B2A (by a16z) is the new M2V & M2M https://t.co/4wf49bCWXf

Crypto x AI adoption is at ATH > GPU marketplaces hit 8 figures of annualized revenue > Onchain Inference is closing in on 2T tok/day > Bittensor DeAI is the most productive they've ever been > Mega (prev. MegaETH) is pivoting into Onchain Inference Existing builders are doubling down, new builders are entering into the sector In today's article, we explore the thesis behind what makes Crypto x AI one of the strongest moats of Crypto in 2026-27

Top 3 most undervalued onchain inference teams @AskSurplus, @dphnAI, @engyai (MC-to-token usage) Both Surplus + Engy growth driven by increase in deepseek-v4.1-flash consumption Dolphin usage spike comes from the usage of large swarms of flipper agents using Gemma 4 model Market ↓ Fundamentals ↑

US midterm wrapped up Market rallied... and dumped hard The blue wave is bringing more oversight into AI and Crypto-alike And the Onchain/Decentralized AI thesis will benefit the most from it In this week's After Hour, we're covering - Latest on the market, on AI, and on the blue wave era - Potential dip-buying opportunities + What I’m personally doubling down on

Citrini's "State of the Themes" article kinda strengthens the Onchain/Decentralized AI thesis   The article argued that the AI infra/semis trades are ending because Neoclouds/Hyperscalers CAPEX is hitting ~$1T while AI eco ARR is only $200B and the huge gap won't be filled any time soon. Plus, financing constraints, construction delays and model price competition add pressure to the trade   That creates an interesting setup for projects helping operators monetize existing capacity and buyers access cheaper compute.   This is where Onchain/Decentralized AI gets interesting   1. Turning idle/crowdsourced GPUs into revenue   GPU marketplaces like @lium_io help operators rent out unused capacity, connecting existing hardware with paying customers.   For data centers and smaller operators, better utilization means more revenue from assets they’ve already paid for. Buyers get more options and competitive pricing (e.g. rent idle B200s on @lium_io for $4.25/hr).   As financing gets tighter, making existing hardware earn more becomes increasingly valuable.   2. Building markets for surplus/excess inference   Inference Providers and AI gateways have inference capacity they can sell. Some are expirables because they run subsidy campaigns but they couldn't get users to fully use them.   @AskSurplus connects these participants through an inference marketplace, routing requests across discounted supply. Selected offers carry discounts of 50–99% relative to listed API prices.   This creates a way for sellers to monetize unused allocations and buyers to reduce their inference bills.   Onchain settlement and tokenized credits add building blocks for developers to create new markets, routing tools and products around that supply.   3. Bootstrapping inference networks with existing GPUs   Decentralized AI crowdsources idle GPUs across the world.   Token emissions are used to pay for compute & inference instead of debt & equity financing huge CAPEX for data center buildouts.   Teams like @engyai and @dphnAI are already proving that it's possible to host top open-weight models on a hybrid/fully decentralized network that's fast, reliable, and cheap for users.   The opportunity is to turn fragmented hardware into useful, competitively priced inference.   The economics become stronger as paying customer demand grows enough to support suppliers with progressively smaller subsidies.   4. Expanding access to GPU financing   @USDai_Official connects onchain capital with operators financing GPU infrastructure.   That gives smaller operators another route to funding hardware purchases and deployments, with loans secured by GPUs and supported by their cash flows.   As traditional financing becomes more selective, access to additional capital sources could become increasingly valuable.   5. Making inference access a tradable asset   Pilots of inference tokenization are happening onchain first   @AskVenice tokenized its platform credits into "DIEM", people are buying & selling inference like a commodity and builders are building on top of it, creating venues where users can speculate & earn from their credits.   This creates the market for future inference access. Users can acquire capacity, trade the token representing that access, and earn revenue from it (if they don't use it).   Other teams are experimenting with similar models, creating more ways to price, distribute and monetize inference.   For tokens, this creates a link with extraneous demand outside of crypto. The more people use the platform, the more valuable the credits, the more valuable the token.   The investment case gets stronger when usage translates into value for token holders   The best part is that this sector is growing fast and teams are starting to align interest with token holders through buybacks & burn, revenue distribution, and more utilities.   Onchain/Decentralized AI provides the opportunities for better GPU utilization, cheaper inference, new financing channels and tradable access to compute/inference.   All real economic problems that crypto teams are building products around them.   2025-26 is the year of AI infra/semiconductors   2027 will be the year of Onchain/Decentralized AI