A $75M+ raise at a $1.5B valuation — still in talks, not closed — tells a larger story: capital is now pricing optionality on structurally different routes to advanced AI, not just writing bigger checks to the incumbents.
TechCrunch reported on July 13, 2026 that Nous Research — the independent AI lab founded in 2023 by Jeffrey Quesnelle, Karan Malhotra, Ryan Teknium, and Shivani Mitra — is in talks to raise at least $75 million at a $1.5 billion valuation. The round is being led by Robot Ventures, with Union Square Ventures among the participants. Nous had previously raised approximately $70 million from a syndicate that included Paradigm, Robot Ventures, North Island Ventures, OSS Capital, and Balaji Srinivasan. Both the round size and the valuation remain unconfirmed; these are reported talks, not a closed deal.
The company’s flagship product, Hermes, is an open-source AI agent with roughly 214,000 GitHub stars and 40,000 forks — adoption metrics that place it among the most widely used independent model projects outside the hyperscaler ecosystem. A hosted version is offered in commercial tiers running from $20 to $200 per month. That disclosed revenue base is modest by any venture standard; the valuation is not pricing current cashflow but the thesis and the option value embedded in the architecture.
Separately, Nous is developing Psyche: a decentralized network that pools hardware across multiple operators to run distributed computation and model training. Psyche is the structural argument made concrete — a direct challenge to the assumption that advancing the frontier requires a single operator’s multi-billion-dollar data center. Together, Hermes and Psyche define Nous’s position: open weights on one axis, decentralized compute on the other.
Founded by Quesnelle, Malhotra, Teknium, and Mitra. Hermes open-source weights released publicly.
Pre-2026
~$70M raised from Paradigm, Robot Ventures, North Island Ventures, OSS Capital, Balaji Srinivasan. Hermes reaches 214K GitHub stars, 40K forks.
Active
Psyche decentralized training network in development. Hosted Hermes tiers launched at $20–$200/month.
July 13, 2026
TechCrunch reports new round in talks: $75M+ at $1.5B valuation. Robot Ventures leading; Union Square Ventures participating. Round not yet closed.
The key insight: Nous’s $1.5 billion reported valuation cannot be explained by its current revenue. It is priced on two structural bets — open weights as a distributionstrategy, and decentralized compute as an infrastructure alternative — at a moment when investors are actively building a diversified portfolio of architecture theses rather than concentrating only in the incumbents.
The Structural Read
For most of the early 2020s, frontier AI investment followed a single organizing logic: close the weights, own the compute, scale the parameters, and charge for API access. That model produced OpenAI, Anthropic, and Google DeepMind as the dominant reference points — hyperscaler-backed, capex-heavy, and architecturally convergent. The field looked, from the outside, like a race with a settled rulebook.
What is changing now is not that the incumbents are weakening — they hold the users, the revenue, and the compute today, and that is not in dispute. What is changing is that capital has begun pricing a second hypothesis: that the current dominant path is one option, not the settled endpoint. The result is a widening field of independent labs — neolabs — each pursuing a structurally distinct architecture and each raising serious capital on the premise that the frontier is genuinely open.
Nous sits on the most direct challenge to the centralized-capex model. Open weights lower the cost of adoption to near zero for developers — a distributionstrategy that trades margin for reach and network effects in the way that open-source databases, operating systems, and programming languages have historically done. Psyche takes the argument one layer deeper: if inference can be distributed, can training be too? If yes, the multi-billion-dollar data center ceases to be a prerequisite for pushing the frontier, and the barrier to entry collapses structurally rather than just commercially. That is the bet Psyche is making, and it is live and unproven.
Nous is not alone in that broader category. Richard Sutton’s Oak Lab is building on continual learning from experience rather than scaled static text — a different architectural premise entirely. Yann LeCun’s trajectory toward world models represents a third path that disputes the primacy of next-token prediction at scale. And open models are already capturing a rising share of real developer usage, as token-level data from platforms like OpenRouter confirms. What these neolabs share is not a common technique but a common premise: that the architecture question is still genuinely open. Their differentiator is the design choice, not the size of the compute bill.
Architecture Thesis
“Capital is no longer betting on a single winner in a single-architecture race. It is building a portfolio of architecture theses — open vs. closed, centralized vs. decentralized, scaled prediction vs. continual learning — because the structural question of which path reaches advanced AI is genuinely unresolved. Nous’s round is one entry in that portfolio, not a verdict.”
Read through the lens of The War of Agents’ Architectures and the Open vs. Closed Meta-Framework, Nous’s position becomes structurally legible: it is placing both levers simultaneously — open distribution and decentralized infrastructure — as a compounding bet against the closed, centralized model. The risk is that neither lever has been proven at scale against well-resourced incumbents with years of production data and established developer trust. The option value is that if either lever proves decisive, Nous arrives at it from a structurally lower-cost position.
Three Implications
IMPLICATION 1 — FOR DEVELOPERS
Open weights with 214,000 GitHub stars represent a distributionflywheel that commercial tiers can monetize without requiring exclusivity. If Psyche delivers on decentralized training, the developer-accessible frontier expands — not because Nous outcomputes the incumbents, but because the coststructure of accessing frontier-adjacent capability drops. That is a structurally different competitive dynamic than API pricing wars.
IMPLICATION 2 — FOR INVESTORS
A $1.5 billion venture valuation on disclosed revenue of $20–$200/month tiers is not a cashflow multiple — it is optionality pricing. Robot Ventures and USV are not buying Nous’s current P&L; they are buying exposure to the scenario in which decentralized compute and open weights prove to be durable structural advantages. That is a legitimate bet, but it requires the thesis to resolve, not just the product to ship. Investors entering at unicorn prices on an unclosed round in a field defined by well-capitalized incumbents are taking on meaningful execution risk.
IMPLICATION 3 — FOR THE COMPETITIVE MAP
The neolab funding wave — Nous, Oak Lab, the world-model researchers, safety-first independents — is functionally a market signal that the AI architecture question has not closed. When capital funds structurally diverse approaches simultaneously, it is expressing uncertainty about which architecture wins, not confidence in a settled outcome. For the incumbents, the implication is that moat-building through scale alone is insufficient if structurally lower-cost alternatives can reach comparable capability. The hedge they need to watch is not the next GPT-class competitor — it is the architecture that makes the coststructure of the current model obsolete.
Business Engineer Framework
The Map of AI: Where Nous Sits in the Stack
The Map of AI traces 200+ companies across 9 layers of the AI stack — from compute and infrastructure through models, agents, and applications. Nous competes at the model and agent layers while building Psyche to challenge the infrastructure layer assumption. Understanding where any lab sits in this map — and which layers it is attempting to vertically integrate — is the clearest lens for reading its strategic logic and its threat surface against incumbents.
Gennaro is the creator of FourWeekMBA, which reached about four million business people, comprising C-level executives, investors, analysts, product managers, and aspiring digital entrepreneurs in 2022 alone | He is also Director of Sales for a high-tech scaleup in the AI Industry | In 2012, Gennaro earned an International MBA with emphasis on Corporate Finance and Business Strategy.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
