The Problem: AI Models Need Data That Doesn't Exist
The gap between AI can do this and AI can do this here, with this data, at this cost is where most environmental monitoring projects die. Vienna-based startup Another Earth just raised €3.5 million to close that gap – not by launching more satellites, but by generating synthetic satellite imagery that doesn't exist yet.
This matters for anyone deploying AI systems in data-scarce environments. The funding, announced this week, comes from Wake-Up Capital alongside existing investors Rockstart, Inovexus, and Stamco AG, with institutional backing from Austria's Research Promotion Agency (FFG), Austria Wirtschaftsservice (AWS), and the European Space Agency (ESA).
Here's the implementation reality that rarely makes it into pitch decks: Earth observation AI systems require massive amounts of labeled training data. Real satellite imagery is expensive, covers limited geographic areas, and requires time-consuming manual annotation. For regions experiencing rapid environmental change – exactly the places where monitoring matters most – the data situation is even worse.
As TechFundingNews reports, Another Earth's Synthetic Data Engine addresses this by combining generative AI with procedural 3D simulations to create physically accurate, customizable Earth imagery. The output is fully labeled, multispectral, and temporally consistent – meaning organizations can train AI models for environmental monitoring without waiting years to collect real-world data.
Founded in 2020 by Maya Pindeus and Felix Geremus, the company has positioned itself at the intersection of two trends: the explosion of AI capabilities and the growing urgency of climate monitoring. But the real value proposition isn't the technology itself – it's what the technology enables for teams with real constraints.
What the Funding Actually Buys
The €3.5 million will fund three specific initiatives, according to Arab Founders:
Expansion of the Synthetic Data Engine. This is the core platform that generates high-resolution synthetic satellite imagery. The expansion will enable more biomes, environmental conditions, and geographic contexts.
Enhanced environmental monitoring and risk simulations. The platform will add capabilities for predicting environmental risks before they escalate – moving from reactive crisis response to predictive intervention.
Scaling operations in Brazil and Sub-Saharan Africa. These regions represent critical ecosystems where real satellite data is scarce but environmental monitoring is urgent. Another Earth has already partnered with GeoTerra Image in Sub-Saharan Africa and NovaTerra in Brazil to deploy its platform.
With this funding, and our deployment into vital ecosystems spanning from Latin America to Africa, we are generating data where there is none. We are giving organisations the tools to transition from reactive crisis response to proactive, predictive intervention.
Maya Pindeus, CEO and co-founder
Implementation Considerations: What Makes Synthetic Data Work
Synthetic data isn't new. What's new is synthetic data that's good enough to train production AI systems for high-stakes environmental decisions. The implementation details matter here.
Automatic labeling eliminates annotation bottlenecks. Traditional satellite imagery requires human annotators to label features – forests, water bodies, infrastructure, land use changes. This is slow and expensive. Another Earth's synthetic data comes pre-labeled, which according to the company cuts labeling costs significantly.
Scalable generation for rare scenarios. Real satellite data has sampling bias – some regions and conditions are overrepresented, others barely exist in training sets. Synthetic data can be customized by biome, location, or environmental conditions, enabling AI models to handle edge cases they'd never see in real-world training data.
Bias-free, ethical data. Synthetic data avoids privacy issues and sampling bias inherent in real-world satellite imagery. For public sector deployments where data governance matters, this is a meaningful advantage.
The competitive landscape includes AI.Reverie (now part of Meta), Rendered.ai, and Sensity AI. Another Earth differentiates by focusing specifically on Earth Observation and climate resilience, with partnerships tuned to local realities in target regions.
What Could Go Wrong
No implementation guide is complete without the failure modes. Synthetic data for Earth observation carries specific risks:
Domain gap. Synthetic data that looks realistic to humans may still confuse AI models trained on it when they encounter real-world imagery. The gap between synthetic and real distributions can cause performance degradation in production. Teams deploying Another Earth's data should validate model performance on held-out real-world test sets before production deployment.
Temporal consistency challenges. Environmental monitoring often requires tracking changes over time. Synthetic data must maintain physically plausible temporal relationships – seasonal changes, gradual deforestation patterns, infrastructure development. If the synthetic data doesn't capture these dynamics accurately, models trained on it may miss critical real-world patterns.
Overconfidence in data-scarce regions. The whole point of synthetic data is to enable AI deployment where real data is limited. But this creates a validation problem: how do teams verify that models trained on synthetic data perform well in regions where, by definition, there's limited real data to test against?
Regulatory uncertainty. As AI governance frameworks mature – particularly the EU AI Act – the use of synthetic training data may face new scrutiny. Teams should document their synthetic data sources and validation procedures now, before compliance requirements crystallize.
The Broader Pattern: Infrastructure for AI Deployment
Another Earth represents a broader trend in the European AI ecosystem: companies building infrastructure that enables AI deployment rather than AI models themselves. The model is the easy part. The hard part is getting the data, the governance, the monitoring, and the organizational readiness in place.
For public sector technologists and policymakers, this funding signals growing maturity in the European climate tech stack. Environmental monitoring AI has moved from research demonstrations to production deployments, and the infrastructure layer is catching up.
For startup leaders and investors, the deal structure is instructive. The combination of VC funding (Wake-Up Capital, Rockstart, Inovexus, Stamco AG) with institutional support (FFG, AWS, ESA) reflects the hybrid funding models that work for deep tech with public benefit applications. This isn't a pure commercial play – it's infrastructure that serves both market and policy objectives.
For AI researchers, Another Earth's approach raises interesting questions about the role of synthetic data in closing the gap between AI capabilities and real-world deployment. The company's focus on physically accurate simulation rather than pure generative AI suggests a hybrid approach that may prove more robust for high-stakes applications.
What to Watch
The next twelve months will reveal whether Another Earth's synthetic data approach scales beyond pilot deployments. Key indicators to monitor:
Validation results from Brazil and Sub-Saharan Africa deployments. Real-world performance data from these partnerships will determine whether synthetic data can actually enable production AI systems in data-scarce regions.
Regulatory developments around synthetic training data. The EU AI Act and related frameworks may impose new requirements on AI systems trained with synthetic data, particularly for high-risk applications like environmental monitoring.
Competitive response from larger players. Meta's acquisition of AI.Reverie signals big tech interest in synthetic data. Whether Another Earth can maintain its focus on Earth observation while larger players enter the space will shape the market.
The company's trajectory from research to large-scale deployment is exactly the transition where most AI projects fail. The funding provides runway. The partnerships provide market access. The question is whether the implementation details – the unglamorous work of making AI systems actually work in production – can match the ambition.
For teams building AI systems that depend on geospatial data, Another Earth's approach offers a potential solution to the training data bottleneck. But as with any infrastructure dependency, the implementation details matter more than the pitch deck. Before committing to synthetic data as a training source, validate performance on real-world test sets, document the data provenance for compliance purposes, and build monitoring systems that can detect domain gap issues in production.
The gap between demo and deployment is where projects die. Another Earth is betting €3.5 million that synthetic data can help close it.
This is precisely the kind of infrastructure question that deserves deeper examination – not just what the technology can do, but how it actually gets deployed, governed, and validated in production. For those building Europe's AI future, these implementation details are the conversation that matters. That conversation continues at Human x AI Europe in Vienna on May 19, where the people shipping AI systems gather to share what actually works.
Frequently Asked Questions
Q: What is synthetic satellite data and how does Another Earth generate it?
A: Synthetic satellite data is artificially generated imagery that mimics real satellite observations. Another Earth's Synthetic Data Engine combines generative AI with procedural 3D simulations to create high-resolution, fully labeled, multispectral imagery that can be customized by biome, location, or environmental conditions.
Q: How much funding did Another Earth raise and who invested?
A: Another Earth raised €3.5 million in funding led by Wake-Up Capital, with participation from existing investors Rockstart, Inovexus, and Stamco AG, plus institutional support from Austria's FFG, AWS, and the European Space Agency (ESA).
Q: What regions will Another Earth expand into with this funding?
A: The company will scale operations in Brazil (partnering with NovaTerra) and Sub-Saharan Africa (partnering with GeoTerra Image) to produce synthetic satellite data for biodiversity monitoring, deforestation tracking, and climate risk analysis.
Q: What are the main risks of using synthetic data to train Earth observation AI models?
A: Key risks include domain gap (models performing differently on real vs. synthetic data), temporal consistency challenges, validation difficulties in data-scarce regions, and regulatory uncertainty as AI governance frameworks evolve.
Q: When was Another Earth founded and who are the founders?
A: Another Earth was founded in 2020 by Maya Pindeus (CEO) and Felix Geremus. The company is headquartered in Vienna with operations in London.
Q: How does Another Earth's synthetic data differ from traditional satellite imagery for AI training?
A: Unlike traditional satellite imagery, Another Earth's synthetic data comes pre-labeled (eliminating manual annotation costs), can be generated at scale for rare scenarios and underrepresented regions, and avoids privacy issues and sampling bias inherent in real-world data collection.