In this article, we aim to take you through some of Solana's recent major moves and, focusing on the AI sector, delve into the current landscape and future prospects of Solana AI.

"Solana's greatest strength lies in its versatility, capable of hosting multiple different types of applications and sectors simultaneously."

— Lily Liu, President of the Solana Foundation

Many relationships in the world are like a seesaw. While ETH's price has recovered on its 10th anniversary, SOL has been going through a period of anxiety — the path of being an "Ethereum killer" has become even more difficult, and it has also encountered a group of its own "killers" — a bit of a role reversal, as fortunes shift.

However, moderate anxiety is not a bad thing for either individuals or projects. For Solana, which has weathered many storms, coping with such challenges is more familiar territory. In this article, we aim to take you through some of Solana's recent major moves and, focusing on the AI sector, delve into the current landscape and future prospects of Solana AI.

01. Frequent Major Moves by the Official Team

1. Alpenglow Upgrade: Replacing PoH to Compete with SUI and Other Rivals at Faster Speeds

Alpenglow is hailed as the most significant upgrade to Solana's core protocol to date, comparable in significance to Ethereum's shift from PoW to PoS consensus, as Alpenglow's mission is to replace the existing PoH (Proof of History) and Tower BFT (Tower Byzantine Fault Tolerance) consensus mechanisms.

Why replace the existing consensus mechanism?

In the past, because PoH did not require synchronizing timestamps like other blockchains, and Tower BFT adopted a "one person packages, others vote" approach, it significantly simplified block synchronization time, allowing Solana to lead all public chains in performance in its early days. However, under high load, PoH caused huge computational overhead, combined with the single leader overload issue in Tower BFT, leading to the frequent downtime problems that Solana has often been criticized for.

In addition, Solana has often been criticized for the high cost of running nodes, resulting in insufficient decentralization.

How does Alpenglow improve this?

Simply put, Alpenglow removes PoH, a potential computational power killer, and uses Votor (stake-weighted voting) combined with node clocks to handle timing and confirmation, reducing the computational burden on the leader node; and addressing the drawbacks of a single leader, the design paves the way for multiple leaders to propose blocks simultaneously in the future.

Another core component, Rotor, optimizes block propagation and synchronization, reducing block confirmation time from 12.8s to 150ms. At the same time, because Rotor reduces communication overhead and computational burden between nodes, even weaker nodes can participate efficiently, achieving "performance upgrade without the need for hardware upgrade."

According to Cogent Crypto's validator yield calculator, after Alpenglow is implemented, the minimum staking amount required for validators to be profitable will drop from 4,850 SOL (approximately \(800,000) to 450 SOL (approximately \)75,000).

Final Expectations

The Alpenglow whitepaper states: "Alpenglow will be a turning point for Solana; it is not just a new consensus protocol, but a key step for Solana to move towards internet infrastructure-level competitiveness."

In simple terms: faster, more stable, cheaper, more scalable, more decentralized, reclaiming the throne of performance king, and escaping the nightmare of downtime.

1. ICM Roadmap: Re-optimizing Transactions to Continue Marching Towards "On-Chain Nasdaq"

After announcing the Alpenglow upgrade, Solana Labs, together with several core development teams in the ecosystem such as Anza and Jito, released a more specific "Internet Capital Markets" (ICM) roadmap.

In the past few years, Solana has attracted a large number of users and DeFi market share with low costs and fast transactions. However, the trading-specific chain Hyperliquid has risen rapidly, capturing more than 70% of the on-chain contract market. The founder of Hyperliquid even publicly criticized Solana's speed as "not fast enough." This undoubtedly made Solana feel a sense of crisis.

Is Solana not fast enough?

Currently, transaction final confirmation on Solana takes 12-13 seconds, while Hyperliquid's average confirmation time is about 0.2 seconds, and SUI's average confirmation time is 0.5 seconds.

It is expected that after the Alpenglow upgrade, Solana's confirmation time will reach 150 milliseconds, regaining its speed advantage and achieving Visa-level transaction confirmation capabilities, but there is still a gap compared to Nasdaq's "microsecond-level" high-frequency trading system.

Beyond being faster, what else needs to be done?

Giving more power to dApps on the platform — the so-called ACE (Application-Controlled Execution) — allowing dApps (smart contracts) to determine transaction priorities themselves, enabling dApps to handle complex issues more flexibly and effectively.

Protecting market makers and addressing MEV (front-running bots and other high-frequency arbitrage attacks): Hyperliquid's order-matching engine gives higher priority to market maker orders, protecting them from MEV attacks, thereby offering better prices to attract retail investors. Solana ICM's roadmap is similar; through the introduction of BAM and the completion of the Alpenglow upgrade, DEXs in the ecosystem will be able to address the cancer of high-frequency arbitrage, improve market health, and bring better prices to retail investors.

A More Ambitious Goal

In addition to aligning with Hyperliquid in DeFi, Solana has a more ambitious goal — a true "on-chain Nasdaq" — helping more enterprises bypass complex IPO processes and conduct on-chain financing.

Solana co-founder Anatoly Yakovenko recently stated in an interview that he hopes to complete the on-chain listing of traditional financial assets (RWA) within a year, provide compliant, open-source on-chain IPOs for entrepreneurs within five years, and ultimately build an open, low-cost, decentralized internet capital market.

It is ambitious, but it seems to be the goal of almost all top public chains. Currently, we see Solana making continuous efforts to reform in terms of underlying architecture, ultimate performance, and application scenarios. However, whether this strategy can break through in this highly competitive era still requires time and market testing.

02. In Troubled Times, Where Will the AI Narrative Go?

Public chains usually do not easily change their consensus strategies, just as no one would casually change their nationality or move their household registration. Once changed, it is often driven by huge threats.

In terms of core on-chain transaction business, whether it is Hyperliquid, as a trading-specific chain, rapidly eroding Solana's advantages in the DEX perpetual contract market, or the rising star Sui, as a general-purpose chain, comprehensively catching up in performance in multiple fields such as DeFi and DeAI, coupled with hot topics such as stablecoins, RWA, and MicroStrategy pushing ETH to the throne again, all have put Solana under significant pressure, forcing it to accelerate upgrades to reverse the situation.

The AI field, which Solana is proud of, is also not optimistic. After the once popular DePIN and AI Agent MEME boom subsided, the Virtuals ecosystem rose rapidly, making Base occupy an absolute advantage in the AI Agent field; BNB Chain, leveraging celebrity effects and exchange resources, diverted much of the attention from AI MEMEs. In terms of decentralized AI infrastructure, the rapid expansion of Bittensor subnets and the rise of many AI Layer 1s are gradually forming new forces in the DeAI field.

For a while, anxiety seemed to have spread to all aspects of the Solana chain, but Solana, which has weathered many storms, is clearly not going to admit defeat easily. Currently, the Alpenglow upgrade is expected to bring strong confidence to the ecosystem. At the very least, Solana officials have shown sufficient humility and pragmatism, clearly benchmarking competitors, gradually making up for shortcomings, and even striving to be stronger.

Next, let's sort out Solana's AI ecosystem to see if Solana's AI narrative can help it usher in a new round of prosperity.

03. Solana's AI Landscape

The AI projects on Solana have been developing for a long time and span a wide range of sectors. Taking stock of them is impressive. Here, we can divide the AI projects on Solana into three stages according to approximate time intervals.

First Stage: Early DePIN Ecosystem Boom, DeAI in Its Infancy

The popularity of the DePIN concept made Solana one of the earliest public chains to explore on-chain decentralized AI. These projects, relying on Solana's strong performance and low prices, built their own decentralized computing power, bandwidth, data, and other networks, and also laid the early foundation for the Solana AI ecosystem.

In terms of AI computing power networks, early projects such as Render, io.net, and Aethir were among the first to attempt on-chain decentralized computing power, but with different focuses:

Render focuses on 3D animation and metaverse rendering tasks, recording the allocation and payment of rendering tasks on Solana, and matching GPU providers and demanders through smart contracts.

io.net mainly provides low-cost decentralized GPU computing power for AI and machine learning, automatically allocating computing power through its own network, with Solana used for settling \(IO tokens and recording node contributions. Aethir mainly optimizes real-time applications such as games through edge computing, with its own network allocating computing power to edge nodes, and Solana used for recording transactions and \)ATH token incentives.

These projects are among the early explorers in the already crowded decentralized computing power network. As early entrants, they have raised large amounts of financing and are highly favored by the market. However, these projects usually require high-performance GPUs to participate in the network, making the threshold too high for ordinary people.

In contrast, Grass, Helium, Roam, and Gradient Network have a better mass foundation.

Grass is a decentralized data collection network. Users can automatically contribute bandwidth by downloading a browser plugin or application, and collect network data through activities such as crawling. Users' contributions are recorded on the Solana chain, and they receive \(GRASS rewards. Roam's network aims to provide seamless, secure global internet coverage through community-provided WiFi nodes. Users can earn \)ROAM tokens by providing WiFi coverage or verifying the network in the APP, with the project using Solana for node management and payments.

Gradient Network is a decentralized edge computing platform. Users only need to install a plugin to contribute idle computing resources (such as mobile phones, laptops, etc.) to support scenarios such as AI inference, content distribution, and serverless functions, and handle coordination and payments between devices through Solana's on-chain mechanism to ensure efficient and secure decentralized computing.

Helium provides IoT and mobile network coverage in a decentralized manner, using Solana to record hotspot data and transactions. Users purchase and deploy Helium Hotspots (hotspot devices, priced at approximately \(200-500) to provide network coverage and receive \)HNT or \(MOBILE token rewards. A more accessible way to participate is to use the SIM card launched by Helium and T-Mobile in the U.S., enable the "Mapping" function, share anonymous location data to help optimize network coverage, earn Cloud Points or \)MOBILE tokens, and get a free trial of the SIM card.

These projects have actively explored the field of infrastructure combining blockchain and AI (such as bandwidth, data, computing power, etc.), not only bringing huge imagination and confidence to the market but also further opening a new chapter in DeAI. Although the token prices of these projects have fallen from their highs, network effects are still evident, and many projects have established cooperation with traditional large companies, winning recognition from the mainstream market.

Second Stage: AI Agents Flourish in Multiple Areas

With the significant improvement in the capabilities of LLMs represented by ChatGPT, applications such as AI Agents have developed rapidly in the past year. Solana, with its excellent performance and rich ecosystem, became one of the earliest public chains to benefit from on-chain AI Agent tokens and the prosperity of related applications.

Wayfinder, developed by Parallel Studios (invested by Solana Ventures, previously mainly known for the Solana-based blockchain game Colony), simplifies cross-chain operations through AI Agents, supporting easy cross-chain asset transfers between Solana, Ethereum, Polygon, and Base.

ElizaOS is an open-source AI Agent framework that helps developers quickly create, deploy, and manage AI Agents on Solana. These AI Agents can quickly process information, execute transactions, and interact with smart contracts, with governance through DAO AI16Z. The DAO token \(AI16Z once had a market capitalization exceeding \)2.5 billion, becoming the leading AI token on Solana. However, its MEME attribute was too strong, and its value later returned to rationality.

Holoworld is a multi-modal AI Agent launch platform that allows ordinary people to create, customize, and trade AI Agents, and supports 3D character and video customization. All digital asset interactions are managed through Solana's smart contracts.

Moby AI initially was an AI Agent focusing on Alpha research and providing real-time crypto information, and has now launched its own terminal, focusing on Alpha mining and trading on the Solana chain.

Hey Anon is a DeFAI protocol deployed on Solana, aiming to simplify DeFi interactions through natural language processing and real-time data aggregation, helping users manage DeFi operations, obtain project updates, and analyze market trends.

In addition, during the AI boom at the beginning of the year, many well-known Agent and application projects emerged on the Solana chain, such as the Agent frameworks \(ARC and \)SWARMS, the DeFAI platforms \(GRIFFAIN and \)BUZZ, the AI Agent launcher \(HAT, and the well-known AI Agents \)PIPPIN and $ZEREBRO, etc. However, after the boom subsided, the token prices and popularity of these projects dropped rapidly, and some projects were put on hold.

Third Stage: Post-On-Chain AI Era, Steadily Moving Towards On-Chain DeAI

As the market gradually returns to rationality, we have noticed that a new batch of projects has emerged on the Solana chain. They have shed the hustle and bustle of the AI MEME era and are building decentralized AI infrastructure on Solana in a more down-to-earth and feasible manner.

Nous Research is committed to training open-source AI models that can compete with mainstream large models such as OpenAI through decentralized means. Its core Psyche network significantly reduces the communication frequency between nodes through compression technology, effectively solving the communication bandwidth bottleneck of decentralized AI training, enabling the network to carry high-computation-demand AI training scenarios such as LLM pre-training; and using the Solana network to record node contributions and distribute incentives. It has currently launched the Hermes series of open-source models fine-tuned based on the Llama model.

Arcium was initially a privacy protocol Elusiv on Solana, later transforming into a broader privacy computing platform. Using MPC and ZKP technologies, it allows computing on encrypted data without exposing the data content, not only providing protection for user privacy in fields such as DeFi and Desci but also offering privacy infrastructure for AI model training, inference that requires attention to training data privacy, and AI Agent applications.

Neutral Trade is a hedge fund platform based on Solana, providing a variety of AI-driven quantitative and automated trading strategies. Currently, in addition to common quantitative neutral strategies and arbitrage strategies, it also has a well-performing CTA Momentum strategy, a multi-asset, high-liquidity strategy operated in cooperation with the well-known quantitative company R* Research, with an annualized return of 95.11%, and the $10 million subscription quota has been fully filled.

Although the popularity of Solana AI has declined in this stage, it is somewhat like sifting through the chaff. The quality of projects has significantly improved — with stronger implementation capabilities and solving more practical problems — evolving from "dream speakers" to "doers."

04. Views on Solana's Current AI Ecosystem

Through the above overview, we can see that the AI ecosystem on Solana has developed early, covers a very comprehensive range, and its overall strength is still very competitive among existing public chains.

While being comprehensive, many excellent projects have emerged in the Solana ecosystem. For example, Nous Research is committed to solving the problem of decentralized large-scale training of AI models, Grass collects massive training data through a decentralized network, and Arcium focuses on building privacy infrastructure. The competitors of these projects often require proprietary chains to support them, but they have developed well on Solana, fully demonstrating Solana's excellent carrying capacity.

Although the current popularity of on-chain AI has been diverted by Base, BNB Chain, etc., in the long run, the possibility of one chain dominating the AI field is almost zero, and there will be more cooperation than competition between chains. At the same time, with the Alpenglow upgrade, Solana's advantages in the AI field will be further strengthened:

1. Fast and Cost-Effective: Speed, throughput, and fees are the most concerned factors for on-chain AI projects. Multiple AI Agents under the management of protocols such as MCP need close cooperation and transactions, and decentralized training, data collection, and other tasks also rely on high-frequency, fast, and massive interactions between nodes. Currently, Solana is already very efficient; after the Alpenglow upgrade, the confirmation time will be shortened to 150 milliseconds, further reducing latency and costs, and better supporting real-time AI applications.

1. Good Liquidity: As the currency of projects, the price stability and smooth circulation of AI tokens have a significant impact on projects. Solana currently has a daily DEX volume of up to $1.4B, ranking second in the entire network only to Ethereum, and has a mature and active DeFi ecosystem such as Raydium and Jito. AI projects can easily access stable liquidity pools. After the upgrade, it is expected to attract more market makers, provide deep liquidity, and help AI tokens circulate and raise funds quickly.

1. Smart Contract Support: Solana supports parallel processing and more flexible programming languages. Its virtual machine SVM can process complex logic

In this article, we aim to take you through some of Solana's recent major moves and, focusing on the AI sector, delve into the current landscape and future prospects of Solana AI.

"Solana's greatest strength lies in its versatility, capable of hosting multiple different types of applications and sectors simultaneously."

— Lily Liu, President of the Solana Foundation

Many relationships in the world are like a seesaw. While ETH's price has recovered on its 10th anniversary, SOL has been going through a period of anxiety — the path of being an "Ethereum killer" has become even more difficult, and it has also encountered a group of its own "killers" — a bit of a role reversal, as fortunes shift.

However, moderate anxiety is not a bad thing for either individuals or projects. For Solana, which has weathered many storms, coping with such challenges is more familiar territory. In this article, we aim to take you through some of Solana's recent major moves and, focusing on the AI sector, delve into the current landscape and future prospects of Solana AI.

01. Frequent Major Moves by the Official Team

1. Alpenglow Upgrade: Replacing PoH to Compete with SUI and Other Rivals at Faster Speeds

Alpenglow is hailed as the most significant upgrade to Solana's core protocol to date, comparable in significance to Ethereum's shift from PoW to PoS consensus, as Alpenglow's mission is to replace the existing PoH (Proof of History) and Tower BFT (Tower Byzantine Fault Tolerance) consensus mechanisms.

Why replace the existing consensus mechanism?

In the past, because PoH did not require synchronizing timestamps like other blockchains, and Tower BFT adopted a "one person packages, others vote" approach, it significantly simplified block synchronization time, allowing Solana to lead all public chains in performance in its early days. However, under high load, PoH caused huge computational overhead, combined with the single leader overload issue in Tower BFT, leading to the frequent downtime problems that Solana has often been criticized for.

In addition, Solana has often been criticized for the high cost of running nodes, resulting in insufficient decentralization.

How does Alpenglow improve this?

Simply put, Alpenglow removes PoH, a potential computational power killer, and uses Votor (stake-weighted voting) combined with node clocks to handle timing and confirmation, reducing the computational burden on the leader node; and addressing the drawbacks of a single leader, the design paves the way for multiple leaders to propose blocks simultaneously in the future.

Another core component, Rotor, optimizes block propagation and synchronization, reducing block confirmation time from 12.8s to 150ms. At the same time, because Rotor reduces communication overhead and computational burden between nodes, even weaker nodes can participate efficiently, achieving "performance upgrade without the need for hardware upgrade."

According to Cogent Crypto's validator yield calculator, after Alpenglow is implemented, the minimum staking amount required for validators to be profitable will drop from 4,850 SOL (approximately \(800,000) to 450 SOL (approximately \)75,000).

Final Expectations

The Alpenglow whitepaper states: "Alpenglow will be a turning point for Solana; it is not just a new consensus protocol, but a key step for Solana to move towards internet infrastructure-level competitiveness."

In simple terms: faster, more stable, cheaper, more scalable, more decentralized, reclaiming the throne of performance king, and escaping the nightmare of downtime.

1. ICM Roadmap: Re-optimizing Transactions to Continue Marching Towards "On-Chain Nasdaq"

After announcing the Alpenglow upgrade, Solana Labs, together with several core development teams in the ecosystem such as Anza and Jito, released a more specific "Internet Capital Markets" (ICM) roadmap.

In the past few years, Solana has attracted a large number of users and DeFi market share with low costs and fast transactions. However, the trading-specific chain Hyperliquid has risen rapidly, capturing more than 70% of the on-chain contract market. The founder of Hyperliquid even publicly criticized Solana's speed as "not fast enough." This undoubtedly made Solana feel a sense of crisis.

Is Solana not fast enough?

Currently, transaction final confirmation on Solana takes 12-13 seconds, while Hyperliquid's average confirmation time is about 0.2 seconds, and SUI's average confirmation time is 0.5 seconds.

It is expected that after the Alpenglow upgrade, Solana's confirmation time will reach 150 milliseconds, regaining its speed advantage and achieving Visa-level transaction confirmation capabilities, but there is still a gap compared to Nasdaq's "microsecond-level" high-frequency trading system.

Beyond being faster, what else needs to be done?

Giving more power to dApps on the platform — the so-called ACE (Application-Controlled Execution) — allowing dApps (smart contracts) to determine transaction priorities themselves, enabling dApps to handle complex issues more flexibly and effectively.

Protecting market makers and addressing MEV (front-running bots and other high-frequency arbitrage attacks): Hyperliquid's order-matching engine gives higher priority to market maker orders, protecting them from MEV attacks, thereby offering better prices to attract retail investors. Solana ICM's roadmap is similar; through the introduction of BAM and the completion of the Alpenglow upgrade, DEXs in the ecosystem will be able to address the cancer of high-frequency arbitrage, improve market health, and bring better prices to retail investors.

A More Ambitious Goal

In addition to aligning with Hyperliquid in DeFi, Solana has a more ambitious goal — a true "on-chain Nasdaq" — helping more enterprises bypass complex IPO processes and conduct on-chain financing.

Solana co-founder Anatoly Yakovenko recently stated in an interview that he hopes to complete the on-chain listing of traditional financial assets (RWA) within a year, provide compliant, open-source on-chain IPOs for entrepreneurs within five years, and ultimately build an open, low-cost, decentralized internet capital market.

It is ambitious, but it seems to be the goal of almost all top public chains. Currently, we see Solana making continuous efforts to reform in terms of underlying architecture, ultimate performance, and application scenarios. However, whether this strategy can break through in this highly competitive era still requires time and market testing.

02. In Troubled Times, Where Will the AI Narrative Go?

Public chains usually do not easily change their consensus strategies, just as no one would casually change their nationality or move their household registration. Once changed, it is often driven by huge threats.

In terms of core on-chain transaction business, whether it is Hyperliquid, as a trading-specific chain, rapidly eroding Solana's advantages in the DEX perpetual contract market, or the rising star Sui, as a general-purpose chain, comprehensively catching up in performance in multiple fields such as DeFi and DeAI, coupled with hot topics such as stablecoins, RWA, and MicroStrategy pushing ETH to the throne again, all have put Solana under significant pressure, forcing it to accelerate upgrades to reverse the situation.

The AI field, which Solana is proud of, is also not optimistic. After the once popular DePIN and AI Agent MEME boom subsided, the Virtuals ecosystem rose rapidly, making Base occupy an absolute advantage in the AI Agent field; BNB Chain, leveraging celebrity effects and exchange resources, diverted much of the attention from AI MEMEs. In terms of decentralized AI infrastructure, the rapid expansion of Bittensor subnets and the rise of many AI Layer 1s are gradually forming new forces in the DeAI field.

For a while, anxiety seemed to have spread to all aspects of the Solana chain, but Solana, which has weathered many storms, is clearly not going to admit defeat easily. Currently, the Alpenglow upgrade is expected to bring strong confidence to the ecosystem. At the very least, Solana officials have shown sufficient humility and pragmatism, clearly benchmarking competitors, gradually making up for shortcomings, and even striving to be stronger.

Next, let's sort out Solana's AI ecosystem to see if Solana's AI narrative can help it usher in a new round of prosperity.

03. Solana's AI Landscape

The AI projects on Solana have been developing for a long time and span a wide range of sectors. Taking stock of them is impressive. Here, we can divide the AI projects on Solana into three stages according to approximate time intervals.

First Stage: Early DePIN Ecosystem Boom, DeAI in Its Infancy

The popularity of the DePIN concept made Solana one of the earliest public chains to explore on-chain decentralized AI. These projects, relying on Solana's strong performance and low prices, built their own decentralized computing power, bandwidth, data, and other networks, and also laid the early foundation for the Solana AI ecosystem.

In terms of AI computing power networks, early projects such as Render, io.net, and Aethir were among the first to attempt on-chain decentralized computing power, but with different focuses:

Render focuses on 3D animation and metaverse rendering tasks, recording the allocation and payment of rendering tasks on Solana, and matching GPU providers and demanders through smart contracts.

io.net mainly provides low-cost decentralized GPU computing power for AI and machine learning, automatically allocating computing power through its own network, with Solana used for settling \(IO tokens and recording node contributions. Aethir mainly optimizes real-time applications such as games through edge computing, with its own network allocating computing power to edge nodes, and Solana used for recording transactions and \)ATH token incentives.

These projects are among the early explorers in the already crowded decentralized computing power network. As early entrants, they have raised large amounts of financing and are highly favored by the market. However, these projects usually require high-performance GPUs to participate in the network, making the threshold too high for ordinary people.

In contrast, Grass, Helium, Roam, and Gradient Network have a better mass foundation.

Grass is a decentralized data collection network. Users can automatically contribute bandwidth by downloading a browser plugin or application, and collect network data through activities such as crawling. Users' contributions are recorded on the Solana chain, and they receive \(GRASS rewards. Roam's network aims to provide seamless, secure global internet coverage through community-provided WiFi nodes. Users can earn \)ROAM tokens by providing WiFi coverage or verifying the network in the APP, with the project using Solana for node management and payments.

Gradient Network is a decentralized edge computing platform. Users only need to install a plugin to contribute idle computing resources (such as mobile phones, laptops, etc.) to support scenarios such as AI inference, content distribution, and serverless functions, and handle coordination and payments between devices through Solana's on-chain mechanism to ensure efficient and secure decentralized computing.

Helium provides IoT and mobile network coverage in a decentralized manner, using Solana to record hotspot data and transactions. Users purchase and deploy Helium Hotspots (hotspot devices, priced at approximately \(200-500) to provide network coverage and receive \)HNT or \(MOBILE token rewards. A more accessible way to participate is to use the SIM card launched by Helium and T-Mobile in the U.S., enable the "Mapping" function, share anonymous location data to help optimize network coverage, earn Cloud Points or \)MOBILE tokens, and get a free trial of the SIM card.

These projects have actively explored the field of infrastructure combining blockchain and AI (such as bandwidth, data, computing power, etc.), not only bringing huge imagination and confidence to the market but also further opening a new chapter in DeAI. Although the token prices of these projects have fallen from their highs, network effects are still evident, and many projects have established cooperation with traditional large companies, winning recognition from the mainstream market.

Second Stage: AI Agents Flourish in Multiple Areas

With the significant improvement in the capabilities of LLMs represented by ChatGPT, applications such as AI Agents have developed rapidly in the past year. Solana, with its excellent performance and rich ecosystem, became one of the earliest public chains to benefit from on-chain AI Agent tokens and the prosperity of related applications.

Wayfinder, developed by Parallel Studios (invested by Solana Ventures, previously mainly known for the Solana-based blockchain game Colony), simplifies cross-chain operations through AI Agents, supporting easy cross-chain asset transfers between Solana, Ethereum, Polygon, and Base.

ElizaOS is an open-source AI Agent framework that helps developers quickly create, deploy, and manage AI Agents on Solana. These AI Agents can quickly process information, execute transactions, and interact with smart contracts, with governance through DAO AI16Z. The DAO token \(AI16Z once had a market capitalization exceeding \)2.5 billion, becoming the leading AI token on Solana. However, its MEME attribute was too strong, and its value later returned to rationality.

Holoworld is a multi-modal AI Agent launch platform that allows ordinary people to create, customize, and trade AI Agents, and supports 3D character and video customization. All digital asset interactions are managed through Solana's smart contracts.

Moby AI initially was an AI Agent focusing on Alpha research and providing real-time crypto information, and has now launched its own terminal, focusing on Alpha mining and trading on the Solana chain.

Hey Anon is a DeFAI protocol deployed on Solana, aiming to simplify DeFi interactions through natural language processing and real-time data aggregation, helping users manage DeFi operations, obtain project updates, and analyze market trends.

In addition, during the AI boom at the beginning of the year, many well-known Agent and application projects emerged on the Solana chain, such as the Agent frameworks \(ARC and \)SWARMS, the DeFAI platforms \(GRIFFAIN and \)BUZZ, the AI Agent launcher \(HAT, and the well-known AI Agents \)PIPPIN and $ZEREBRO, etc. However, after the boom subsided, the token prices and popularity of these projects dropped rapidly, and some projects were put on hold.

Third Stage: Post-On-Chain AI Era, Steadily Moving Towards On-Chain DeAI

As the market gradually returns to rationality, we have noticed that a new batch of projects has emerged on the Solana chain. They have shed the hustle and bustle of the AI MEME era and are building decentralized AI infrastructure on Solana in a more down-to-earth and feasible manner.

Nous Research is committed to training open-source AI models that can compete with mainstream large models such as OpenAI through decentralized means. Its core Psyche network significantly reduces the communication frequency between nodes through compression technology, effectively solving the communication bandwidth bottleneck of decentralized AI training, enabling the network to carry high-computation-demand AI training scenarios such as LLM pre-training; and using the Solana network to record node contributions and distribute incentives. It has currently launched the Hermes series of open-source models fine-tuned based on the Llama model.

Arcium was initially a privacy protocol Elusiv on Solana, later transforming into a broader privacy computing platform. Using MPC and ZKP technologies, it allows computing on encrypted data without exposing the data content, not only providing protection for user privacy in fields such as DeFi and Desci but also offering privacy infrastructure for AI model training, inference that requires attention to training data privacy, and AI Agent applications.

Neutral Trade is a hedge fund platform based on Solana, providing a variety of AI-driven quantitative and automated trading strategies. Currently, in addition to common quantitative neutral strategies and arbitrage strategies, it also has a well-performing CTA Momentum strategy, a multi-asset, high-liquidity strategy operated in cooperation with the well-known quantitative company R* Research, with an annualized return of 95.11%, and the $10 million subscription quota has been fully filled.

Although the popularity of Solana AI has declined in this stage, it is somewhat like sifting through the chaff. The quality of projects has significantly improved — with stronger implementation capabilities and solving more practical problems — evolving from "dream speakers" to "doers."

04. Views on Solana's Current AI Ecosystem

Through the above overview, we can see that the AI ecosystem on Solana has developed early, covers a very comprehensive range, and its overall strength is still very competitive among existing public chains.

While being comprehensive, many excellent projects have emerged in the Solana ecosystem. For example, Nous Research is committed to solving the problem of decentralized large-scale training of AI models, Grass collects massive training data through a decentralized network, and Arcium focuses on building privacy infrastructure. The competitors of these projects often require proprietary chains to support them, but they have developed well on Solana, fully demonstrating Solana's excellent carrying capacity.

Although the current popularity of on-chain AI has been diverted by Base, BNB Chain, etc., in the long run, the possibility of one chain dominating the AI field is almost zero, and there will be more cooperation than competition between chains. At the same time, with the Alpenglow upgrade, Solana's advantages in the AI field will be further strengthened:

1. Fast and Cost-Effective: Speed, throughput, and fees are the most concerned factors for on-chain AI projects. Multiple AI Agents under the management of protocols such as MCP need close cooperation and transactions, and decentralized training, data collection, and other tasks also rely on high-frequency, fast, and massive interactions between nodes. Currently, Solana is already very efficient; after the Alpenglow upgrade, the confirmation time will be shortened to 150 milliseconds, further reducing latency and costs, and better supporting real-time AI applications.

1. Good Liquidity: As the currency of projects, the price stability and smooth circulation of AI tokens have a significant impact on projects. Solana currently has a daily DEX volume of up to $1.4B, ranking second in the entire network only to Ethereum, and has a mature and active DeFi ecosystem such as Raydium and Jito. AI projects can easily access stable liquidity pools. After the upgrade, it is expected to attract more market makers, provide deep liquidity, and help AI tokens circulate and raise funds quickly.

1. Smart Contract Support: Solana supports parallel processing and more flexible programming languages. Its virtual machine SVM can process complex logic