The generative AI explosion’s dust has yet to settle, but a new wave is already hitting the shores. Companies treating models as core infrastructure are outcompeting those still running pilots for business tools treated as SaaS.
In tech, the infrastructure layer is whatever layer becomes the standard that everyone else builds upon. Once, the backbone of a company was the facilities. Then with digital transformation, it was IT. Entering into 2026, treating AI as SaaS is a massive pitfall as AI becomes the foundation of business operations.
The true value of this shift will be captured not the application layer, but at the model layer. In the coming decade, the most valuable assets won’t be the apps on our screens, but the specialised models they run on. To demonstrate this point, this blog draws on the wonderful world of web3 to create an analogy that compares fine-tuned AI models with Layer 2 blockchains.
[ 👋 Hi there! If you’re here to find out more about FLock.io, follow us on X, read our docs and sign up to AI Arena on train.flock.io]
The first wave was efficiency tools, the second is reinvention
There have so far been two waves. The first wave strived towards efficiency and cost-cutting. In the second wave, businesses undergoing a new kind of digital transformation are reinventing entire processes.
Many CEOs piloted AI early on in the hype, only to decide that it was an optional tool. It was a chatbot, plug-in or dashboard that could shave off some of the overheads and save time if the workforce knew how to use it properly. How times have changed! Entering into 2026, forward-thinking leaders are building robust data pipelines, training their teams up and aligning their multi-year strategy around models.
In medicine, AI’s impact has been tremendous. AstraZeneca’s AI is already embedded in national healthcare systems, while many of its competitors are still simply optimising internal R&D pipelines. The NHS is building a world-first AI system based on its IT software, the NHS Federated Data Platform. In financial services, JPMorgan has announced its blueprint to become the world’s first fully AI-powered megabank. Amazon uses AI extensively as a foundational element of nearly every part of its operations to reimagine the customer experience, from internal logistics to customer-facing services.
The UK government is planning ahead – anticipating, and paving the way for, the business world to follow suit. In January, parliament set out a blueprint to turbocharge AI and unleash it across the UK. September saw the historic UK-US tech prosperity deal signed by Trump and Starmer.
The third wave will be decentralised AI
The third wave is forecasted to be decentralised AI (DeAI). A Forbes article this year argued that AI infrastructure “must become more democratised, domain-specific and decentralised.”
DeAI is quietly building an ethical, on-chain future, where AI tokens represent true ownership in transparent, community-governed ecosystems. It boosts privacy, security, and fairness by keeping data local, offering transparent operations, and rewarding contributors directly. And the greatest part? None of it compromises scalability!
Keeping costs low as you implement AI infrastructure
You won’t need your own compute, since cloud is easy to rent from providers like AWS and Alibaba Cloud. The problem is that traditional cloud AI costs compound as you scale. This is because they cannot price inference below hosting costs without making a loss.
FLock.io’s launchpad is almost here to solve that problem. It’s called FOMO, and finalises our evolution into a full-cycle AI platform. Its power lies in its disruptive pricing. By leveraging the token economy, its franchise model presents a unique solution to the cost obstacle that hampers traditional cloud or gateway providers.
Stay tuned for our launch!
What’s needed for core AI infrastructure
- Clean data pipelines
Your company’s data is no longer just a record of your past. It’s the training material that will tailor and fine-tune your future models. If your data is messy and siloed, your infrastructure is flimsy.
- AI-trained managers
Employees need to be retrained to work with AI in the loop, to understand the capabilities and limitations of your model.
- Strong governance and security
As AI takes on a bigger role, the risks get magnified. Bias, lack of transparency and misuse must all be considered. By decentralising governance, you ensure the model is transparent and aligned with the company’s values.
The “1 model = 1 chain” web3 analogy
In web3, a Layer 1 blockchain like Ethereum is a foundational protocol. You don’t build your own blockchain to launch an app, you build on top of the existing infrastructure. If the chain is successful and the apps are heavily used, the network thrives. AI models are becoming comparable to Layer 1s.
The model is the chain.
Whether it’s a massive base model like Meta Llama 3 or a specialised fine-tune for medical research, the model is the foundational infrastructure. It is the mainnet where intelligence is generated.
Base models are like a Layer 1
Base models (like Llama 3 or Qwen) are functioning as the L1s of AI. They provide the universal compute logic and reasoning capabilities that everything else requires. Just as an L1 secures a network, these are the mainnet where intelligence is generated.
Fine-tuned models are like a Layer 2
In crypto, Layer 2s (like Arbitrum) make the L1 faster, cheaper and more specialised for certain transaction types. They benefit from the security of an L1 but are tailored to a use case.
In AI, a general base model is powerful but expensive and broad. A fine-tuned model can be customised to use cases like medical research. Like an L2, they are more efficient for its specific tasks, has lower inference costs, and its on top of the base model’s foundational logic.
Inference is the gas.
Here we compare the costs of running the chain and the model.
In a blockchain, you pay a gas fee to execute a transaction. In AI, every time a user asks an AI agent a question, it consumes compute – this is called inference. In the AI-as-infrastructure wave, inference is the gas that powers the model’s economy.
The token is the capture value.
Just as ETH captures the value of the Ethereum ecosystem, the model token (issued via FOMO) captures the value of the model’s usage. As more developers build on a specific model, and as inference calls increase, the revenue share flows back to the token holders.
More about FLock.io
Find out more about FLock.io by reading our docs and blogs. See a sneak peak of new launches coming in Q4 2025 and our FLock 2025 earnings report: performance & growth overview. Stake, train and earn on train.flock.io.
For future updates, follow FLock.io on X.
