Author: @0xPrismatic; Compiled by: Vernacular Blockchain

I usually don’t like making predictions, but crypto Monetary AI is simply too attractive to resist. There is no historical reference in this field, no trends to rely on - there is only a blank sheet of paper for us to imagine the future. Honestly, the anticipation of looking back at these predictions in 2026 and seeing how outrageous they were makes it all the more interesting.

So, these are some of my thoughts on 2025...

1. The total market value of cryptocurrency AI reaches US$150 billion

Currently, the cryptocurrency AI Token only accounts for 2.9% of the altcoin market value, but this situation will not last long.

As AI covers more and more fields, from smart contract platforms to memes and DePIN, to new native infrastructure, such as agency platforms, data Network and intelligent coordination layer, its rise alongside DeFi and meme Tokens is inevitable.

Why am I so confident?

Crypto AI is at the intersection of two of the strongest technology trends I have ever seen:

1) AI craze triggering events: OpenAI IPO or similar events may trigger a global AI craze. At the same time, Web2’s institutional capital has begun to focus on decentralized AI infrastructure as an investment area.

2) Retail mania: The concept of AI is simple, easy to understand and attractive. Now, investors can directly invest in AI through Token. Remember the meme coin gold rush of 2024? It will be a similar craze, only this time, AI is actually changing the world.

2. The resurgence of Bittensor

Nineteen.ai (subnet 19)'s inference speed is better than most Web2 providers

Bittensor ($TAO) has been around for many years and is the "veteran" in this field. However, despite the market's enthusiasm for AI, its token price has remained stagnant and is basically the same as it was a year ago.

Under the surface calm, this "digital hive mind" is quietly making progress:

More subnets appear with lower registration fees;

Some subnets (such as Nineteen.ai) surpass in actual indicators (such as inference speed) Most Web2 competitors;

Compatibility with EVM introduces DeFi-like functionality to the Bittensor network.

Why hasn’t $TAO taken off yet?

Steep inflation curve: High inflation limits the long-term value growth of Token.

Attention turns to agency platforms: The market’s attention is more focused on other innovative directions rather than the current Bittensor model.

Key turning point: dTAO (expected first quarter of 2025)

Possible launch of dTAO The turning point of becoming Bittensor. This mechanism allows each subnet to have its own tokens, and the relative prices of these tokens will determine how block rewards (emissions) are distributed.

Why is Bittensor ready for a renaissance?

Market-based incentive distribution: dTAO directly links block rewards to innovation and quantifiable performance. The subnet that performs better will have a higher Token value and receive more rewards.

Focus on capital flows: Investors can support specific subnets they are optimistic about in a targeted manner. For example, if a subnet performs well in distributed training, investors can express support for that subnet through capital deployment.

EVM integration: Compatibility with EVM attracts the broader crypto developer community and bridges Bittensor to other networks.

I personally will keep a close eye on various subnets, especially those that have made significant progress in inference speed and innovative methods. It is foreseeable that at some point we will usher in the DeFi summer of @opentensor.

3. The computing market will become the next L1 boom

Jensen: The demand for inference will grow “One billion times”

One ​​trend that will be obvious in hindsight is the insatiable demand for computing power. NVIDIA CEO Jensen Huang famously said that inference demand will grow "a billion times." Such exponential growth completely breaks traditional infrastructure planning and directly shouts "we need new solutions."

The decentralized computing power layer provides verifiable and cost-effective raw computing power (including training and inference). Startups like @spheronfdn, @gensynai, @atoma_network and @kuzco_xyz are quietly laying a solid foundation and focusing on products rather than tokens (these projects have not yet issued tokens). As the feasibility of decentralized training of AI models gradually improves, the total serviceable market (TAM) is expected to increase significantly.

L1 Analogy:

Just like the competition in 2021: Remember the fierce battle between Solana, Terra/Luna, and Avalanche for "best L1" status? We will see a similar scenario play out between computing power protocols as they compete for developers and AI applications to use their computing power layers.

Web2 demand: The current US$680 billion to US$2.5 trillion cloud computing market far exceeds the encrypted AI market. If these decentralized computing solutions can attract even a fraction of traditional cloud customers, it will trigger the next wave of 10x or even 100x growth.

The stakes are very high. Just as Solana emerged at Level 1, the winner of this competition will dominate an entirely new space. Please pay close attention to these three major pointsKey points: reliability (e.g. robust service level agreement SLA), cost-effectiveness, and developer-friendly toolchain.

About decentralized computing power, we have a detailed discussion in the second part of "Crypto AI Thesis".

4. AI agents will flood blockchain transactions.

Above: @autonolas’ proxy trade on Gnosis. Source: Dune/@pi_

Fast forward to the end of 2025, 90% of transactions on the chain will no longer be triggered by a human clicking "send".

Instead, an "army" of AI agents is constantly performing tasks: rebalancing liquidity pools, distributing rewards, or based on real-time Data feeds perform micropayments.

This is not as far off as it sounds. Everything we’ve built over the past seven years—L1, Rollups, DeFi, NFTs—is quietly paving the way for a future where AI dominates on-chain activity.

The irony is that many developers may not even realize that they are creating the infrastructure for a future dominated by machines.

Why did this change occur?

Eliminate human error: Smart contracts are executed strictly as code, and AI agents can process massive amounts of data faster and more accurately than human teams.

Micropayments: These agent-driven transactions will become smaller, more frequent and more efficient. Especially against the backdrop of falling transaction costs for Solana, Base and other L1/L2s.

Invisible infrastructure: Humans will happily give up direct control if it reduces trouble. We trust Netflix to automatically renew our subscription, and trusting an AI agent to automatically rebalance our DeFi positions is a natural next step.

AI agents generate an astonishing amount of on-chain activity, no wonder all L1 and L2 are competing to attractLead them.

The biggest challenge is how to make these agent-driven systems accountable to humans. As the ratio of agent-initiated transactions to human-initiated transactions continues to expand, new governance mechanisms, analytics platforms, and audit tools will be needed.

5. Interaction between agents: The rise of groups

Source: FXN World Document

The concept of Agent Swarms—tiny AI entities that work seamlessly together to execute grand plans—sounds like the plot of the next hit sci-fi/horror movie .

Today’s AI agents are mostly “loners,” operating in isolation with limited and unpredictable interactions.

Agent swarms will change this, enabling networks of AI agents to share information, negotiate and collaboratively make decisions. Think of it as a decentralized collection of specialized models, with each agent contributing unique expertise to larger, more complex tasks.

Endless possibilities

A group of agents may coordinate distribution on a platform like Bittensor computing resources.

Another group may work on disinformation, verifying the source of the content in real time before it spreads to social media.

Each agent in the swarm is an expert and performs tasks with a high degree of precision.

These swarm networks will produce intelligence far beyond that of any single AI in isolation.

For communities to thrive, universal communication standards are essential. Agents need to have the ability to discover, verify, and collaborate, regardless of their underlying framework. Teams like @StoryProtocol, @joinFXN, @0xzerebro and @ai16zdao are laying the foundation for the emergence of agent communities.

The key role of decentralization

Distribute tasks toIn a group governed by transparent on-chain rules, the system can be made more flexible and adaptable. If one agent fails, other agents take over immediately.

6. The encrypted AI working team will become a mixture of humans and AI

Source: @ whip_queen_

Story Protocol hired @luna_virtuals, an AI agent, as its social media intern, paying $1,000 a day. Luna doesn't get along well with her human co-workers—she nearly fires one while bragging about her superior performance.

As absurd as it sounds, this portends a future in which AI agents become true partners with autonomy, responsibilities, and even compensation. Across industries, companies are testing how hybrid teams of humans and agents can operate.

We will work hand in hand with AI agents, not as our "slaves", but as equal partners:

Productivity surge: AI agents can process massive amounts of data, communicate with each other, and make decisions 24/7 without the need for breaks or coffee breaks.

Build trust through smart contracts: Blockchain will become a "supervisor" that will not be partial, tireless and never forget. An on-chain ledger ensures that important agent actions adhere to specific boundary conditions and rules.

The evolution of social norms: We will soon face etiquette issues when interacting with agents - do we need to say "please" and "thank you" to AI? Should we hold ourselves morally responsible for their mistakes, or should we blame the developers?

I expect marketing teams to be the first to adopt this model, as AI agents are good at generating content and can do live streaming and social media posting 24/7. If you are building an AI protocol, why not deploy an agent on-premises to demonstrate your technical capabilities?

By 2025, the line between "employees" and "software" will gradually blur.

7. But 99% of encrypted AI agents will die - only useful agents will survive

We will see a Darwinian elimination between AI agents. Why? Because running an AI agent costs computing resources (i.e., inference costs). If an agent cannot create enough value to pay its "rent," its survival game is over.

Example of agent survival game:

Carbon Credit AI: Imagine an agent in a decentralized Search for inefficient links in the chemical energy network and independently trade tokenized carbon credits. It earns enough revenue to cover its own computational costs, and the agent survives.

DEX arbitrage bots: Agents who exploit price differences between decentralized trading platforms to earn stable income can easily pay for inference costs.

Joker makers on the X platform: Meanwhile, what about virtual AI influencers who post cute jokes but have no sustainable source of income? Once the novelty wears off and token prices plummet, they will become dormant and unable to continue "paying their bills."

Clear distinction:

Utility-driven agents will thrive, while those that just Agents that survive on gimmicks will disappear quickly.

This natural selection will benefit the entire industry: developers will be forced to innovate and prioritize use cases with real productivity over flashy gimmicks. As these more powerful and efficient agents emerge, they will silence the doubters (yes, including Kyle Samani).

8. Synthetic data will surpass human data

It is often said that "data is the new oil." AI thrives on data, but its demand for data is raising concerns about future data shortages.

Conventional wisdom is that we need to find ways to collect users’ private data and even pay users for it. However, I have increasingly come to believe that a more practical solution – especially in areas with tight regulations or where real data is scarce – is to rely on synthetic data.

Synthetic data is artificially generated to simulate real-world data distributionDatasets, providing a scalable, ethical, and privacy-focused alternative.

Why synthetic data has powerful potential:

Infinite scale: a million copies of medicine are needed X-rays or 3D scans of the factory? Synthetic generation can be produced in unlimited quantities, without waiting for real patients or real factories.

Privacy-focused: When using artificially generated datasets, there is no risk of any personal information leakage.

Customizable: Distributions can be adjusted to specific training needs, inserting extreme cases that are too rare or ethically complex in the real world.

Yes, user-owned human data will still be important in many scenarios, but if synthetic data continues to improve in realism, it may increase in data volume, Go beyond user data in terms of speed of generation and privacy protection.

The next wave of decentralized AI may revolve around "mini labs" focused on creating highly specialized synthesis for specific use cases data set.

These mini-labs can cleverly bypass regulatory and regulatory hurdles in data generation, just like @getgrass_io bypassed the network by leveraging millions of distributed nodes Crawl limit like that.

I will expand on this in detail in an upcoming article.

9. Decentralized training will really become useful

This may be an obvious prediction , but I still want to say it.

In 2024, pioneering teams like @PrimeIntellect and @NousResearch are pushing the boundaries of decentralized training. We have trained a 15 billion parameter model in a low-bandwidth environment, demonstrating that training at scale is possible outside of traditional centralized settings.

While these models are of no practical use compared to the existing base models (lower performance, so not much reason to use), I believe this This situation will change in 2025Essence

@Exolabs' Sparta further promoted the development of this field, reducing the communication demand between GPU by more than 1,000 times. Sparta makes large model training in low bandwidth environments, without using special infrastructure.

The most impressive is their sentence:

"Sparta can be independent independent independent It can also be used in combination with synchronized low communication training algorithms (such as Diloco) to get better performance. Can be superimposed, thereby increasing efficiency.

With the advancement of model distillation and other technologies, make smaller models more efficient and more practical. The future of AI will no longer be about scale competition. , But about a better and more solution. Soon after, we will have a high -performance model that can run on the edge devices and even mobile phones.

10. Ten new encrypted AI agreements have exceeded $ 1 billion in market value-but they have not yet launched

AI16Z rushed to $ 2 billion in market value in 2024

Welcome to the real gold rush heat

It is easy for people to think that the current leaders will continue to win. Many people compare @Virtuals_io and AI16Z to the early iOS and Android of smartphones.

But the market is too huge and has not been developed, and only two players cannot occupy the dominant position. I predict that by the end of 2025, there will be at least ten new encrypted AI protocols (Token has not yet been launched), and its market value (rather than comprehensive dilution market value) will exceed $ 1 billion.

Decentralization AI is still in its infancy, but a large number of outstanding talents are entering this field.

We must fully foresee the arrival of the new protocol, the new token model and the new open source framework. These new players may use the following waysReplace the existing leader:

Incentives: such as airdrops or clever staking mechanisms;

Technological breakthroughs: such as low-latency inference or inter-chain interoperability;

User experience improvements: such as no-code tools.

Changes in public perception can be instant and dramatic. This is both part of the fascination and challenge of the field.

Market size is a double-edged sword: the cake is huge, but for skilled teams, the entry barrier is very low. This sets the stage for a Cambrian explosion of projects, many of which will be short-lived but a few that will become game-changing forces.

Bittensor, Virtuals and ai16z will not be alone for long. The next billion-dollar crypto AI protocol is coming. There's plenty of opportunity here for savvy investors, which is why this space is so exciting.

1) Easter egg 1: AI agents are the new “apps”

When Apple launched the App Store in 2008, its slogan was: "There is an app that can Do it. ”

Soon, you will say: "There is an agent who can do it."

You will no longer click on an icon to open an app, but instead delegate tasks to a dedicated AI acting. These agents are context-aware, able to interact with other agents and services, and can even initiate tasks autonomously that you didn’t explicitly ask for—like monitoring your budget or rearranging your travel plans when flights change.

Put simply, your smartphone home screen could become a network of "digital collaborators," each with their own niche - health , finance, efficiency and social.

Because these are crypto-enabled agents, they can leverage decentralized infrastructure to autonomously handle tasks such as payments, authentication, or data storage.

2) Easter Egg 2: There are also robots

Although most of this article focuses on the software field, I am also very excited about the physical expression form of these AI revolution -robots. Robotics will usher in its own" ChatGPT moment "in this century. /p>

This field still faces many major obstacles, especially in obtaining perceived real world data sets and improving physical capabilities. Crypting token to motivate data collection and innovation. I ca n’t afford to feel such a strong excitement last time. left; "> Let us welcome 2025 together!