From Bittensor revival to the rise of AI Agent, in 2025, the Top Ten Forecast of AI

Original title: What we 're watching in 2025 (Crypto AI)

Original Author: teng yan

Compilation: Asher

The future of the encryption AI sector is full of attractiveness. Although it lacks historical examples and clear trends, it also means that it is at a new starting point and waiting for future development. Thinking of all this in 2026, seeing the gap between expectations in early 2025 and the actual situation will be even more exciting.

1. The total market value of encrypted AI sector will reach 150 billion

US dollars

At present, the tokens of the encrypted AI sector account for only 2.9%of the market value of the cottage, but I believe this proportion will not last long. With the gradual expansion of artificial intelligence to new areas such as smart contract platforms, MEME, decentralized physical infrastructure (DEPIN), proxy platforms, data networks, and intelligent coordination layers. Trend.

I have confidence in the encryption AI sector because it is at the intersection of the two most powerful technical trends:

• AI fanatic trigger event: OpenAI's first public offering or similar incidents may trigger global fanaticism to AI. At the same time, Web2's institutional capital is paying attention to decentralized AI infrastructure as investment.
• Retail frenzy: The concept of artificial intelligence is easy to understand and exciting, and now they can invest through token. Remember the gold rush of MEME coins in 2024? This will be the same fanatic, but artificial intelligence changes the world more practically.

Second, the revival of Bittensor

Bittensor (token name is TAO) has been for many years. It is an old player in this field. Despite the boom of artificial intelligence, its token prices have been hovering at the level a year ago. But in fact, the digital honeycomb thinking behind Bittersor has quietly advanced. More subnets have emerged, the registration costs are reduced, and the actual performance of subnets has surpassed the corresponding technology of web2 in the actual performance of the reasoning speed. Further enriched Bittersor's network.

So why did TAO soar? The steep inflation plan and the concern of the platform -oriented platform for AI Agent have limited it. However, DTAO (expected to be in the first quarter of 2025) may be a major turning point. Through DTAO, each subnet will have its own tokens, and the relative price of these tokens will determine how to distribute and release.

Why is Bittersor a chance to explode again:

• Market release: DTAO directly linked to the performance of block rewards to innovation and actual measurement. The better the sub -network performance, the more valuable its tokens -therefore, the more release it has.
• Focus on capital flow: Investors can finally invest in specific sub -nets they are optimistic. If a subnet adopts an innovative method in distributed training and has achieved success, capital can flow into the subnet to express their investment views.
• EVM Integration: Compatible with EVM has attracted a wider range of encrypted native developer communities into Bittersor, which has brought about the gap with other networks.

At present, they are paying attention to each sub -network and recorded their actual progress in their respective fields. It is expected that at a certain time, it will usher in the DEFI summer similar to the @Opentersor version.

Third, the calculation market is the next L1 transaction

Calculating demand is not met to become an obvious giant trend. Huang Renxun, CEO of Nvidia, once said that the demand for reasoning will surge "billion times." This exponential growth will break the planning of traditional infrastructure and urgently call "new solutions."

The decentralized calculation layer provides original calculations (for training and reasoning) in a validable and cost -effective way. Start -up companies like@SPHERONFDN,@Gensynai,@Atoma_network and @Kuzco_xyz are quietly establishing a strong foundation to use this to focus on products instead of tokens (none of these companies currently). As decentralized AI model training becomes feasible, it is expected to rise sharply for the addressable market.

For encryption AI sectors, compare the L1 section:

• Just like 2021: Remember solana, Terra, and Avalanche to fight for the "best" L1? We will see similar battles between the calculation protocols to compete for developers and AI applications in order to build on their computing layers.
• Web2 demand: 680 billion US dollars to $ 2.5 trillion cloud computing market far exceeds the encryption AI market. If these decentralized calculation solutions can get even a small part of traditional cloud customers, there will be a chance to see the next 10 or 100 times growth wave.

Just as Solana stands out in the field of L1, the winners here will dominate a new cutting -edge, and need to pay close attention to three standards: reliability, cost benefits, and developers' friendship tools.

4. AI Agent will pour into the blockchain transaction

By the end of 2025, 90%of the chain transactions will no longer be manually clicked by human beings. Instead, these transactions will be performed by an AI Agent army, which will continue to balance the liquidity pool, allocate rewards, or perform micro -payment based on real -time data source.

All this is not as far -fetched as it looks like. In the past seven years, we have built everything (L1, Rollups, DEFI, NFT, etc.) in the past seven years quietly paved the way for a chain on the chain led by AI.

@Autonolas proxy on GNOSIS transaction

So why is there such a change?

• No one is wrong: Smart contracts are accurately executed according to coding. AI Agent can process a lot of data faster and accurately than a set of humans.
• Micro -payment: AI Agent -driven transactions will become smaller, more frequent and more efficient. Especially with the transaction costs of Solana, Base, and other L1/L2 declined.
• Invisible infrastructure: Human beings will be willing to give up direct control, if this means reducing trouble. Trusting Netflix automatically renews, and trusts a AI Agent to automatically re -balance the user's DEFI position is the next step.

AI Agent generates amazing amounts of activity, but the biggest challenge will be to make these AI Agent -driven systems responsible for humans. With the increase of transactions initiated by AI Agent and transactions initiated by humans, new governance mechanisms, analysis platforms and audit tools will be required.

Fifth, interaction between agents and agents: the rise of the concept of

AI cluster

The AI ​​Agent group refers to the seamless collaboration of small artificial intelligence entities to implement a grand plan, which sounds like the next popular science fiction or horror movie plot. Most of the current AI Agent operates in isolation, with less interaction and unpredictable. However, the AI ​​Agent group will change this situation, allowing multiple AI Agents to exchange information, negotiate and joint decisions in the network.

These AI agent groups can be regarded as decentralized professional model collectives, and each model contributes its unique professional knowledge for larger and more complicated tasks. Its potential is shocking. For example, a group may coordinate distributed computing resources on platforms like Bittensor, and another group can verify the source of content in real time to prevent false information from spreading on social media. Each AI Agent in the group is an expert and accurately performs their respective tasks.

The intelligence of these group networks will far exceed any single isolation AI. In order to make the agent group vigorously develop, general communication standards are crucial. The agent needs to be found, certified and collaborated without restrictions on the underlying framework. Teams such as Story, FXN, ZEREBRO, and AI16Z are committed to laying the foundation for the rise of the agent group.

At the same time, this also leads to a key role in decentralization, allocating tasks to the agent groups managed by the rules on the transparent chain, so that the system has higher flexibility and adaptability. If an agent fails, other agents can intervene in the replacement and maintain the continuous operation of the system.

6. The encrypted AI work team will be a mixture of human and artificial

intelligence

Story hired Luna (an AI Agent project) as their social media intern and paid her $ 1,000 per day. It may sound strange, but this is a precursor of the future. In this future, AI Agent will become real collaborators, with its own autonomy, responsibility, and even salary. In various industries, the company is conducting the test of the human -machine hybrid team.

We will cooperate with AI Agent, not as our slaves, but as equal partners:

• Property surge: AI Agent can process a lot of data, communicate with each other, and make decisions within 24/7 without sleep or coffee rest.
• Establishing trust through smart contracts: Blockchain is a supervisor who is not favored, not tired, and never forgets. A chain account to ensure that important AI agent behaviors follow specific boundary conditions/rules.
• Social norms are evolved: soon facing the etiquette of interaction with smart body -should you say "please" and "thank you" to AI? Should they assume moral responsibility for their mistakes or blame their developers?

The boundaries between "employees" and "software" began to blur in 2025. I believe that more encrypted teams will participate, because AI Agent performed well in terms of content generation, can be broadcast live broadcast around the clock and publish content on social media. If an AI protocol is being developed, why not show its ability through internal deployment AI Agent?

7. 99%of AI Agent will die (only useful can survive)

We will see the Darwin -style elimination between AI Agent. This is because running an AI agent requires consumption power, that is, the cost of reasoning. If an AI agent cannot create enough value to cover its "rent", it will face elimination.

Take the AI ​​Agent Survival Game as an example. The first is carbon credit AI: Assuming that there is a AI Agent looking for inefficient links in decentralized energy networks and trading to token carbon credit independently. If it can earn enough income to pay its calculation fee, this AI Agent will flourish. Another example is the Dex arbitrage robot: this AI Agent earns a stable income by using the price difference between the decentralized exchange, which is enough to cover its inference costs. In contrast, the spoof on the X: a virtual AI net red with an interesting but no sources of income, with the fading of freshness and the decline of the tokens, it will gradually disappear and cannot maintain their livelihood.

The difference is obvious that the practical -oriented AIAGENT will be booming, and those AIAGENT, which dependence on interference and gimmick, will become irrelevant. Such a natural choice is beneficial to the industry, and it prompts developers to continue to innovate and give priority to productive applications rather than fancy technology. With the rise of more powerful and productive forces, the suspect will gradually be silent.

8. AI synthetic data beyond human data

The saying "data is new oil" is widely circulated. However, the high dependence of artificial intelligence on data has also triggered concerns about the shortage of data shortage. Traditional views believe that it should be found to collect private practical data from users, and even pay them for it.

However, in the case of highly regulated industries or real data scarcity, a more actual solution may be synthetic data. Synthetic data is artificially generated and aims to simulate the data distribution of the real world. It provides a scalable, ethical, friendly, and privacy and security solution for replacing human data. The advantage of synthetic data is:

• Unlimited scale: It requires a million medical X -rays or factories to scan 3D. Synthetic data can be generated infinitely without relying on real patients or factories.
• Privacy and Friendship: When processing synthetic data, personal privacy information is not threatened.
• Customization: You can adjust the data distribution according to specific training needs, and even insert those marginal cases that are scarce or ethical in reality.

Although the data owned by humans is still important in many cases, if synthetic data is continuously improved in terms of authenticity, it may surpass user data in terms of quantity, generating speed, and advantages of non -privacy restrictions. In the future, decentralized artificial intelligence may be carried out around the "mini laboratory". These laboratories focus on creating a highly professional synthetic dataset to meet specific use cases.

Nine, decentralized training began to play a role

In 2024, pioneers like Prime Intellect and Nous Research pushed the boundary of decentralized training. For example, successfully trained a 15 billion parameter model in a low bandwidth environment, proves that large -scale training can be implemented in addition to traditional centralized settings. Although these models do not perform well in actual applications compared to existing basic models and have low performance, there are not many reasons to use them, it is expected that this situation will change in 2025.

EXO Labs has further promoted progress through Sparta, reducing communication between GPUs by more than 1000 times. Sparta is possible to train large model training under low bandwidth without relying on dedicated infrastructure. The most impressive thing is their statement: "Sparta works independently, but it can also be combined with synchronized low communication training algorithms such as Diloco to obtain better performance." This means that these improvements are superimposed, and efficiency improves efficiency. It is gradually accumulated.

With the continuous progress of model technology, smaller and more efficient models will become more useful. The future of artificial intelligence no longer pays attention to the scale, but pays more attention to quality and acquisition. Soon, there will be high -performance models that can run on the edge devices and even mobile phones.

10. At least ten new encryption AI super protocol

Although many people compare Virtuals and AI16Z with the early stages of smartphones (such as iOS and Android), they believe that the current leaders will continue to win, but the market is huge and has not yet developed. Only two participants cannot dominate. By the end of 2025, at least ten new encrypted AI agreements (not yet issued) market value will exceed $ 1 billion.

Decentralized artificial intelligence is still in its infancy and a large number of talents are gathering. The new agreement, the new token model and the new open source framework will continue to emerge. These new participants may be through incentives (such as airdrops or clever pledge), technical breakthroughs (such as low delay inference or inter -chain interoperability), and user experience Improve (such as non -code) to replace existing participants. The changes in public cognition may be instant and dramatic.

Bittensor, Virtuals and AI16Z will not be alone for too long. The next one billion -dollar encrypted AI agreement is coming. Smart investors will face many opportunities. This is why this market is so exciting.