A nonprofit called Current AI is building open infrastructure for AI agents to talk to each other without paying a toll to Big Tech. Think TCP/IP but for AI. If they succeed, the companies that control AI access today could lose their grip on a market projected to reach $1.3 trillion by 2030, according to Goldman Sachs.
Why This Matters Right Now
The web was supposed to be free. Tim Berners-Lee gave away HTTP and HTML so anyone could build on top of them. That decision created trillions of dollars in wealth spread across millions of businesses worldwide. Then came the platform era. Google, Apple, Facebook, and Amazon built walls. They took the open web and turned it into tollways.
AI is following the exact same pattern. OpenAI is valued at $300 billion, according to Reuters. Anthropic raised $9 billion from Amazon. Google has Gemini locked inside its own products. The infrastructure for AI agents to operate is controlled by a small number of private companies charging per call, per token, per seat.
Current AI wants to break that model. They’re building what they call the World Wide Web of AI: open protocols that let AI agents communicate, share data, and work together without routing through a corporate middleman. No tollbooth. No rate limits set by shareholders. Free, just like the original web.
The Real Play Here
Most people see this as an idealistic nonprofit project. I see it as the biggest threat to incumbents in tech right now.
When TCP/IP became the standard protocol for the internet, every business that built on top of it got rich. The protocol itself was free. The value sat in the applications and services layered on top. Email. E-commerce. Video. Social networks. All of them worth nothing without the open foundation underneath.
If Current AI builds the TCP/IP equivalent for AI agents and it gets adopted, the same wealth transfer happens again. The protocol is free. The businesses built on top capture enormous value. The losers are the companies selling access to the protocol itself today.
OpenAI charges roughly $15 per million output tokens for GPT-4o, according to their published pricing page. Anthropic charges $15 per million output tokens for Claude Sonnet 4. These prices are not permanent. When Linux became free, it killed the Unix licensing business almost completely. The same logic applies here.
According to Andreessen Horowitz, AI infrastructure costs consume 60 to 70 percent of early revenue for AI startups. A free, open protocol for AI agent communication would collapse that cost structure fast. That’s terrible news for the incumbents. It’s excellent news for everyone building on top of them.
The rich mindset here is simple. Watch who controls the rails. When the rails become free, the value moves up the stack. That’s where you want to be positioned before the shift happens, not after.
If you’re already building content or products on top of AI tools, something like InVideo AI lets you create professional video content without touching the infrastructure layer at all. That’s exactly the kind of application layer business that wins when the protocol becomes a commodity.
What This Means for You
Here’s what I would do with this information right now.
First, stop treating today’s AI access prices as permanent. They’re not. The open source movement drove down the cost of software to near zero over two decades. The same pressure is building in AI infrastructure at a much faster rate. Meta released Llama in 2023. Mistral released models matching GPT performance at a fraction of the cost in 2024. Current AI is the next step in that compression.
Second, build on the application layer. You don’t want to compete with Current AI or OpenAI on infrastructure. You want to use both as the foundation for something specific that people will pay for. Specific beats general. Targeted beats broad. Every time.
Third, watch which major platforms adopt Current AI’s protocols first. If one big player integrates their standard, others follow fast. That’s the moment the market tips. The businesses already integrated at that point get a meaningful lead.
Fourth, keep your software costs low while the infrastructure battle plays out. AppSumo regularly runs lifetime deals on AI tools and software that let you lock in today’s prices before the market stabilizes around winners. That’s how you play defense while others fight over the rails.
Current AI’s nonprofit status is not a weakness. It’s structural freedom. Tim Berners-Lee could give the web away because CERN didn’t need a return on investment. Current AI has the same latitude. They don’t answer to a board demanding monetization. That makes them more credible to potential adopters and more dangerous to incumbents, not less.
The Bottom Line
The internet was built on free protocols. The companies that got rich weren’t the ones who invented TCP/IP. They were the ones who built on top of it first and fastest. Current AI is making the same bet for AI agents. If they win, the $300 billion valuations of today’s AI infrastructure companies are built on sand. The next wave of wealth gets created one layer up. Get there early or pay full price later.
Frequently Asked Questions
What exactly is Current AI building?
Current AI is a nonprofit building open protocols for AI agent communication, similar to how HTTP and HTML created the open web. The goal is to let AI agents from different companies and platforms work together without routing through a single corporate gatekeeper. It’s the plumbing layer for AI, designed to be free and open to anyone.
How does this threaten OpenAI and Anthropic?
OpenAI and Anthropic sell access to their AI models through paid APIs. If an open protocol becomes the standard for AI agent communication, it creates competitive pressure on closed API pricing the same way Linux pressured Unix vendors. Neither company disappears, but their pricing power shrinks significantly.
Who funds Current AI if it’s a nonprofit?
Nonprofits building open AI infrastructure typically receive support from foundations, academic institutions, and donors who believe open standards create broader public value than closed commercial alternatives. The same model funded early internet standards through CERN and DARPA, which is the direct historical precedent here.
Will the World Wide Web of AI actually be free?
The protocol itself would be free, exactly like HTTP is free today. Businesses built on top of it are not. This is the standard open source model where free infrastructure creates conditions for expensive, valuable applications built by private companies on top. The protocol being free doesn’t stop people from making money. It just moves where the money gets made.
What should someone building an AI business do right now?
Build on the application layer, not the infrastructure layer. Infrastructure is becoming a commodity faster than the mainstream expects. Watch Current AI’s adoption closely and integrate early if their protocols gain traction among major platforms. Position yourself where users pay for specific outcomes, not for access to the underlying pipes.
