In the latest issue of Hivemind Podcast, Delphi’s core team members had an in-depth conversation about the prospects of the encryption market in 2025. The host Jose (head of Delphi Labs), Ceteris Parabus (head of institutional research department), Duncan (small currency investment expert) and Jason (market director of research department) discussed the evolution of AI narrative, Bitcoin L2 ecology, and infrastructure Value return and other hot market topics.
* Note: The podcast was recorded on January 16, which is quite different from current market data and popular tracks.
market observation
Ceteris:
The market has been very volatile these past few weeks. Looking back, after the FOMC (U.S. Federal Reserve Open Market Committee) meeting at the end of December last year, there was a wave of selling in the market. The price of Bitcoin was at a high of nearly $100,000 at the time, but plummeted to below $90,000 in just a few hours.
My main takeaway is that this market volatility has not changed the fundamentals. For example, in the past decline, the best-performing asset was still Bitcoin, while some DeFi projects and old Meme coins (OG Memes) still did not make much movement. This shows that the market's interest points have not fundamentally shifted. The best-performing projects in the rebound, such as Virtuals and ai16z-related frameworks, are still those assets that performed well at the bottom.
In summary, I believe that the market fluctuations this time are affected by macroeconomic factors, such as CPI data and other policy rumors, as well as technical adjustments. Price charts for many tokens have been “crazy” over the past few weeks, and these extreme price moves naturally trigger pullbacks and liquidations.
Jason:
Many fluctuations are related to market positioning. For example, last week's news that the DOJ wanted to sell $6 billion in Bitcoin now seems to be false or over-interpreted. As Trump's inauguration approaches, it's actually a positive if they do intend to sell but haven't done so.
From my perspective, last week's macro data was better than expected, which could boost near-term targets. However, I feel that the more it rises before a major event, the greater the possibility of a "sell message." Like Trump is inaugurated next Monday, if Bitcoin continues to rise before then, I would be worried about a short-term pullback, just like what happened when the ETF was approved.
But on the other hand, Trump may issue a series of executive orders as soon as he takes office, as he did in 2016, some of which may target cryptocurrencies. Given the relatively small liquidity of crypto markets, he is fully capable of influencing crypto markets as much as he affects traditional financial markets, perhaps even more.
Ceteris:
I see that Trump’s youngest son and his circle are involved in cryptocurrencies, which may mean that he sees Bitcoin prices as a “stock market indicator” for young people. While this claim needs to be verified, it is believable considering his children were involved. If he does start talking about being the first “crypto president” or “Bitcoin president,” there are indeed many ways he can help advance cryptocurrencies.
Jose:
There are a lot of levers that Trump can pull to help crypto, but obviously we had a massive post-election rally and now the market is pricing in that surge. I think the size of future market moves will largely depend on what Trump actually does with cryptocurrencies and the impact of his appointees on the larger macro market , such as Scott Benson's performance as Treasury Secretary.
The decline leading up to Trump's election felt like a good buying opportunity . I think there is a very simple corollary to Trump's election - he can easily pass some policy measures to benefit cryptocurrencies. A cryptocurrency bill is currently being pushed forward in Washington, and the industry has gone from being ignored or even generally disliked to now having a “Crypto Bowl” and a soon-to-be-established Crypto Advisory Council. So really without overthinking it, I think Trump is clearly bullish on cryptocurrencies . Of course, there is also the possibility that he will do nothing, and it is not uncommon for politicians to fail to deliver on their promises, especially Trump, who is not the most trustworthy person. But I think it's in his interest to advance cryptocurrencies , so I'm still quite bullish.
AI bubble
Jose:
In fact, I bought a lot of currencies related to the AI framework during this decline. As far as I'm concerned, I think AI itself will be the biggest narrative in 2025. My overall view is that we may achieve artificial general intelligence (AGI) by 2025. The investment in capital expenditures by hyperscale technology companies is staggering. Microsoft alone plans to spend $80 billion. Other companies have invested almost all of their free cash flow into the construction of computing infrastructure for training AI models. Progress in 2025 may be faster than in 2024.
To me, the definition of AGI is AI that reaches or exceeds human levels in multiple fields and has a certain degree of autonomy to perform tasks for the user. Simply put, it's like a research assistant or investment analyst with PhD-level intelligence.
Currently, only a few people in Silicon Valley and the technology community truly understand how AI will impact the world, but in the next few years, the world will realize this. I believe this will be one of the biggest bubbles in financial history because the fundamentals of AI are so strong and the most compelling stories can be built around it. Never before has a technology had such strong fundamentals at such an early stage, such as OpenAI's revenue and the potential future revenue from coding assistant tools.
Despite the strong fundamentals of AI, humans will always be overly excited about some emerging things, leading to cyclical fluctuations in the market. I don’t think the big AI bubble has really arrived yet, but its emergence is almost inevitable . Even assuming there are no fundamentals at all for crypto AI, I believe the capital flows themselves are enough to move the market. Just imagine, how do ordinary investors participate in the AI narrative? Although the valuation of the "MAG7" (the seven major technology companies including Microsoft, Apple, and Google) is high, it is not unaffordable; however, buying stocks of large technology companies cannot easily achieve a counterattack of wealth. In addition, stock selection in the stock market is very difficult. Many people think it is a "rigged game" and venture capital is out of reach for 99.9% of people. So, all that’s left is to bet on crypto AI projects.
I think this area, such as the Agentic (autonomous) sector, currently has a market value of about US$10 billion, and the entire crypto AI market has a market value of about US$50 billion. As we mentioned in our forecast, this size will only continue to grow.
Ceteris:
My strategy has always been "act first, study later". I think the focus now is to get in touch with those popular tracks, and when truly valuable projects emerge, determine their status in the industry. At the same time, I'm betting on the growth of the industry as a whole. However, last week's market performance reminded us that volatility in this space can be very high. For example, the prices of some projects may drop by 50% in two days, and such violent rotations may occur repeatedly in the next year. Therefore, I will not stick to AI tokens all year long, but will be flexible based on market dynamics .
The main leaders in the market right now are Virtuals and ai16z, but I don't think they have insurmountable moats.
- The advantage of Virtuals lies in its built-in token value capture mechanism. For example, users need to pay Virtuals tokens to activate agents, which provides them with stable cash flow.
- Ai16z, on the other hand, relies more on the widespread usage of its framework, but currently lacks a clear profit model.
Ultimately, the dominant force in the market lies in the attractiveness of the platform, especially its ability to attract more developers and users. It is worth noting that this is also the first time that users’ attention to the chain itself has declined. Many agent functions are executed off-chain, and the chain mainly provides support for token trading or liquidity. Therefore, users care more about the functions and convenience of the platform itself rather than which chain it executes on.
Moat for Proxy Tokens
Jason:
I think the main moat right now is traction, which is the ability to serve as a launchpad, which is key to capturing value. A Launchpad's value depends on how many users and attention it attracts, and whether those users think the platform's fees are reasonable. This is why Binance can charge projects 3% to 10% of the token supply as Launch Pool fees, and Virtuals has proven that it can do this. But such moats are fragile and competition can be fierce.
There will be brutal capital rotation in the market in the future, and it may be difficult to fully keep up with these changes. I hope there will be time in the future to see different platforms emerge and be able to reduce risk. Take TAO, for example. It started out as the dominant option in the field, but as better options continued to emerge, its position began to erode. Current leaders, such as Virtuals and ai16z, will likely face the same situation. Right now, they are the main bets on this narrative, but in the future there will be better solutions and new narratives emerging in the AI space. Once that happens, I think it's wise to keep reallocating the portfolio. This is more of a trading move than a long-term holding option.
Ceteris:
Personally, I have yet to find any team that I feel confident about long-term. Although some people will say that infrastructure such as "mining tools" should be invested, whether these "tools" will always be effective will not be determined even by the end of this year.
Duncan:
For example, Pump.Fun is quite dominant in the Launchpad space, but I doubt Virtuals can maintain that position long term.
Ceteris:
Judging from the information I have learned, the actual technical framework of Virtuals is not excellent, but it is more about the convenience of starting the agent that attracts users. For example, a team originally released an agent through Virtuals, but recently switched to the Eliza framework. Although this agent still belongs to the Virtuals ecosystem and is paired with Virtuals to execute on the Base chain, the technical core has shifted to Eliza.
Jose:
This situation is actually not uncommon. Some people who were originally bullish on Virtuals switched to the ai16z framework after using its products. This shift has caused many originally bullish investors to become less optimistic. I think this phenomenon may appear repeatedly in the market. For Virtuals and other AI tokens, the market is rife with uncertainty about their future. Still, these projects are currently seen as credible options, so a lot of money is pouring into them, fueling their short-term growth.
In addition, from my communication with industry insiders, I found that there are still many people in the market who are skeptical of AI narratives and have not participated in investment. This shows that there is still a lot of potential capital that may enter this market . Although current AI projects are considered the only options worth buying, they may perform better in the future if new projects with greater long-term value and greater moats emerge.
Jason:
As for the technical aspects, my team is currently studying these platforms in depth, including their technical trade-offs. Larry's point - "encrypted AI agent frameworks have almost no technical advantages and big tech companies can replicate these capabilities in a matter of weeks" - may be correct, but it doesn't say much because the tech giants have extremely high Resource advantages. However, I agree with the current state of the market that these projects are the only investment options that seem credible, and capital is pouring into them.
The Decline of OG Memes
Jose:
I think AI is essentially a bigger Meme than WIF. It's easier to tell stories and easier to get excited about because there are at least some fundamental possibilities. I think that in 2025, not only Meme coins, but most of the capital flows in the cryptocurrency market will turn to AI. The shift in the meme coin market may be more immediate, as AI agent memes are somehow better than traditional memes and can tell larger stories.
Jason:
I basically agree with this point of view. This may not be because AI is inherently better than Meme, but because the market is tired of traditional Meme. People have been playing FROG, WIF, BONK, etc. for over a year. While making money is fun, most people want to buy something that has at least some actual value or path to value.
But that doesn't mean I have to buy those things. You can hold both beliefs at the same time when the market clearly tells you that this is not what people are paying attention to at a certain time. This is also the key to being a good tactical trader – it’s not what you believe or wish to be true, but what the market tells you.
For most of 2024, the market has been telling us that it favors meme coins, whether it’s because of a lack of new users entering the market, people chasing patterns from previous cycles, or a poor regulatory environment. But starting in mid-November, the market made it clear that it preferred AI.
I think some of the OG Meme coins may still do well, but the March wave of Meme coins may be dead. In fact, I think most meme coins and these proxy projects will eventually die. Someone should tell the Coinbase listing team that the era of meme coins is over. But looking at the DOGE chart, I don't think it's peaked yet.
I think DOGE is a good benchmark for evaluating the overall meme coin market . The meta narrative has definitely shifted, and the easy money opportunities are now in the agent and AI space. This doesn’t mean that all meme coins are doomed, just that they won’t be the only area worth investing in like they were last year.
Now that Trump is coming back, people may be more willing to invest in cryptocurrencies, so I could consider DeFi coins or projects like Ethena. Now that there is someone in the White House who supports cryptocurrencies and AI, maybe I should buy all AI coins because there may be some positive catalysts.
Duncan:
In the long run, except for a few projects like Bitcoin or DOGE that have reached escape velocity and mind share, any project without actual business support may eventually tend to zero, especially considering the speed at which new projects are launched now.
return to fundamentals
Ceteris:
I agree with the view that projects without fundamental support are unsustainable in the long term. I think we're entering a new era on the infrastructure side, and it's difficult to sustain these projects without a long-term revenue story. We've seen that projects that can actually generate revenue, such as some L1/infrastructure projects, are getting more attention.
According to the latest data from Blockworks, more than 50% of the revenue in the encryption field now comes from Solana applications. (Of course, not all of these incomes belong to Solana. JTO will take a small part, but this part is not included in the income statistics. These are directly attributed to the pledgers.) Specifically:
- Solana accounts for 56.8%
- Ethereum accounts for 17.9%
- Base accounts for 8.7%
- BNB accounts for 4%
- Arbitrum accounted for 1%
If Solana can hit 80% on these metrics by the end of the year, it's hard not to believe it has the potential to overtake ETH . There are two key indicators here:
- The revenue generated by the application is the so-called “on-chain GDP”
- R value, which is the most objective indicator to measure the economic value generated by the chain
When Solana generates 80% of the industry's revenue, other softer metrics such as "Ethereum has more developers" or "Ethereum is more decentralized" become less important.
Duncan:
I agree with this point of view. We did see bubbles in infrastructure tokens, high FDV low liquidity tokens, and meme coins. The market is moving towards some kind of equilibrium, where base coins that generate real revenue will find buyers at a certain price, while meme coins or overvalued infrastructure tokens will have a hard time finding a price bottom because they essentially have no revenue. .
Jose:
I would also like to point out that there are some issues with these metrics. For example, the income indicator mainly only calculates DEX and on-chain currency markets. And as new mechanisms like Pump.Fun emerge, the revenue distribution model may change.
Ceteris:
On the topic of app-specific sequencing, I think crypto researchers may be pushing it a little too far. Many people believe that applications will obtain 100% of the economic value on the chain, but this is not the case. There is a synergistic relationship between the application, L1, and verifiers, which requires some form of benefit distribution. While L1 does take 100% now, this won't last. Even if L1 only gets 30% in the end and applications get 70%, with the expansion of the chain and the addition of more applications, this model is still very beneficial to L1.
I prefer app-specific ordering over app chains because it maintains the same shared state. But there is a trade-off – every time we invent a new mechanism like this, it brings with it new centralization power to some extent. UniChain is the first project to implement pure application ordering from the beginning, so I'm very interested in its development.
People always say that application chains should be built to capture the full economic value, but if you look at the GDP chart, you will know that the most profitable applications are on Ethereum . Some people say that these applications should build their own application chains, but this completely goes against what the fundamentals tell us – the most profitable applications are on Solana , why leave? Just because you can earn more income?
This does not take into account the trade-offs that come with building independent chains. For example, Pump performs very well on Solana, and maybe in the long-term perfect world perspective, it will be better as an independent chain, but for now it is able to take advantage of the capital pool and shared state on Solana.
We did see bubbles in infrastructure tokens, high FDV low-circulation tokens, and meme coins, and now it may be the turn of the AI concept. Although the overall prospects for AI may be great, with these skyrocketing tokens now, over time, the market will eventually return to equilibrium. Those fundamental projects with real income will find buyers at a certain price, while those meme coins or overvalued infrastructure tokens are difficult to determine the bottom because they basically generate no income and can only be measured by subjective indicators. .
The market is maturing, which is a good sign. Those severely overvalued infrastructure tokens, high FDV low-circulation tokens or meme coins are losing momentum, and investors are beginning to focus on projects with real business support and sound fundamentals.
Duncan:
If there was a Chain Abstraction wallet with a particularly good user experience, would the differences between these chains become less important?
Ceteris:
My point is that it really depends more on user experience than Shared State. However, chain abstraction cannot completely eliminate the differences between multiple chains because of inevitable latency issues when using different state machines. This means that performance will be reduced due to cross-chain operations.
For example, a portfolio tracking tool like the Lighthouse.one app is a manifestation of the chain abstraction. It integrates users' assets on different chains and displays them in a front-end interface, which is very convenient. But if cross-chain interaction is required, such as switching between different state machines, performance will always be affected. Therefore, although the chain abstraction improves the user experience, from a technical perspective, it cannot completely replace shared state.
Next generation L2 and L3
Jose:
Regarding L2 development, I think they did well last year. Although there is no unified token to invest in, and this success does not directly benefit ETH, it is still very successful from a product perspective. L2’s total locked volume (TVL) doubled from 22 billion to 40 billion. However, it should be noted that a large part of Base's TVL comes from Coinbase's customer deposits, which is more like a centralized exchange doing accounting processing on its own proprietary blockchain.
Looking forward, I am bullish on Solana and L2, and I think that by 2025, at least one large financial institution will announce the launch of its own L2. Institutions like BlackRock may put tokenized securities on L2 as this provides them with greater flexibility while solving some of the limitations of traditional blockchains. Of course, this "private blockchain" model may cause controversy about decentralization, but if securities are really going to be put on the chain, this will be a more realistic option.
Ceteris:
Currently, only three L2s have more TPS (transactions per second) than Ethereum, which is incredible. Ethereum’s throughput is very limited, only about 12 TPS. We see a winner-take-all effect, where the strong get stronger. Specifically, the lower TPS of these L2s is due to a lack of demand rather than a lack of technical capabilities. I expect Mega ETH may do well when it launches, but its funding will likely be diverted from existing L2 projects . There is actually not much ETH currently bridged to L2, only about 2.5%.
Ceteris:
I am also optimistic about the development of the application level, especially projects related to AI agents. This is also infrastructure to some extent, but it is different. As I said before, people are starting to care less about whether a project is built on BSC or Solana, which is a good trend.
I think it's interesting that a lot of people thought Solana wouldn't succeed because they thought the technology wasn't important, which has always seemed to me to be a weird perspective. It's clear now that technology does matter and you can see the entire industry moving towards these high-efficiency systems. My main conclusion is: there is no reason not to build great apps now .
Bitcoin dominance
Duncun:
Against the backdrop of an increasingly relaxed regulatory environment, will we see an Altcoin boom, or will Bitcoin’s dominance continue to rise?
If you look at the data of BTC.D (Bitcoin Dominance), it has been rising steadily since 2022. It was 40% in 2022, peaked at 61% a few weeks after the election, and is now around 58%. We did experience a wave of small Altcoin in November and early December, somewhat similar to what happened in March.
The key question is: will this become the new normal? Just like in March, when Bitcoin formed a temporary top near $72,000 and then fluctuated in the $60,000-70,000 range, Altcoin generally fell. Or will this be a stronger cycle that brings in new capital and allows Altcoin to truly outperform?
Jose:
I feel like even if BTC dominance remains, you may still see some small cap coins outperform. In 2021 and 2022, when Bitcoin peaked, we saw crazy Altcoin runs. However, after Bitcoin hit a new high in March this year, most Altcoin suffered heavy losses during Bitcoin's consolidation period. Traditionally this would have been Altcoin season, but things have changed.
I think the concept of “Altcoin season” in the traditional sense may be outdated. Instead, projects with strong fundamentals and a clear narrative will outperform , while others will underperform.
Jason:
Speaking of the top prediction for BTC dominance, this is basically discussing the ETH/BTC ratio, because it is mainly a battle between these two currencies. Of course, XRP and BNB are also at the forefront now. Theoretically, if the entire AI sector surges 10 times, it may also have a little impact on BTC's dominance.
From this perspective, I think Bitcoin may still outperform Ethereum this year , especially if there are some favorable executive orders. It was a bold prediction, but I made a similar prediction last year and it came true. However, the gap may not be that big this year, as factors such as an improving regulatory environment and a crypto-friendly president will benefit ETH, as can be seen from the flow of funds in ETFs.
Six months ago I expected BTC dominance to reach 66%-68%, but now it seems that it may not reach that high. If Bitcoin rises to $125,000-150,000, I think BTC dominance may return to 62%-63% and then start to fall back.
Duncun:
It is worth noting that BTC dominance was 73% in January 2021, and later dropped to 40%. In contrast, the recent wave from November to December only fell from 61% to 55%, and the fluctuation range is only about one-eighth to one-tenth of 2021.
Bitcoin L2 feasibility
Jose:
I think Bitcoin L2 may become a dark horse in 2025. In the past, Bitcoin holders rarely did anything with their Bitcoins, and WBTC only accounted for about 6% of the Bitcoin supply at its peak. The adoption rate of existing multi-signature L2 is not high, but we may see some real ZK L2 launched on Bitcoin in 2025, such as projects such as Citrea or Alpen Labs .
This could have a huge impact because Bitcoin holders do value security and the market is huge. Bitcoin L2 not only provides expansion capabilities, but also enables programmability. You can already use smart contracts on Ethereum L1, but there is almost nothing you can do on Bitcoin L1. Therefore, the changes brought by Bitcoin L2 are even greater – not only greater expansion benefits, but also programmability. Even if only 1-2% of Bitcoins enter the L2 ecosystem, it is enough to increase L2's TVL several times.
Ceteris:
I am looking forward to the development of Bitcoin rollup. I’ve never believed people who say “Bitcoin holders don’t want to use DeFi” or “don’t want to do anything with their Bitcoin”, it’s always been ridiculous to me. Sure, there are some loud people on Twitter who may never do it, but this is a $2+ trillion asset that a lot of people are going to want to use, even if it's just to get an on-chain loan, and it doesn't require any Package.
Currently, a trusted bridge is required to achieve this, but if OP_CAT passes, trustless bridging can be achieved through ZK certification and other methods. OP_CAT has been running on the Bitcoin testnet for about a year, and some say it may be officially launched in 2025. This feature actually existed in the early days of the protocol and was later removed due to a catastrophic vulnerability, but this problem no longer exists.
I think Starkware's Starknet may be a dark horse for Bitcoin L2 . Once OP_CAT passes, they can start submitting proofs to Bitcoin. The space is vibrant right now, which can only be good for Bitcoin. Even if people don't use Bitcoin DA (Data Availability), it makes sense to just submit proofs to Bitcoin.
Even if only 2.5% of Bitcoin holders use these services, that would be equivalent to a $50 billion market. And not only existing holders will use it, but there will also be people who specifically buy Bitcoin to use these services.
Jose:
I think this year we'll see some non-managed solutions. It's worth writing a new report to update on all Bitcoin L2 developments, as our existing report is almost a year old. If Bitcoin L2 does take off, this could put a lot of pressure on Ethereum.
There are currently about 10,000 Bitcoins in existing Bitcoin L2, such as W-BTC (Wrapped Bitcoin) and so on. I think this number can grow about 10 times in the future, especially if it is not multisig. Bitcoin holders are actually very concerned about this issue, because many of these L2s may turn to EVM in the future. If there is a ZK bridge, or an extremely reliable code library (such as Aave or Morpho) that allows you to deposit Bitcoin and borrow stablecoins such as USDC, or do a Maker CDP (debt warehouse) operation, I think this kind of solution will definitely have a certain market.
Jason:
From a financial perspective, the need for lending is there, and BlockFi used to do a lot with deposits. After the launch of Bitcoin ETF, traditional finance also began to provide Bitcoin custody solutions. Will it also provide Bitcoin credit solutions? Will more people tend to use these crypto-native solutions in the future, or will they choose a platform like JP Morgan, which although it is centralized, has more safeguards and room for remediation if something goes wrong? .
I believe that the main source of demand for Bitcoin in the future is likely to be non-crypto-native markets . There are currently approximately $32 billion in Bitcoin in an L2-like state, including Bitcoin on Ethereum, Arbitrum, Base, Solana and other chains. This information is very interesting, although most of the solutions are not really "secure" solutions, except for the Lightning Network. However, if a real solution emerges for Bitcoin, this number will increase significantly.
