AxionOrbital Space’s first technical claim is a number: 0.06 seconds. That is the inference latency the company says its foundation model achieves, translating raw radar satellite data into a human-readable optical image in near real-time [Y Combinator LinkedIn]. The promise is to turn the unintelligible squiggles of Synthetic Aperture Radar (SAR) into something a commodity trader or military analyst can understand at a glance, without retraining their vision pipelines.
Founded in 2025, the San Francisco-based startup emerged from Y Combinator’s Winter 2026 batch. Its core proposition is straightforward. Legacy optical satellites are blind at night and useless through cloud cover, which occurs roughly 70% of the time [Y Combinator]. SAR satellites, which use radar pulses, see through both. But their data is not optical. It is a complex backscatter signal that breaks standard computer vision models and requires specialist interpretation. AxionOrbital says it bridges that gap, delivering “analysis-ready optical imagery” around the clock [Y Combinator].
The technical wedge
The company’s differentiation rests on a proprietary architecture it calls Deterministic One-Step diffusion. In a LinkedIn post, Y Combinator cited the model’s performance benchmarks: an FID score of 30.24 and an SSIM of 0.6, which the post claimed “resets the SOTA” [Y Combinator LinkedIn]. The startup has announced two initial models:
- Hubble. An open-source model offering 10-meter resolution, positioned as an entry point for developers and researchers [Y Combinator LinkedIn].
- Orion. A restricted, higher-fidelity model boasting 0.5-meter resolution for commercial and government applications [Y Combinator LinkedIn].
The architecture is designed to anchor image generation to physical spatial priors, constraining outputs to reality rather than artistic generation.
The founders and the early capital
The team is a compact, technically dense unit of two co-founders. Dhenenjay Yadav, the CEO, is an MBA candidate at IIM Ahmedabad with a background that includes a stint as an ML Engineer at ISRO’s National Remote Sensing Centre and prior founder experience with an AI wearables company in India [YC Tier List, Fortune, 2026]. Atharva Peshkar, the CTO, is a Computer Science PhD candidate at CU Boulder and a former research assistant at Harvard University [Y Combinator]. Their path to YC was not straightforward; according to a Forbes profile, Peshkar’s team was initially rejected by the accelerator, rebuilt their model, and were accepted on the second attempt [Forbes, 2026].
| Round | Amount | Lead Investor | Year | Source |
|---|---|---|---|---|
| Pre-Seed | $100,000 | Lobster Capital | 2025 | [Lobster Capital, 2026] |
The ambition beyond translation
The radar-to-optical translation is the wedge, but the founders’ ambition stretches further. Yadav has described work on “NeoEarth,” a persistent world model for Earth observation designed to predict how terrain and infrastructure evolve up to a month in advance [Dhenenjay Yadav - AxionOrbital Space (YC W26) | LinkedIn, 2026]. The company also claims to offer clients access to an archive of over ten years of radar data, with structured pricing for either mission-specific analysis or long-term continuous monitoring partnerships [Preqin, 2026].
Where the wheels could come off
For all its technical promise, AxionOrbital operates in a field of well-funded, operational competitors. The commercial and execution risks are pronounced:
- Unproven commercial traction. Public records show no named customers, no disclosed revenue metrics, and no evidence of a deployed product with paying users [YC Tier List].
- Capital-intensive competition. The company is up against established SAR satellite operators and data providers like ICEYE, Capella Space, and Umbra.
- The integration challenge. Even with perfect translation, convincing large, risk-averse organizations in defense or finance to integrate a new, unproven data feed into core decision-making workflows is a formidable sales undertaking.
The next twelve months
The immediate roadmap is about moving from technical claims to commercial proof. Key milestones to watch will be the announcement of a first enterprise or government pilot customer, a subsequent funding round to scale beyond the current pre-seed capital, and more detailed, public case studies demonstrating Orion’s 0.5-meter resolution in action. The $100,000 pre-seed from Lobster Capital provides runway, but the jump to a meaningful seed round will require a signed contract.