital allocation.
- Beyond OpenAI and Anthropic: Where iCapital sees the next wave of AI winners
- Tokenomics: The Economics of the AI Boom
- Market Pulse: Will AI Capex Vigilantes Emerge?
- The Bridge Ep. 23: Can AI Deliver Returns, Not Just Hype?
- How trust may drive a competitive edge in the age of AI
- Stock winners and losers as Anthropic passes OpenAI as hottest AI upstart
i10x Perspective
We are witnessing the rapid commoditization of cognition. The market is slowly realizing that building the smartest frontier model is a capital-intensive utility game, while the real wealth is generated by those who supply the picks and shovels—or use the utility to reinvent legacy industries. As capex vigilantes tighten the leash on infrastructure spending, the next phase of the AI race won’t be defined by parameter counts, but by the relentless economics of token elasticity and secure enterprise integration.
