Opinions expressed by Entrepreneur contributors are their own.
Key Takeaways
- The AI boom has become a trillion-dollar infrastructure arms race, yet most companies are seeing little return and growing dependence on a handful of centralized platforms.
- As cloud costs rise and trust issues deepen, founders are realizing that building on closed AI systems creates structural risk.
- History suggests the next wave of innovation will come from distributed, sovereign AI models that shift control, resilience and ownership back to users
The race to build artificial intelligence has become an infrastructure arms race. In recent months, Oracle signed a $300 billion deal with OpenAI. Microsoft spent $35 billion on AI infrastructure in just three months. Nvidia poured $100 billion into OpenAI, which then used that money to buy Nvidia’s chips — a flow of investments that has some experts questioning the entire model.
The numbers are staggering. Total AI spending from major U.S. companies is expected to reach $1.1 trillion between 2026 and 2029. Yet despite this, a Massachusetts Institute of Technology report found that 95% of organizations are getting zero return. Even Sam Altman, CEO of OpenAI, has acknowledged that an AI bubble may be underway.
The scale seems impossible to compete with. But history suggests that when an industry becomes this closed and concentrated, it’s ripe for disruption.
The limits of centralized AI
For founders, centralized AI systems pose serious risks. You’re not just dependent on OpenAI’s or Google’s infrastructure — you’re betting your company on their ability to sustain trillion-dollar spending indefinitely. If the bubble bursts, the platforms you’ve built on could be repriced, restructured or shut down entirely.
When cloud providers raise prices, thousands of startups see their margins evaporate overnight. When OpenAI changes its API terms, entire businesses built on top of it will face existential risk. Founders are essentially building features that can be turned off with a policy change.
Beyond that, there’s the trust problem. When you use ChatGPT, you have no idea what data it was trained on or whether you can trust the output. The entire process happens in OpenAI’s servers, hidden from view. Every document analyzed by Claude, every image generated by Midjourney — all of it flows through corporate servers. Users are pouring intimate details of their lives, businesses and ideas into systems they don’t control.
The distributed alternative
Rather than asking how to build bigger systems, a growing number of researchers are asking how to build smarter ones — systems that are decentralized, transparent and resilient. AI doesn’t need to live exclusively in massive data centers. It can run across networks, on devices people already own.
Consider this: There are over 2 billion smartphones globally, each with processing power that would have been considered a supercomputer decades ago. Add laptops, tablets and other devices, and you have a massive, distributed computing network sitting idle most of the time. What if AI could tap into that?
This is the bet behind a new generation of distributed AI platforms, including projects like Gradient, that are developing “sovereign AI.” Instead of concentrating power in a single model or platform, AI becomes something that anyone can run, contribute to and verify. These systems are already beyond theory and can run sophisticated AI models — the kind that typically require massive data centers — across ordinary consumer hardware. These systems can process hundreds of responses per second while coordinating over a million participating devices globally.
When countries invest billions in AI infrastructure, or when financial institutions demand on-premises deployments, they’re responding to a fundamental problem: dependence. Sending proprietary data through another company’s servers is a strategic vulnerability. For governments, it means foreign jurisdictions potentially accessing sensitive information. For enterprises, it means building competitive advantages on infrastructure controlled by competitors. Distributed systems address this by shifting the locus of control. A single entity holds the keys.
Lessons from history
If this sounds familiar, it should. We’ve seen this pattern play out repeatedly across industries. The early internet was controlled by a few telecom companies and closed networks like CompuServe and AOL. Then it opened up, and an explosion of innovation followed. Banking followed a similar trajectory. Centralized institutions controlled all financial services until distributed fintech platforms and mobile payments democratized access.
Transportation went from centralized taxi dispatch systems to distributed rideshare networks. Energy is shifting from centralized power grids to distributed solar and renewable microgrids. Media moved from centralized broadcast to distributed content creation on platforms like YouTube and podcasts.
In each case, the companies that understood openness could scale faster than control built enduring businesses. The same pattern is emerging in AI.
What this means for entrepreneurs
The critical insight for founders is that closed systems always open up eventually. And when they do, the opportunities are massive. The winners aren’t necessarily those with the most capital. They are the ones who recognized the shift and built for the new paradigm.
Right now, AI is a closed system dominated by a handful of players spending trillions to maintain their position. That’s not sustainable — economically, politically or technologically. As that system opens up, new opportunities emerge.
In any industry where control is centralized, there’s an opportunity to distribute it. Where systems are fragile, there’s a chance to make them resilient. Where value concentrates at the top, there’s potential to spread it more broadly. Startups that design for independence — through open participation, clear incentives and user control — can move faster and last longer than those renting access from giants.
Sign up for the Entrepreneur Daily newsletter to get the news and resources you need to know today to help you run your business better. Get it in your inbox.
Key Takeaways
- The AI boom has become a trillion-dollar infrastructure arms race, yet most companies are seeing little return and growing dependence on a handful of centralized platforms.
- As cloud costs rise and trust issues deepen, founders are realizing that building on closed AI systems creates structural risk.
- History suggests the next wave of innovation will come from distributed, sovereign AI models that shift control, resilience and ownership back to users
The race to build artificial intelligence has become an infrastructure arms race. In recent months, Oracle signed a $300 billion deal with OpenAI. Microsoft spent $35 billion on AI infrastructure in just three months. Nvidia poured $100 billion into OpenAI, which then used that money to buy Nvidia’s chips — a flow of investments that has some experts questioning the entire model.
The numbers are staggering. Total AI spending from major U.S. companies is expected to reach $1.1 trillion between 2026 and 2029. Yet despite this, a Massachusetts Institute of Technology report found that 95% of organizations are getting zero return. Even Sam Altman, CEO of OpenAI, has acknowledged that an AI bubble may be underway.
