The global artificial intelligence boom has reached a decisive inflection point. After years of experimentation and speculative enthusiasm, investors are now asking a blunt question: Where are the profits? With capital spending accelerating at historic levels, the AI sector is being pushed out of its promise-driven phase and into an era defined by measurable returns.
That pressure was underscored this week after Goldman Sachs warned that the world’s largest technology companies are collectively spending hundreds of billions of dollars annually on AI infrastructure. Hyperscale data centers, advanced chips, power-hungry training clusters, and proprietary models have turned AI into one of the most capital-intensive technology cycles in modern history. For investors, the implication is clear: valuations now depend less on future narratives and more on near-term cash generation.
Yet revenue growth from AI products is lagging far behind investment growth. While leading firms report AI-driven revenue streams in the single-digit or low double-digit billions, their capital expenditures often exceed $40–$60 billion per year individually. The gap between spending and monetization has become a growing source of market anxiety, particularly as interest rates remain elevated and tolerance for long-duration bets shrinks.
“AI investment has changed dramatically and is entering a ‘show me the money’ era,” says Jiahao Sun, CEO of FLock.io. “Investors are no longer looking for potential future capabilities. They want measurable ROI, strong current user bases, and real-world utility.” According to Sun, this shift explains why enthusiasm around broad, centralized AI platforms is cooling, while attention is quietly moving toward alternative architectures.
Market signals are already reflecting that rotation. In recent trading sessions, Bitcoin has outperformed several high-profile technology equities, reinforcing the idea that capital is flowing toward decentralized assets and foundational infrastructure rather than pure application-layer AI plays. For some investors, this is less about speculation and more about hedging against concentration risk in Big Tech–dominated ecosystems.
Sun argues that regulation is accelerating this structural rethink rather than slowing it. “Governments are trying to regulate Big Tech by hiring Big Tech. This is a sovereignty trap,” he says. “You cannot achieve data independence using the infrastructure of the data monopolies.” As AI becomes embedded in healthcare, finance, defense, and public administration, questions of data ownership and national control are becoming inseparable from questions of profitability.
Looking ahead to 2026, Sun believes the winning strategy will be vertical integration. “The thesis is owning the compute, the data, and the model weights end-to-end,” he says. This mirrors a broader shift already visible in fintech and payments, where companies are abandoning the so-called “rent-a-charter” or outsourced-infrastructure model. By owning their rails, firms gain better unit economics, regulatory clarity, and insulation from platform risk.
The same logic now applies to AI builders. “Renting intelligence from centralized APIs is a categorical liability,” Sun warns. Dependence on a handful of providers exposes developers to pricing shocks, policy changes, and opaque governance decisions. Decentralized AI infrastructure, by contrast, offers a path toward cost predictability and long-term strategic control.
For markets, the message is sobering but clear. AI is no longer being judged by how impressive it appears in demonstrations, but by how efficiently it turns capital investment into durable cash flow. The period of open-ended patience has ended. What follows may be less speculative and less driven by narrative momentum—but for investors demanding evidence rather than promises, it may finally offer a sustainable foundation.
Korean version available here.
