The global market for what gets loosely called "market intelligence" is worth roughly $70 billion a year. That number comes from S&P Global's own investor materials — and even they, one of the recognized leaders, capture only about six percent of the total. Bloomberg generates an estimated $10 billion annually, mostly through its Terminal. The rest is split among dozens of brands, platforms, and niche providers.
So the first question anyone asks us is obvious: why enter a segment that looks completely occupied by global blue chips? The second question follows immediately: isn't the whole thing about to be replaced by AI anyway?
My answer to both is the same. We don't compete with those firms, and we're not threatened by AI. Delphi occupies a different position in the value chain — one that barely exists as a category today, but that everything downstream depends on.
The intelligence value chain has a gap
Think of the market intelligence process as three layers. At the bottom, raw data: filings, press releases, patent texts, sensor feeds, web signals. In the middle, data systems: Bloomberg, PitchBook, S&P Global Market Intelligence, LexisNexis — the companies that normalize, store, and distribute data. At the top, strategic synthesis: McKinsey, BCG, corporate strategy teams, investment banks — the people who interpret data and make recommendations.
Most people assume this system works end to end. It doesn't.
The handover from raw data to structured data systems has become the main bottleneck. The companies in the middle layer — the Bloombergs and S&P Globals — are expected to do everything: harvest raw data, clean it, structure it, and serve it to end users. That quadruple burden is where the system breaks down.
What the missing layer looks like
We call what we do data aggregation and structuring — or, less formally, the part of the value chain that sits between raw information and finished intelligence products.
We operate four interlocking engines. Signal pipelines ingest raw sources — financial disclosures, transport records, legal filings, web-scraped data, scientific publications — with everything time-stamped and traceable. Entity resolution reconciles the chaos of company names that split, merge, change spelling, and appear differently across jurisdictions into canonical identifiers. An industrial ontology maps the relationships: a membrane supplier feeds an electrolyzer manufacturer, which supplies a hydrogen production project, which signs an offtake agreement with a utility.
Why the incumbents haven't solved this
The obvious objection: why can't Bloomberg or S&P Global just tighten their own pipelines? In theory, they can. But their core strength is the creation and delivery of finished insight — terminals, dashboards, research notes, client service. Re-architecting the raw data plumbing is a different craft entirely.
This is where a company like ours fits. Because our entire organization was designed around the structuring layer, we move faster and lighter.
AI makes this more urgent, not less
Large language models impress everyone with fluent answers, but they are ruthless with bad input. When training data conflicts or arrives without provenance, the model hallucinates. The more organizations lean on automated reasoning for high-stakes decisions, the more they need verified, explainable facts underneath.
The structuring layer is not a victim of AI — it's the fuel that high-stakes AI will run on.
What we've proven so far
We chose the hydrogen economy as our first sector because the data gaps were wide open. In two years, we mapped over 4,000 low-carbon hydrogen projects globally. We've since expanded to 37 sectors across energy, decarbonization, defense, and industrial technology.
Where this goes
We built Delphi because we experienced this friction firsthand. The structuring layer isn't a glossy report or a miracle algorithm. It's entity IDs, source links, and relationship graphs — unglamorous pieces that, combined, let everyone else move faster with more confidence.