The next AI infrastructure cycle will not be defined by how agents access data. It will be defined by how intelligence is priced, owned, and settled at machine speed.
By David Arnež, Co-Founder & CEO at Inflectiv
For most of 2024 and 2025, the question that defined AI infrastructure was whether agents could reach the data they needed. The answer turned out to be yes. Model Context Protocol opened the door. Connectors followed. By the end of 2025, every serious enterprise stack assumed agents would have read access to internal systems. Access stopped being the interesting problem.
The interesting problem now is what happens after the agent reaches the data.
Because agents are no longer just reading. They are buying. They are paying for inference, for tooling, for context, and increasingly for the intelligence that lets them make a decision. The major payment networks have shipped agent commerce products over the last twelve months. x402 has gone from a research concept to live transaction flow across multiple wallet stacks. The agent is not a hypothetical buyer. It is an active participant in markets that did not exist eighteen months ago.
The infrastructure around those transactions, however, is not built. There is no standard way to price a piece of intelligence per query. There is no standard way to attribute the value of that intelligence back to whoever produced it. There is no standard way for an agent to verify that what it just bought is what it asked for. The plumbing is missing.
This is the gap that defines the next cycle.
Why aggregation cannot solve this
The instinct in the market is to assume the existing data infrastructure will absorb the agent economy. It will not. The reason is structural, not technological.
Platforms like HuggingFace, Kaggle, and Snowflake were designed for human users running pipelines on a quarterly cadence. Their economics assume seat licenses, subscription tiers, and human decisions about what to buy. None of that maps to an agent that needs to query a thousand sources in a second, settle each one programmatically, and pass on attribution to whatever it produces downstream.
The problem is not that those platforms have failed to add agent features. Several have. The problem is that the underlying contract between buyer and seller assumes a human in the loop. Once you remove the human, the contract has to be rewritten. That rewrite is the new infrastructure layer.
What it means for intelligence to be a balance sheet item
When intelligence can be priced, attributed, and settled at machine speed, it stops behaving like an input and starts behaving like an asset.
Three things change.
First, intelligence compounds. A dataset that earns revenue every time an agent queries it has a yield. The producer who built it has an incentive to keep improving it. The dataset accrues a track record, which lets the market price its reliability over time. This is not a property of static data. It is a property of intelligence that exists inside a market.
Second, intelligence becomes ownable across systems. Today, when a piece of operational knowledge moves from one platform to another, it loses its lineage. The provenance is the contract, not the file. An intelligence asset with on-chain attribution can move freely. It carries its history with it. That changes who can produce it, who can sell it, and who can extend it without permission.
Third, intelligence becomes collateralizable. Once an asset has a price, a yield, and a verifiable history, the standard primitives of capital markets become available. Intelligence pools can be staked. Producers can borrow against future query revenue. Buyers can hedge against quality drift. None of this is theoretical. The same playbook applied to compute, to liquidity, to attention. It will apply to intelligence next.
The shift the market is underestimating
The dominant framing of the agent economy still treats intelligence as a cost center. Pay for the data, feed it to the agent, get the output. That framing is going to look quaint within eighteen months.
The protocols that win this cycle will not be the ones that store the most data or connect to the most systems. They will be the ones that let intelligence be owned, priced, and settled at machine speed, with attribution that survives composition and provenance that survives transfer. That is not a data infrastructure problem. It is a market structure problem.
The first agent economies are already being built on that assumption. The protocols designed for human-speed software will not be the ones that absorb them.
https://x.com/inflectivAI/status/2040052519854817476
Proprietary Knowledge Is the Real Enterprise Moat
Enterprises are not sitting on a lack of AI models. They are sitting on decades of proprietary knowledge that agents still cannot properly use.
Contracts, product documentation, customer records, support tickets, research archives, compliance documents, market intelligence, internal reports, supply chain data, financial models, operational playbooks, and sector-specific expertise all carry enormous value.
But most of that value is invisible to agents.
It was created for human storage and retrieval, not machine execution. It sits across disconnected systems, buried in formats that are hard to query, hard to verify, and hard to reuse. The knowledge exists, but it does not yet behave like infrastructure.
That is the opportunity.
The companies that convert proprietary knowledge into structured, agent-ready intelligence will not simply improve internal productivity. They will create a new strategic asset layer. Their agents will become more reliable because they operate on better context. Their workflows will become more efficient because knowledge can persist and improve. Their data will become more valuable because it can be accessed, updated, governed, and monetized.
In an agent-driven market, the question is no longer only “which model are you using?”
The stronger question is: “what does your agent actually know, and do you own that intelligence?”
From Passive Data to Compounding Intelligence
Most companies still think about data as something they store, back up, occasionally search, and protect. That mindset made sense in previous cycles. But agents change the function of data.
When an agent can act on information, data becomes operational. When an agent can improve that information, data becomes compounding. When other builders, agents, or applications can access that intelligence, data becomes economic.
This is where the investment opportunity becomes much larger than file storage or retrieval.
A structured dataset can serve one agent or many agents. It can stay private or become accessible. It can power APIs, workflows, assistants, analytics, automation, and vertical applications. It can improve as agents and users interact with it. It can become part of a larger intelligence market.
That creates a different kind of network effect.
Every useful dataset increases the value of the platform. Every agent query creates more demand for structured intelligence. Every write-back loop can improve the underlying dataset. Every contributor adds supply. Every developer building on top of that intelligence expands distribution.
This is how data moves from passive resource to compounding infrastructure.
Inflectiv is not competing for the model layer. We are building the intelligence layer beneath it, where proprietary information becomes structured, reusable, verifiable, and monetizable.
Why Verifiability Matters
If agents are going to act on data, trust becomes part of the infrastructure layer.
It is not enough for data to be available. Teams need to know where it lives, whether it has changed, who can access it, and whether another system can depend on it. Provenance becomes a product feature. Verifiability becomes a requirement. Data control becomes part of the business model.
This is why our work with Walrus matters. Walrus gives Inflectiv a verifiable storage layer for agent data, replacing centralized storage assumptions with cryptographically verifiable and immutable infrastructure. The Walrus case study now reports over 7,000 datasets stored on Walrus, alongside a 60% cost reduction compared to AWS S3.
https://walrus.xyz/
That matters because the agent economy will not be built on invisible blobs sitting in black boxes. It will require stronger relationships between data structure, storage, identity, access, and provenance.
For investors, this is an important distinction.
The opportunity is not simply that more data will be stored. The opportunity is that more data will become usable by agents, trusted by systems, and valuable to markets.
That is a much larger category.
The Investment Case for Structured Intelligence
From a founder’s perspective, the pattern is familiar.
The strongest technology markets emerge when a new behavior creates demand for new infrastructure. Cloud became inevitable once software needed to scale globally. Payments infrastructure became inevitable once commerce moved online. Data infrastructure became inevitable once companies needed to operationalize information at scale.
Agents create the next infrastructure demand.
They need structured context. They need persistent memory. They need verifiable sources. They need controlled access. They need data that can be updated, reused, and monetized. They need a layer between raw information and autonomous execution.
That is the layer Inflectiv is building.
We turn fragmented information into structured datasets that agents can actually use. A company, builder, researcher, or community can bring in documents, PDFs, spreadsheets, reports, or internal knowledge and convert them into queryable intelligence. Once structured, that dataset becomes something agents can access, build on, and improve over time.
This creates a path toward a new kind of data economy.
Creators and organizations can turn knowledge into assets. Developers can access intelligence without rebuilding it from scratch. Agents can query what they need at runtime. Enterprises can unlock proprietary knowledge without giving up control. Ecosystems can form around useful datasets, usage, and demand.
That is why this category matters.
It is not another AI wrapper. It is infrastructure for the agent economy.
The Companies That Structure Intelligence First Will Lead
The next wave of AI will not be defined only by who has access to the best model. Models will keep improving, and access to them will become more common.
The durable advantage will come from the intelligence layer around them.
Who owns the best proprietary knowledge?
Who can structure it?
Who can verify it?
Who can make it usable by agents?
Who can improve it over time?
Who can monetize it without losing control?
These are the questions that will define the next phase of AI infrastructure.
The internet made information searchable. Cloud made software scalable. Blockchain made digital ownership programmable. AI made generation cheap. Agents will make structured intelligence actionable.
Once that happens, data stops being a passive resource. It becomes infrastructure, memory, distribution, and economic value.
The companies that structure their intelligence now will own the agent economy later.
At Inflectiv, we are building exactly that layer.
The model era asked: what can AI generate?
The agent era asks: what does the agent actually know, and can it trust it?
That is the question every founder, enterprise, investor, and ecosystem will have to answer.
Further Reading
Anthropic: Introducing the Model Context Protocol
Google: MCP Toolbox for Databases
Walrus × Inflectiv Case Study
Inflectiv App
About the Author
David Arnež is the Co-Founder & CEO of Inflectiv, a data infrastructure platform turning unstructured information into structured, queryable intelligence for AI agents. He is a strategist and entrepreneur with over a decade of experience across Web2 and Web3. Before Inflectiv, David led two startup cycles from inception to exit, Foora and RingoX, and raised over $1M in capital. He is also a PhD Candidate who has delivered more than 200 lectures on innovation and provided over 2,000 hours of consultancy to early-stage startups.
