Y Combinator, the accelerator that backed Airbnb, Stripe and Coinbase, has issued a new request for startups that reads like a time capsule from 2017 with a 2026 twist. The October Crypto Request for Startups, shared publicly by visiting partner Nemil Dalal, asks founders to make capital raising for artificial intelligence companies easier, and it points to a surprising precedent for how to do it: Ethereum’s initial coin offering.
The connection between AI funding and crypto rails may sound odd at first, but the logic is straightforward once laid out. Y Combinator describes the current wave of investment in AI as one of the largest capital-raising events in human history. Training frontier models, buying compute clusters and paying research teams requires sums that were once the province of nation-states. Yet the machinery for raising that capital remains slow, gated and expensive, dominated by a small circle of institutional investors.
Ethereum’s ICO era, whatever its excesses, proved something important: that blockchain networks could move billions in capital from a global pool of small backers into new projects in weeks rather than months. Y Combinator’s October document asks founders to build the modern, compliant version of that machinery, deliberately leaving the exact financing structure open for entrepreneurs to invent.
A continuing agenda, not a pivot
The request is not a sudden conversion to crypto. Y Combinator and Coinbase published a joint request in September 2025 seeking startups working on stablecoins, tokenization, trading infrastructure and on-chain capital formation, a document that emerged alongside a partnership with Coinbase’s Base network. The October update gives AI capital raising an explicit seat inside that continuing agenda, slotting the biggest financing story of the decade into the crypto toolkit the accelerator has been quietly championing for a year.
Stablecoin banking for the underserved
Beyond the headline AI item, the October request devotes significant attention to stablecoins, which Y Combinator sees as the most immediate product-market fit in crypto. The document asks for dollar stablecoin products that reach customer or merchant groups the existing financial system has failed to serve, along with stablecoins denominated in additional currencies.
Neobanks serving global contractors and international businesses appear as an explicit target. The accelerator also flags gaming as an underserved segment for dollar-based digital money, an interesting angle given how much of the gaming economy already runs on virtual currencies. Notably, the request demands a distribution strategy, not just a product: founders must identify a specific group that benefits and explain how to reach it.
Agents as the next class of trader
Perhaps the most forward-looking section concerns AI agents. Y Combinator expects software agents to become active financial participants and prolific traders, sometimes operating under human direction and sometimes acting on their own standing instructions. Today’s exchange interfaces and APIs were built for human developers clicking buttons and writing code. The October document asks founders to build trading applications and infrastructure designed for agents from the ground up.
The API category goes further, focusing on financial connections purpose-built for software actors. Y Combinator argues that blockchain access alone will often cover only part of an agent’s task, and that completing it will require stitching together multiple services, including fiat-to-crypto on-ramps. The accelerator’s broader thesis expects entirely new asset classes to be issued on-chain, including startup equity and small securities, with trading, lending and derivatives services blossoming around them.
Institutional use cases are cited as evidence the tide has already turned, with Y Combinator pointing to oil hedging and faster cross-border movement of enterprise money as drivers of crypto uptake over the past year.
What it means for everyday investors
For retail crypto investors, the significance of this request is validation from the heart of Silicon Valley. When the accelerator that funded a generation of tech giants starts asking for crypto infrastructure that serves AI fundraising, stablecoin banking and agent-driven trading, it signals that the next wave of crypto startups may be built for customers with enormous budgets and concrete problems.
The Ethereum ICO comparison also carries a warning worth remembering. The 2017 boom produced a handful of durable protocols and a mountain of failed projects and regulatory enforcement actions. If AI capital raising does move on-chain, the winners will be the infrastructure providers who build compliance in from the start rather than the projects chasing a quick raise.
Founders who want in on the Winter 2027 batch, which runs January through March in San Francisco, should apply by November 2 for guaranteed review timing. For everyone else, the document is worth reading simply as a map of where the smartest money in Silicon Valley thinks crypto is going next.
yc pointing AI founders at the 2017 ETH ICO playbook is hilarious. half those icos went to zero and theyre citing the era as a success story
They cite it because it actually moved billions from small global backers in weeks. The excesses got filtered out, the rails worked.
The rails worked, yes, and the scams worked too. The honest read is YC wants the 2017 mechanics with 2026 disclosure law bolted on, which is a very different product than what actually shipped back then
2017 mechanics with 2026 disclosure law is generous, half that cohort shipped nothing but a whitepaper. the contractor payroll wedge is the only piece that survives contact with regulators
the doc explicitly says build the compliant version. the speed is what they want copied, the 2017 scams were the bug in the shipping
YC pointing at the ethereum ICO as the template is wild. half those 2017 raises were outright scams and now it is a fundraising blueprint lol
the agents-as-traders section is the underrated part. every exchange api out there assumes a human clicking buttons on the other end
right, agent wallets need machine speed rate limits and non custodial keys. half the exchange api docs out there still gate everything behind human kyc flows lol
nemil dalal basically wants token sales for compute clusters. launchpads for gpus. ok, im listening
launchpads for gpus is exactly it. collateralize the cluster, trade the claims, borrow stables against it. sengupta basically wrote the pitch deck already
everyone obsessing over the ICO line while the stablecoin neobank section quietly prints. dollar accounts for global contractors without the swift headache is the actual wedge
stablecoin neobank wedge is real but its also the most regulated bite of the apple. contractor dollar accounts means licensing in thirty jurisdictions, the ico speedrun crowd is not built for that grind