Crypto AI’s top ten predictions for 2025: Total market value reaches $150 billion, 99% of AI Agents will die

Original title: "What I'm Watching in 2025"

Author:Teng Yan , researcher (focusing on Crypto x AI)

Compiled by: Felix, PANews

With the explosion of the AI ​​industry this year, Crypto x AI has risen rapidly. Teng Yan, a researcher who focuses on Crypto x AI, published an article and made 10 predictions for 2025. Below are the forecast details.

1. The total market cap of crypto AI tokens reaches $

150 billion

The current market capitalization of crypto AI tokens only accounts for 2.9% of the market capitalization of altcoins, but this ratio will not last long.

AI covers everything from smart contract platforms to memes, DePIN and Agent platforms, data networks and intelligent coordination layers. There is no doubt that its market position is comparable to DeFi and memes.

Why are you so confident about this?

• Crypto AI is at the fusion of two of the most powerful technologies
• AI mania triggering event: The OpenAI IPO or similar events may trigger a global mania for AI. At the same time, Web2 Capital has begun to focus on decentralized AI infrastructure
• Retail Frenzy: The AI ​​concept is easy to understand and exciting, and retail investors can now invest in it via tokens. Remember the meme gold rush of 2024? AI will be the same craze, except AI is actually changing the world.

2. Bittensor renaissance

The decentralized AI infrastructure Bittensor (TAO) has been online for many years and is a veteran project in the field of encrypted AI. Despite AI's popularity, its token prices have been hovering around the levels of a year ago.

Now Bittensor's Digital Hivemind has quietly made a leap forward: more subnets have lower registration fees, subnets perform better than their Web2 counterparts in actual indicators such as inference speed, and EVM compatibility will be similar to DeFi's Function introduced into Bittensor’s network.

Why isn’t TAO token soaring? Sharp inflationary plans and the market's focus on the Agent platform have hindered its rise. However, dTAO (expected to be launched in the first quarter of 2025) may be a major turning point. With dTAO, each subnet will have its own token, and the relative prices of these tokens will determine how emissions are allocated.

Why Bittensor can make a comeback:

• Market-Based Emissions: dTAO ties block rewards directly to innovation and real measurable performance. The better the subnet, the more valuable its tokens are.
• Concentrate capital flows: Investors can ultimately target specific subnets in which they believe. If a particular subnet wins with an innovative distributed training approach, investors can deploy capital to represent their perspective.
• EVM Integration: Compatibility with EVM attracts the wider community of crypto-native developers within Bittensor, bridging the gap with other networks.

3. The computing market is the next “L1 market ”

The obvious megatrend right now is the insatiable demand for computing.

NVIDIA CEO Jensen Huang has said that demand for inference will grow "a billion times." This exponential growth will disrupt traditional infrastructure plans and new solutions are in short supply.

The decentralized computing layer provides raw computation (for training and inference) in a verifiable and cost-effective manner. Startups like Spheron, Gensyn, Atoma, and Kuzco are quietly building a solid foundation, focusing on products rather than tokens (none of these companies have tokens). As decentralized training of AI models becomes practical, the entire addressable market will rise dramatically.

Compare with L1:

• It's like 2021: Remember Solana, Terra/Luna, and Avalanche competing for the "best" L1? A similar competition will emerge among compute protocols for developers and AI applications built using their compute layers.
• Web2 demand: The cloud computing market size of US$680 billion to US$2.5 trillion dwarfs the encrypted AI market. If these decentralized computing solutions can attract even a small percentage of traditional cloud customers, we could see the next wave of 10x or 100x growth.

Just as Solana triumphed in L1, the winner will dominate a whole new realm. Pay close attention to reliability (such as a strong service level agreement or SLA), cost-effectiveness, and developer-friendly tools.

4.AI agents will flood blockchain transactions

 _Olas agent transaction on Gnosis; source: Dune_ 

By the end of 2025, 90% of on-chain transactions will no longer be clicked “send” by real humans, but will be executed by a swarm of AI agents who are constantly rebalancing liquidity pools, distributing rewards, or executing micropayments based on real-time data feedback.

It doesn't sound far-fetched. Everything built in the past seven years (L1, rollup, DeFi, NFT) has quietly paved the way for a world where AI runs on the chain.

Ironically, many builders may not even realize they are creating the infrastructure for a machine-dominated future.

Why did this shift occur?

• No more human errors: smart contracts execute exactly as coded. In turn, AI agents can process large amounts of data faster and more accurately than real humans.
• Micropayments: These agent-driven transactions will become smaller, more frequent, and more efficient. Especially as transaction costs on Solana, Base and other L1/L2 are trending downward.
• Invisible infrastructure: Humans will gladly relinquish direct control if it reduces some hassle.

AI agents will generate a lot of on-chain activities, so it’s no wonder that all L1/L2 are embracing agents.

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

5. Interaction between agents: the rise of clusters

 _Source: FXN World_ 

The concept of Agent swarms—tiny AI agents working seamlessly together to execute grand plans—sounds like the plot of the next big sci-fi/horror movie.

Today’s AI agents are mostly “lone wolves,” operating in isolation with minimal and unpredictable interactions.

Agent clusters will change this status quo, allowing a network of AI agents to exchange information, negotiate and make collaborative decisions. Think of it as a decentralized collection of specialized models, each contributing unique expertise to larger, more complex tasks.

A cluster might coordinate distributed computing resources on a platform such as Bittensor. Another cluster can handle misinformation, verifying the source of content in real time before it spreads to social media. Each Agent in the cluster is an expert and can perform its tasks with precision.

These swarm networks will produce intelligence more powerful than any single isolated AI.

For clusters to thrive, common communication standards are essential. Regardless of its underlying framework, agents need to be able to discover, authenticate, and collaborate. Teams such as Story Protocol, FXN, Zerebro and ai16z/ELIZA are laying the foundation for the emergence of Agent clusters.

This reflects the key role of decentralization. Under transparent on-chain rule management, tasks are assigned to various clusters, making the system more flexible and adaptable. If one Agent fails, other Agents will step in.

6. Crypto AI **workforce will be a hybrid of humans and

machines**

 Source: @whip_queen_ 

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

As strange as it sounds, this is a harbinger of a future in which AI agents become true collaborators, with autonomy, responsibility, and even paychecks. Companies across industries are beta testing hybrid human-machine teams.

In the future, we will work with AI Agents, not as slaves, but as equals:

• Productivity surge: Agents can process large amounts of data, communicate with each other, and make decisions around the clock without sleep or coffee breaks.
• Building trust through smart contracts: The blockchain is an unbiased, untiring, and never-forgetting overseer. An on-chain ledger ensures that important Agent operations follow specific boundary conditions/rules.
• Social norms are constantly evolving: Soon you will start thinking about the etiquette of interacting with agents - will you say "please" and "thank you" to an AI? Will they be held morally responsible for mistakes, or will they blame their developers?

The line between “employees” and “software” will begin to disappear in 2025.

7. 99% of AI Agents will die – **only the useful

ones will survive**

The future will see a “Darwinian” elimination among AI agents. Because running AI agents requires expenditure in the form of computing power (ie, inference cost). If the Agent cannot generate enough value to pay its "rent," the game is over.

Agent survival game example:

• Carbon Credit AI: Imagine an agent searching a decentralized energy network, identifying inefficiencies, and autonomously trading tokenized carbon credits. It thrives only if it makes enough money to pay for its own computing expenses.
• DEX arbitrage robot: Agents that exploit price differences between decentralized exchanges can generate stable income to pay for their reasoning fees.
• Shitposter on X: Virtual AI KOL with cute jokes but no sustainable revenue stream? Once the novelty wears off (token prices plummet), you can't pay for yourself.

Utility-driven agents thrive, while distraction-driven agents fade into irrelevance.

This elimination mechanism is beneficial to the industry. Developers are forced to innovate and prioritize production use cases over gimmicks. With the emergence of these more powerful and efficient agents, skeptics can be silenced.

8. Synthetic data surpasses human data

“Data is the new oil.” AI thrives on data, but its appetite raises concerns about looming data depletion.

Conventional wisdom is to find ways to collect users’ private real-life data and even pay for it. But a more practical route, especially in highly regulated industries or where real data is scarce, is to use synthetic data.

Synthetic data are artificially generated data sets designed to mimic real-world data distributions. Providing a scalable, ethical, and privacy-friendly alternative to human data.

Why synthetic data works so well:

• Infinite scale: Need a million medical X-rays or 3D scans of a factory? Synthetic generation can be made in unlimited quantities without waiting for real patients or real factories.
• Privacy Friendly: No personal information is at risk when using artificially generated datasets.
• Customizable: Distributions can be customized to your exact training needs.

User-owned human data will still be important in many cases, but if synthetic data continues to improve in reality, it may surpass user data in volume, speed of generation, and lack of privacy constraints.

The next wave of decentralized AI may center around “microlabs” that create highly specialized synthetic data sets tailored to specific use cases.

These micro-labs will neatly bypass policy and regulatory hurdles in data generation—much like Grass bypassed web scraping restrictions by leveraging millions of distributed nodes.

9. Decentralized training is more useful

In 2024, pioneers like Prime Intellect and Nous Research are pushing the boundaries of decentralized training. A 15 billion parameter model was trained in a low-bandwidth environment, demonstrating that large-scale training is possible outside of traditional centralized settings.

While these models are of no practical use (lower performance) compared to the existing base models, this will change in 2025.

This week, EXO Labs made further progress with SPARTA, reducing inter-GPU communication by more than 1,000 times. SPARTA enables large model training on slow bandwidth without requiring specialized infrastructure.

What impressed me was the statement: “SPARTA can run independently, but can also be combined with synchronization-based low-communication training algorithms such as DiLoCo for better performance.”

This means these improvements can be stacked, resulting in increased efficiency.

As technology advances and tiny models become more practical and efficient, the future of AI lies not in scale but in getting better and easier to use. It is expected to soon have high-performance models that can run on edge devices and even mobile phones.

10. Ten new crypto AI **protocols have a circulating market

value of** $ 1 billion (not yet launched)

 ai16z will achieve a market value of US$2 billion in 2024 

Welcome to the real gold rush.

It's easy to think that the current leaders will continue to win, with many comparing Virtuals and ai16z to early smartphones (iOS and Android).

But the market is too large and untapped for just two players to dominate. By the end of 2025, at least ten new crypto AI protocols (that have not yet launched tokens) are expected to have a circulating (not fully diluted) market capitalization of over $1 billion.

Decentralized AI is still in its infancy. And the talent pool is growing.

Expect new protocols, novel token models, and new open source frameworks to come. These new players can displace existing players through a combination of incentives (like airdrops or clever staking), technological breakthroughs (like low-latency inference or chain interoperability), and user experience improvements (no code). Changes in public perception can be instantaneous and dramatic.

This is both the beauty and challenge of this field. Market size is a double-edged sword: the pie is huge, but for technical teams, the barriers to entry are low. This sets the stage for an explosion of projects, many of which will fade away but a few that will have transformative power.

Bittensor, Virtuals, and ai16z won’t stay at the forefront for much longer, the next $1 billion crypto AI protocol is just around the corner. There are tons of opportunities for savvy investors, which is why it's so exciting.