In case you missed it: Halcyon has closed a $21 million Series A investment led by Energize Ventures. You can read about the raise here.

Energy is 10% of global GDP. Up to $850 billion will be invested in U.S. power and digital infrastructure this year alone. That kind of capital deployment demands an information layer that doesn't exist yet — one that can make millions of unstructured documents not just searchable, but genuinely useful. To borrow from my LinkedIn earlier this week: we're connecting millions of unstructured data sources into an infinite recombination of informational possibilities.

From the jump, Halcyon’s goal is to make a vast amount of unstructured energy information accessible using artificial intelligence and smart, directed software engineering. We are building the AI platform for energy, solving information challenges wide in breadth and deep in scope. Fresh capital helps us towards that goal. There’s much more to come.

That ambition requires a phased approach. Three-phase alternating current connects generation, transmission, and distribution. We think in three phases too, similarly connective: Crawl, Walk, Run. Each builds on the last. None of them are small.

[Editor's note: We'll be expanding on all these topics at our "Time to Power" event during San Francisco Climate Week. Join us!]

Crawl: Smart Alerts and Automated Query Delivery

Where we’re at now

The energy regulatory landscape generates a staggering volume of filings, dockets, and documents — across 50 state public utility commissions, every ISO and RTO, and FERC. Keeping up with all of it is a full-time job, and most organizations are still doing it manually or not at all.

Crawl is about solving that problem today. We've rolled out agentic alerts — persistent, AI-powered queries that continuously monitor our data catalog and deliver results when something relevant surfaces.

Think of it this way: instead of a person logging in every morning to run the same search, Halcyon runs it for them (across every document, and over every night) and pushes a synthesized, cited answer to their phone, inbox, or dashboard. A user tracking small modular nuclear reactor costs disclosures, for example, gets notified the instant a new filing touches on candidate project capex, with a full AI-generated summary of what changed and why it matters.

This is not a keyword alert. It is a natural-language query that understands context, disambiguates between a feasibility study and an operational filing, and delivers a briefing. It is an active relationship with dynamic information.

Walk: Totality Queries with Human-in-the-Loop

What's next

Alerts handle the monitoring problem. Observation is a start, but it is not research, and we know that research is the path towards enacting billion-dollar investment decisions. Research in energy means wrestling with sprawling, multi-dimensional questions that span dozens of jurisdictions, hundreds of dockets, and thousands of pages. 

Walk is about better reconciling the way that people are used to asking questions with the way that our system works. Many people like to ask what we call “Totality Queries,” which require a lot of data to answer. By giving our users input into an agentic workflow that constrains search filters and then executing many, many queries against that data, we can provide a more complete response, faster.

Here's an example. A user asks something like: "Find all the in-development nuclear energy production projects in the Northeast and extract the most recently published budgeted costs." Before Halcyon runs that query, it asks clarifying questions: What the threshold for project inclusion should be, which cost components matter, and what metadata to pull alongside the results. The user makes their selections, Halcyon decomposes the request into sub-queries, runs them across nearly 5,000 documents, and returns a structured, exportable table, with every data point traced back to its source filing.

As more data is ingested, the table updates in real-time. As business objectives change, revised parameters and new data fields are easily configurable. We are moving beyond just “energy search engine” and into “energy research system.”

Run: Custom Data Products, Generated Agentically

Where it’s all going

Here's where it gets exciting.

Today, there are myriad structured energy datasets that are effectively hand-loaded. Analysts at research firms, consultancies, and internal strategy teams spend months compiling trackers, benchmarks, and market maps, pulling from regulatory filings, news, corporate disclosures, and expert interviews. The output is valuable but static; it is slow to update, and expensive to produce.

Run is about agentic data subscriptions: custom, continuously updated intelligence products that Halcyon builds for the user, with the user, on demand.

Imagine asking Halcyon:

"Build a benchmarking tracker for micro nuclear development in the Northeast United States. Include cost modeling, licensing and regulatory costs, public comment sentiment analysis, and project status."

Halcyon doesn't just search; it architects. It asks clarifying questions, validates source parameters, breaks the request into sub-queries, and then assembles a living dashboard: structured tables, interactive maps, technology breakdowns by plant type, and project-level detail — all continuously refreshed as new filings land in the system.

This is the "system of action" we reference in our announcement. It is not a tool that you query but rather a platform that builds and maintains the analytical infrastructure energy professionals need to allocate capital, evaluate risk, and proceed with confidence.

What the Series A Means for All of This

The $21 million lets us accelerate across all three phases simultaneously. Crawl features are shipping now. Walk capabilities are in active development. And Run is the architectural north star that informs every decision we make about our data pipeline, our AI stack, and how we hire.

Want to hear about it in person? Join us for our “Time to Power” event during San Francisco Climate Week. If you’re reading this, we’d love to see you there.