ETHDenver 2026 Observation: Side Events Cut in Half, AI Explosion, Builders Seeking New Paradigms

Author | kyle Wu on Blockchain

The Number of Side Events Halved

Let’s start with the most direct data: the number of side events this year has clearly halved. According to statistics from Luma, compared to the 668 side events in 2025, this year's side events during the conference period (2026/2/18--2026/2/21) are only around 215, a drop of 68%.

There are several reasons behind this. On one hand, it is the industry cycle itself; on the other hand, the timing of the conference is close to the Lunar New Year, which is not friendly for Asian project teams and attendees. Additionally, some external events are also diverting industry attention, such as the WLFI Forum at Mar-a-Lago, which has drawn away some OGs and core practitioners.

Thus, while attendees can still feel that many North American Web3 people are gathering at ETH Denver, there are noticeably fewer people coming from outside the United States, and the overall spillover influence of the conference has weakened.

From the perspective of specific ecosystem activity, it is also evident that the industry has entered a phase that emphasizes efficiency and cost-saving.

Monad and X Layer have a strong presence; the former hosted three events, while the latter sponsored the main stage (but the team only participated in the first day of the exhibition); Aptos and Sui also had some exposure; Solana only hosted one event, which was small in scale but of good quality. Overall, many public chain ecosystems are clearly no longer pursuing the past strategy of holding numerous events to maximize presence, but instead are retaining only a few side events, holding one or two activities to make a point, focusing on cost-effectiveness.

AI is Everywhere

One of the biggest changes this time is that AI is almost everywhere.

In addition to the main stage, ETH Denver has also divided the main venue into five stages—ETHERSpace (Applications and User Ownership), Devtopia (Infrastructure and Developer Tools), New France Village (Finance / Compliance / Institutionalization), Futurllama (AI / DePIN and other cutting-edge trends), and Prosperity (Privacy, Public Goods, and Community). After three days of exploring the exhibition, Futurllama had the highest foot traffic.

The main venue also featured the Claws Out Summit focused on the recently popular "lobster" OpenClaw, as well as a small roadshow aimed at AI project founders; outside the venue, there was the Open AGI Summit hosted by Sentient, which had a scale and foot traffic density even higher than some main venue areas. This year's ETH Denver feels less like a traditional crypto industry conference and more like a mixed exhibition of AI × Crypto.

Even the forms of exhibiting projects have changed. When seeing robots and mechanical arms at the exhibition, one might have the illusion of "Did I walk into CES?" Projects like PrismaX and Gensyn, which focus on embodied intelligence, had independent booths. Additionally, many projects still carry the Web3 label, but their core narratives are no longer solely centered around public chains, DeFi, and wallet infrastructure; instead, they are moving towards areas like agents and chatbots, which are closer to the application layer of AI products.

A strategic manager from an exchange stated that at this conference, he was not focused on mature projects, but rather on very early-stage directions that were just taking shape. Regarding AI + Crypto, he emphasized that he was not looking for projects that immediately generate revenue, but rather whether the combination of AI and Crypto would lead to new production relationships, new distribution methods, new workflows, and new on-chain collaboration mechanisms.

He also shared with me an internal strategic judgment from their company's senior management: everyone has already assumed that AI will bring significant productivity improvements, but in the Web3 space, what is truly worth investing in is not the so-called large models, but whether AI technology can be embedded within exchanges to enhance the current product experience. They are currently developing a large language model embedded in the exchange that provides trading recommendations based on real-time market news and allows direct trading of related products within the chat box.

ETH Denver Remains Builder-Oriented

Although AI has become the hottest topic, one thing that hasn’t changed about ETH Denver is that it remains one of the strongest builder-oriented crypto conferences.

This is evident from the agenda. On the last day, apart from the closing ceremony, the entire exhibition time was dedicated to hackathons and Builder Workshops, and side events during the exhibition period for public chains like Base were also organized around developers.

It is worth mentioning that Base showcased a small product called Braindate during the conference. Base specifically set aside an area in its exhibition space for socializing and chatting. However, unlike traditional casual chat areas, you can join some interesting groups on Braindate or create a group to attract like-minded partners for discussions.

Inside the venue, this atmosphere is even more palpable: compared to discussing market trends, people coming to ETH Denver are more focused on "What are you building lately?" Especially student builders, who care more about "Can this thing be built?" and "Is this product interesting?" rather than short-term market fluctuations. Because of this, although ETH Denver has seen an overall scale contraction, it has not completely turned into a conference that only discusses narratives without talking about products. At least here, the true builders remain at the core.

The best embodiment of this spirit is still its hackathon, BUIDLathon, which has a noticeable change this year: the overall competition design has been moved forward.

According to feedback from participants, BUIDLathon has introduced an Online Hacking phase, where specific topics will be communicated a week in advance, moving away from the traditional "start working only when you arrive on-site" rhythm.

This is actually quite an interesting innovation. It aligns better with the current working methods of many builders and makes it easier to improve project completion rates. However, it also means that the in-person communication time has been shortened, from last year's 8 days to 4 days, and much of the preparatory work must be completed in advance. The offline event is no longer the starting point for building but rather resembles the final concentrated wrap-up of the project.

The changes in the competition format may only reflect a shift in the organizers' preferences, but the changes in the prize pool starkly showcase the cycle the industry is in.

After interviewing several participants, they all complained that the number of sponsors and their level of engagement this year were significantly lower than in previous years. Last year's total prize pool reached $1.03 million, while this year it has shrunk to $132,000. Sponsor budgets are more concentrated and lean towards the directions they care about most: AI.

One participant mentioned that a champion project in the hackathon had the gimmick of an "AI girlfriend." Users can tip the AI girlfriend with cryptocurrency, and through more frequent tipping interactions, establish a kind of behavioral incentive mechanism to bring the "two" closer together. This tipping mechanism sounds quite like a meme, but the project's award also highlights an issue: the judges are no longer limited to viewing the combination of AI + Crypto from the perspective of wallets, settlements, and other strong Web3 scenarios; they are more interested in seeing projects outline usage scenarios that are closer to ordinary, mainstream users.

I also interviewed Justin, a participant in the hackathon who won third place in the Base sponsorship prize. His project is an open protocol aimed at AI agent advertising monetization.

The judgment behind this project is that many AI products that are freely available to the public in the future may need to achieve sustainable monetization through advertising rather than subscriptions. Advertisers pay for promotional tasks, and AI agents distribute recommendations based on user context, driving conversions and earning revenue after task completion.

However, the biggest challenge is how to prove that the AI has indeed completed the advertising task.

Justin's solution is to introduce on-chain validators to verify advertising results. In this hackathon demo, he primarily validated verifiable behaviors such as on-chain purchases: only when the relevant delivery results are confirmed as genuinely occurring by validators will the protocol trigger settlement, and the advertiser's budget will truly be paid to the agent that completed the task. In this sense, this project is not just an "AI taking ads" product, but is attempting to build the infrastructure for verification and settlement in the future AI-native advertising market.

Another noteworthy point about this hackathon is the diversity of participants.

The on-site crowd included both university students and senior builders; there were individuals with deep experience and mature industry understanding, as well as creators with new and interesting approaches. Among the participating projects, in addition to AI and DeFi-related directions, there were also some interesting GameFi projects.

ETH Denver's hackathon has not turned into a game solely for the crypto circle. One can still feel many newcomers entering, and they do not necessarily carry the burden of having to work on traditional crypto usage scenario-related projects; instead, they are more willing to naturally integrate AI, gaming, advertising, social interaction, and other scenarios with on-chain technology.

Somewhat Makeshift Judging Process

Of course, this hackathon is far from perfect, and the judging process seems somewhat "makeshift."

Participants reported that the judging process was divided into a main track and a sponsor track. The main track was relatively clear: participants queued from 9 AM, usually facing 2 to 3 judges, and the entire pitch and demonstration process was not particularly tense. However, the problem is that the presentation time is only five minutes, which means that judges naturally lean towards teams that can clearly explain their projects and demonstrate them in a very short time. Under time constraints, having strong punchlines and memorable points often matters more than technical polish, and even bugs do not affect the chances of winning.

After completing the main track, teams also had to continue to the sponsor track to present their projects again to sponsors like Base. This stage was relatively more chaotic, with unclear management of the order of arrival, which tested the teams' on-the-spot judgment and execution.

Prediction Markets

Aside from AI, another theme from this conference that I think is worth mentioning separately is prediction markets.

At the Frontier Markets hosted by Monad, I heard several key discussion points shared by guests that left a deep impression. First, they believe that the biggest barrier to prediction markets remains liquidity. Secondly, since each market has an end time, liquidity constantly migrates, making it difficult to settle long-term like perpetual contracts. Furthermore, how to attract liquidity providers to long-tail markets remains a challenging problem.

At the same time, prediction markets have clear distinctions from traditional trading. Traditional trading tends to have gradual fluctuations, but prediction markets can often drop to zero at a certain point, complicating leverage design, market-making strategies, and risk management, which makes many traditional large market makers hesitant about entering.

However, prediction markets also have their own peculiarities: many popular markets naturally give rise to liquidity providers primarily composed of retail investors. In other words, the key in this space may not just be "to create another prediction platform," but rather who can continuously create sufficiently attractive markets for retail participation and reorganize liquidity around these markets to enhance user experience.

Resilience of Crypto

If I were to conclude this year's ETH Denver, I would say: it reflects the myriad of experiences in the industry during a bear market and a transitional phase. This is not a stage where there are clear hot topics and everyone quickly follows popular narratives. The excitement has diminished, the heat in the crypto circle has lessened, and the so-called "beauty of an economic upturn" is also less present. Yet, at the same time, there is still a group of builders seriously exploring, some investors are still looking at very early-stage projects with significant potential, and some projects, although not perfect and still having bugs, have already begun to reveal feasible business models and outlines for the next stage of the crypto space.