Most announced low-carbon molecule projects never get built. The gap between press release and commercial operation is wide, and for anyone allocating time, capital, or analytical attention across a pipeline of thousands of projects, knowing which announcements are credible matters enormously.
That is the problem the Delphi Project Viability Score was built to solve.
What the score does
The Project Viability Score assigns every project in the Delphi Hydrogen Project Database a calibrated probability of reaching commercial operation. It is not a qualitative rating or an analyst judgment call. It is the output of a statistically calibrated ensemble of logistic regression models, one per technology group, trained on the full history of projects that have either reached operation or been cancelled.
Six factors drive the score: technology and capacity, EPC contract commitment, offtake agreements, grant and subsidy linkage, financial milestones, and geographic attributes. Each factor receives a data-driven weight. The output is a single comparable number, bounded between 0 and 0.9 - a deliberate ceiling that reflects the reality that no single project in a nascent industry can be rated at absolute certainty.
The score covers hydrogen, CCS, biofuels, and beyond. I led its development at Delphidata, and we launched it as a free beta for all Delphi dashboard subscribers at the end of Q1 2026.
Use case: ATOME Villeta Green Fertiliser Project
In the scoring cycle preceding April 2026, ATOME's Villeta Green Fertiliser Project in Paraguay received a predicted probability of 0.80 - the highest score recorded across all of Latin America. Globally, it ranked 17th out of 2,650 projects in the estimation set.
At the time of scoring, the project was still in FEED.
In April 2026, ATOME publicly announced Final Investment Decision (FID).
This is what the methodology calls an exogenous, ex-post outcome: a result that happened independently of the model, after the score was assigned, and that can therefore serve as a genuine test of its discriminative ability. The score held.
Why this matters
A single case is illustrative, not conclusive. We are transparent about that. Systematic validation rests on aggregate model diagnostics - ROC-AUC, Brier scores, Events-Per-Variable ratios - across the full training set, not on individual examples.
But individual examples are how credibility becomes legible. When a model flags a project as the most likely to be built across an entire continent, and that project reaches FID while still in FEED, it is worth noting.
The score will continue to improve. Every project that reaches FID or gets cancelled feeds fresh outcome data back into the model, tightening the calibration over time. The hydrogen industry is still early. The dataset of observed outcomes is growing. So is the model's accuracy.
Try it
The Project Viability Score is available now as a free beta for all Delphi dashboard subscribers. If you are tracking hydrogen projects and want a systematic way to distinguish credible pipelines from noise, it is a good place to start.
If you have questions about the methodology, I am happy to walk you through it.