techcrunch.com détecté sur le web

Flapping Airplanes lève 1 Md$ à 6 Md$ neuf mois après le seed

7 médias qui couvrent ce sujet
Funding Chips NVIDIA ai-business

TL;DR

  • Flapping Airplanes quadrupled its valuation from $1.5B to $6B in nine months with no product, on a megakernel framework that fuses training and inference into a single GPU job.
  • The full investor syndicate includes Index, Lux, Sequoia, GV, Kleiner Perkins, Valor, Atreides, A*, and Nvidia, a near-consensus top-tier signal for a pre-revenue lab.
  • Nvidia backed the round despite Flapping Airplanes keeping AMD and Amazon chip compatibility open, suggesting it views efficiency-layer software as additive to GPU demand rather than a threat.

Neuf mois après un seed de 180 M$ à 1,5 Md$, Flapping Airplanes a bouclé une série A de 1 milliard de dollars à 6 milliards de valorisation, rapporte TechCrunch. Index et Lux co-mènent le tour, avec Sequoia, GV, Kleiner Perkins, Valor, Atreides, A* et Nvidia au tour de table.

Le labo n'a pas de produit public. Il revendique « the world's most complete megakernel framework. It can perform entire training runs, serve inference, play video games, and even do all three simultaneously in a single kernel ». L'idée technique : « overlap memory-bound inference with compute-bound training » sur une même puce.

Les trois cofondateurs cochent les cases que la Vallée récompense. Aidan Smith est « a former founder and Thiel fellow who worked at the brain implant company Neuralink during college ». Asher Spector est « a statistics PhD at Stanford ». Ben Spector est « a computer science PhD student at Stanford » et a fondé Prod au MIT, décrit par TechCrunch comme « an almost secret-society, five-year-old student-run startup accelerator ». David Cahn, partner chez Sequoia, les présente comme bâtissant « the young person's AGI lab ».

La thèse économique tient en une phrase d'Asher Spector : « A model that's a million times more data efficient is probably a million times easier to put into the economy. » L'ambition revendiquée est d'entraîner de grands modèles « without ingesting half the internet ».

L'opération s'inscrit dans un flux soutenu de levées IA — 269 tours suivis par notre tracker sur les 90 derniers jours, au lendemain des 15 Md$ bouclés par DeepSeek.

Ce qu'en disent les autres médias

Couverture consolidée 3h après publication

  1. Bloomberg Lire →

    Bloomberg wire confirmation of the $6B valuation with independent sourcing, adding institutional credibility to the investor consortium details reported by other outlets.

  2. SiliconANGLE Lire →

    Deepest technical breakdown: megakernel framework explained, full founder CVs (Stanford PhD, Thiel Fellow, Neuralink), and investor targeting of robotics and life sciences.

    A model that's a million times more data efficient is probably a million times easier to put into the economy. — Asher Spector
  3. The Next Web Lire →

    Frames the round as a valuation-trajectory story (4x since January seed) and places it within the emerging neolab trend of research-first AI startups that shun early commercialization.

    The basic thesis is demand for AI is going to outpace chip and data supply. — Asher Spector
  4. Tech Funding News Lire →

    Only outlet with a named Sequoia partner quote and the detail that Nvidia invested despite AMD and Amazon chip openness; also surfaces Prod accelerator lineage to Cursor and Etched.

    This is not a one-of-many company; this is a one-of-one company. — Sequoia partner David Cahn
  5. Dealroom Lire →

    Full investor consortium named alongside market context: AI took 64% of global venture funding in Q3 2026, framing this round as a sector-concentration data point.

  6. Runtime Wire Lire →

    Leads with the efficiency mission over the valuation number, emphasizing the founders' intent to pursue radically different AI architecture approaches.

    We want to try really radically different things.