LEAP's startup winners tell a different story from its AI-heavy exhibition floor

Two very different LEAPs happening at once Walk LEAP 2026's exhibition floor or read through the event's official announcements and one theme dominates: artificial intelligence. HUMAIN, AWS, Microsoft, Adobe and other technology companies used the event to announce cloud regions, data centres, AI models and partnerships, part of a set of announcements — also spanning manufacturing and venture capital — totalling nearly $15 billion, according to my.gov.sa's official summary of the event . Yet look instead at who actually won LEAP's Rocket Fuel startup competition, and a noticeably different picture emerges — one dominated by satellites, cooling fluid, telemedicine, water technology and devices rather than large language models or AI agents. The two audiences overlapped on the same days: the event drew 1,289 investors from 1,016 firms, according to TechScoop's roundup, the same room the Rocket Fuel finalists were pitching to. The winners, again, for context As TechScoop has covered across its Rocket Fuel coverage, this year's seven winners were Waspito (telemedicine and mobile labs), SARsatX (synthetic-aperture-radar satellites), Ferveret (immersion cooling for data centres), Carevision (healthtech market entry), Melt Water (water technology), Imagine Devices (early-stage devices) and Kanz (recognised for founder leadership). The seven were chosen from 100 finalists, themselves selected from more than 3,000 applications spanning more than 40 nationalities, and shared $1 million in equity-free prize money. None of the seven is described in TechScoop's reporting as, at its core, an AI model or AI-agent company. The full detail on each winner sits in TechScoop's roundup of every 2026 Rocket Fuel winner and in TechScoop's follow-up on what happened to all seven winners since the event . Why the mismatch is not actually a contradiction The gap between an AI-saturated exhibition floor and a hardware-and-infrastructure-heavy winners' list is less surprising once you separate what each part of LEAP is actually optimised for. The keynote-stage announcements and corporate partnerships — the HUMAIN, AWS, Microsoft and Adobe deals TechScoop catalogues in its full rundown of everything AI announced at LEAP 2026 — are large, capital-heavy commitments from established companies and government-linked entities. The Rocket Fuel competition, by contrast, is judged by a panel evaluating early-stage founders on the merits of a specific product and team, which is a fundamentally different filter than the one that produces headline corporate partnership announcements. What a pitch panel is judging A competition judging panel evaluating dozens of pitches is looking for a defensible, hard-to-copy technical advantage and evidence a founder can execute — criteria that often favour deep, physical-world engineering problems like satellite radar or immersion cooling fluid dynamics over a thinner software layer built on top of someone else's foundation model. Agentic AI was the most common product framing on the LEAP floor this year, as TechScoop's roundup noted, but the specific companies that made it through to win prizes were, this year at least, mostly working on problems where the underlying hard technology, not an AI wrapper, was the actual differentiator. What this says about where genuine technical scarcity lives There is also a simpler explanation worth taking seriously: the skills required to build a synthetic-aperture-radar satellite, an immersion-cooling system, or a cross-border telemedicine and mobile-diagnostics network are scarcer and harder to replicate than the skills required to build another AI-powered chat interface, and judging panels tend to notice that scarcity. When many teams are pitching some version of an AI product, differentiation becomes the hardest thing to prove — whereas a working satellite radar system or a functioning immersion-cooling deployment is, almost by definition, evidence of a harder-to-fake technical achievement. Two signals worth taking from LEAP 2026, not one The honest read of LEAP 2026, then, is that it produced two genuinely different signals rather than one contradictory one. The corporate and government-linked announcements describe large commitments to AI infrastructure and models in Saudi Arabia. The Rocket Fuel winners say that, at the individual-startup level, some of the most credible and fundable technology being built around the event is not AI-branded at all. Founders and investors reading LEAP's outcomes for signal should treat both facts as true simultaneously, rather than assuming the loudest theme on the exhibition floor was also the theme that won the competition judges over. What this means for founders pitching the next LEAP For a founder weighing whether to apply to next year's Rocket Fuel competition, this year's results carry a fairly direct lesson: an AI label attached to a pitch is not, on its own, a differentiator when a judging panel is looking at dozens of similarl