Investing in Crypto has become significantly more difficult ever since Trump took office, with broader uncertainty pushing capital into risk-off assets.

The whole world watches as the tariff situation keeps getting worse and worse. Crypto is no different—BTC has shown signs of strength, and Fartcoin has shown an even bigger sign of strength, outperforming everything.

Everything Else Is Struggling

But beyond these two assets, everything (and I mean literally everything) is struggling—Crypto AI, which has consistently been the #1 mindshare sector, drastically declined, maintaining around $6BN in total MC. DeFi is no better, with over $50BN erased from on-chain TVL as capital flees to other safe assets outside of crypto.

So, Where Do We Invest?

This begs the question: How & what do we invest in during a choppy market?

Most people I know will probably point to yield farming on Berachain, Sonic, yadayada—which is fine. But for me, there are a lot more interesting opportunities to tap into with better R/R, especially in times of crisis.

To me, the most asymmetric bets right now lie in the intersection of DeAI infra and AI agent (more on this in a bit).

Sticking to the mantra: “Be cheerful when others are greedy, be needy when others are fearful.” That doesn’t sound right. “Be fearful when others are greedy, be greedy when others are fearful.” Yes!

Crypto AI Segments I’m Watching

The way I see it, there are multiple sub-segments in Crypto AI that are particularly interesting right now:

• Dev tools — frameworks, vibe coding tools, MCP infra
• Decentralized AI infra — decentralized compute, verifiability, deployment, confidentiality, storage, ownership
• Consumer AI — AI agents, alpha tools, gaming, DeFAI, GambleFAI, personality / companion

(This is not inclusive of all sub-segments, but you get the picture.)

Framework Trends

For more granularity on Consumer AI / AI agents and dev tools, I created this thread in March (initially planning to do monthly threads like this but doesn’t seem like the general agent market is moving at a pace fast enough to update monthly):

Frameworks were valued at a pretty high FDV earlier in the szn (Oct–Nov last year), but when devs realized there are a lot of things you can’t do with off-the-shelf frameworks, and LLMs may not be the best for financial use cases (prone to prompt injection), demand for frameworks waned.

Regardless, we’re still seeing growth in open-source frameworks & tools like @elizaOS (15.5k stars on GitHub), @arcdotfun (3.4k stars), and @sendaifun (1.2k stars), gaining 434 stars, 197 stars, and 110 stars respectively over the past month.

Why Agent Distribution Networks > Frameworks

I personally don’t find frameworks that exciting since there’s not much value accrual in them. It’s much better to be investing in distribution networks / agent hubs since there’s clear value accrual there—i.e. trading fee derived from trading volume that stems from speculators/investors trading AI agent tokens. @virtuals_io remains the best at this. Even though the daily volume fell from 8–9 digits range to 7 digits, Virtuals remains the most trusted ecosystem for developers as well as the most diverse ecosystem with many teams trying to build unique agent products.

@elizaOS is starting to look more interesting, especially since @autodotfun (their launchpad) just went live. The team now has a distribution network to directly accrue value back to the $ai16z token. @shawmakesmagic teased a few interesting features (that have yet to be rolled out):

What they need to nail down is the execution for high-quality partner project launches in order to meaningfully differentiate from what Virtuals offer (otherwise they’ll be stuck with low-effort shitters with 4–5 digit MC).

Anyhow, taking a step back, while these AI agents, frameworks, and distribution networks are interesting, the best R/R segment to invest in right now is Decentralized AI infra.

Why?

If you’ve been in the AI agent trenches for a while, you’ve probably noticed that the progression of agent products has been something like:

Conversational “agent” for entertainment ➔ Conversational “agent” for alpha analysis / tools ➔ trading agents ➔ DeFAI abstraction layers ➔ other smaller narratives ➔ agents with smarter context, multi-agent / swarm, etc.

The Death Flywheel Trap

The reason why many teams struggle is because there wasn’t any proper core AI product within those “agent products”. The only AI was to automatically prompt LLMs to yap every x interval.

Obviously a lot has changed from the early days, but the reliance on LLMs or off-the-shelf frameworks/workflows remains the same so in every progression of the agent products / every narratives create sub-par agent products without proper use cases. (Similar vibes to teams forking major defi protocols & fading into irrelevance a year later)

This led to many teams generating hype with their agent launches alongside their tokens, but then failing to retain that attention (because there’s no actual product), resulting in a death flywheel (attention goes down, token goes down).

Agent Builders Need Infra; Infra Builders Need Agents

But, while these teams may fail, they’re good at one thing—it’s GTM/generating hype.

If there are many teams who are good at GTM with agents, know how to play the token game/build the community, but lack a proper AI product—what should they do? They should tap into specialized AI models & ML capabilities from inference networks and DeAI infra providers.

On the other hand, DeAI infra teams aren’t good at GTM. They aren’t in the trenches, some aren’t crypto-native, and have no idea how to build community.

So… why not combine the two?

The missing link between deep AI infra and viral agent distribution is where I think the real opportunity lies.

My Crypto AI Investment Thesis

This brings us to my Crypto AI investment thesis:

Which is investing in DeAI infra as well as agent teams that are introducing new, unique Web3 workflows that change the way people interact with existing crypto products (DeFi, on-chain).

In Web2, workflow automation & augmentation—boosting productivity (and therefore profit) while minimizing cost—is very common across vertical agent sectors, especially for mundane tasks (the more mundane, the better the value). e.g.

Legal AI agent ingesting raw paper documents, creating databases for legal cases, working with lawyers to help their clients succeed in court

Accounting agents going over receipts, invoices, GL, TB, and generating unaudited FS and tax filings

Architecture agents reviewing building floor plans, estimating costs, and suggesting ways to reduce construction/material costs while maintaining durability & design alignment with client needs.

There are so many case studies like this in Web2, and these startups grow to 7–8 fig ARR very quickly (in months) because of this—they really use AI agents to automate & augment workflows offering real value to other businesses/clients.

In Web3, this is still quite new & complex. To really enhance workflows in DeFi, you need domain expertise. You need to understand the pain points that DeFi users (and normies) face—and how to make it better. DeFAI abstraction layers tackle this to a certain extent, but most remain unusable with sub-par reasoning capabilities (you gotta prompt very specific prompts to make it work—which really defeats the purpose, cuz ideally you want normies to use it, and normies usually have no idea what they want to do, so they naturally don’t know what to prompt).

This is why I think teams that understand how to meaningfully change Web3/Crypto workflows are pretty rare. BUT if you can spot them & invest in them early (right now), you’ll receive a lot of the upside in the future.

On the other end of the spectrum, we have DeAI infrastructure. Most of it isn’t investable because it’s still early stage.

These teams tend to raise millions from VCs & take a few years to go to TGE. A few that have launched saw 50–80% price declines due to market conditions. The ones that hang on really well need to generate substantial revenue to maintain token price (or hire a really good MM lol).

A good example of this is @getgrass_io—supposedly 8–9 figs in revenue & a great consumer-facing product (anyone can contribute bandwidth for airdrops).

Projects like Grass are pretty rare in VC-backed DeAI infra, and usually the only way you can participate early is by using the product / farming the airdrop. They’ll likely pump the token price at TGE (low-float, high FDV style) since VCs got in at relatively low valuations. And if you decide to invest in something like this, there’s a higher chance you lose money than make money.

Investable Community-First DeAI Ecosystem

Which brings us to the alternative—pure community / no VC DeAI ecosystem. Yep, it’s @opentensor (Bittensor).

Pre-dTAO upgrade, the ecosystem was quite boring. Validators acted as sort of capital allocators because they decided which subnet got the $TAO emissions (capital).

But ever since the dTAO upgrade went live on Valentine’s this year, the dynamic drastically shifted. The market now determines which subnets receive emissions. The community—the people—is now the capital allocator. If the community decides your subnet has no product & doesn’t provide much value, you don’t get emissions (capital). This pushed subnets to build publicly, ship faster, and build products that people actually want.

The shift also spurred the creation of hedge funds specifically investing in Bittensor subnets:

@BarrySilbert betting on the Bittensor ecosystem with @YumaGroup (subsidiary of DCG), which invests in, builds, and incubates Bittensor subnets. The most recent interview between @RaoulGMI and @BarrySilbert sparked a lot of excitement in the community (because a major crypto institution has now entered the Bittensor eco):

From an investment perspective, liquidity in the Bittensor ecosystem is a lot better than in the AI agent eco. The core problem for agent ecosystems like Virtuals is that LP is paired with Virtuals, which leads to higher volatility & more IL for liquidity providers.

This is why liquidity is usually thin—you can usually deploy $1k–$5k and experience 3–7% slippage on these agent tokens. On the other hand, deploying similar amounts into subnet tokens yields something like 0.05%–0.1% slippage (or even less).

Quick Summary:

Crypto AI agent hype cycle is fading, real product + retention still rare

DeAI infra is undervalued, misunderstood, and mispriced

Best plays combine infra + agent GTM to unlock new workflows

$VIRTUAL leads agent meta, Bittensor leads infra meta

Watch for teams merging these—huge upside if spotted early

Wrapping things up

I believe DeAI will define the next trend in Web3 AI. We’ll see more teams that change how we interact with each other & with protocols, shift how value is created, and generate new segments that reach more users & mindshare (more mainstream). Now is the best time to get up to speed with DeAI infra and how it changes things. Make sure to keep your eyes peeled on teams that can successfully combine DeAI & agents.

Please bear in mind, my thesis isn’t rigid. I’m continuously learning & refining it. I’m trying my best to make sure we’re positioned to capture the next big trend in Web3 AI. Again, this is not financial advice—do your own research and take everything here with a grain of salt.

In the next Bittensor article, I’ll be spending more time diving into interesting subnets and uncovering more opportunities.

Disclaimer:

1. This article is reprinted from [0xJeff]. All copyrights belong to the original author [0xJeff]. If there are objections to this reprint, please contact the Gate Learn team, and they will handle it promptly.
2. Liability Disclaimer: The views and opinions expressed in this article are solely those of the author and do not constitute any investment advice.
3. The Gate Learn team does translations of the article into other languages. Copying, distributing, or plagiarizing the translated articles is prohibited unless mentioned.