CEO Column | C1 | It’s Not PR, It’s Verifiable Narrative: Investors Only Buy What’s “Traceable”
Stories expire; only the chain of evidence remains
At two in the morning, I sat in a hotel room in Brno, Czech Republic. On the desk was black coffee I hadn’t finished, and beside it a pre-investment memorandum (IM) from a cross-border fund. Page by page I fed our compute-center data, utilization rate, depreciation model and carbon-credit revenue estimates into their AI diligence agent. Ten seconds later it replied with a line as cold as a Prague winter: “Narrative detected. Insufficient verifiable milestones.” At that moment I understood: PR is not an asset; being verifiable is. A story can only attract people, but a chain of evidence is what makes investors sign.
01 | Investors in the AI Era No Longer Believe in “Speaking Beautifully”
In the past you could make videos, issue press releases, talk vision and tell brand stories, and investors would feel you had drive, mission and scope. But not now, because the first gate now is not a person but an *AI first-pass screen*.
The AI first-pass screen asks you three questions:
1. Is your narrative verifiable? (Verifiable)
“You say you’re building a compute center → what is your PUE?” “You say you have technology → where is your power-consumption curve?” “You say you have a moat → where is the evidence of your data volume and process complexity?” Without evidence, the agent will directly judge: “Unsupported claim.” No matter how wide your network, it cannot push this through.
2. Are your assumptions traceable? (Traceable)
Where does the revenue forecast come from? Why are costs falling? Are depreciation, amortization and expensed items reasonable? Can the data sources be checked? An AI agent’s job is not to believe you; it is to break you down until you can’t be broken down further.
3. Can your milestones be checked? (Checkable)
You say go-live in Q3, and the agent will check your engineering schedule, the actual manpower committed, supplier delivery dates, server-room completion dates and the timeline of government-document reviews. If they don’t match, it will write it straight into the report: “Timeline inconsistency detected.” This is the IR (Investor Relations) of the AI era.
02 | A Story Is Not an Asset; a “Verifiable Narrative” Is
What is a verifiable narrative? Not one that is beautifully written, touching, deep or philosophical, but one with three things:
1. The narrative has assumptions (Assumptions)
* We expect to go live in 12 months
→ Assumptions: 90 days for equipment delivery, 120 days to build the server room, 45 days for compliance review
* We estimate 85% utilization
→ Assumptions: 11 customers in the current pipeline, negotiation cycle of 4 weeks
Narrative = assumptions × methodology.
2. Assumptions have evidence (Evidence)
Supplier quotes, contract terms, equipment specs, measured data, comparable industry cases, status of official documents, customer LOIs, risk items and countermeasures. The longer the chain of evidence, the more at ease investors feel.
3. Evidence can be traced (Traceability)
Your forecasts cannot skip steps; every number must trace back to one document, one record, one source, one person, and AI does not look at “what you wrote” but at “how it came about.”
03 | A Real Case: Why Have PSF’s Projects Passed Cleanly in Recent Years?
Not because we speak bigger or more beautifully, but because all of our narratives are “verifiable.”
For example:
The Shiga, Japan compute center
* Depreciation model for 256 H200s
* PUE estimates and energy-price sensitivity
* Draft green-power supply contract
* Local government MOU
* Construction drawings, building timeline
* Customer pre-filled capacity (pre-allocation)
All checkable, calculable and verifiable.
After reading, the AI agent generates: **“High traceability. Medium risk. Clear milestones.”**
The Brno, Czech Republic project
We didn’t say “build a compute center”; what we gave was:
* Mapping to EU subsidy mechanisms
* GPU price cycles
* A green power + heat-recovery model
* A local talent-development pipeline
* A five-year financial model (full chain)
* Back-calculated risk and stop-loss mechanisms
Investors don’t need to like me; they only need to believe that this model “won’t go off the rails.”
04 | Why Have Investors Changed? Because Their First Firewall Is AI
In the past investors relied on experience to judge whether the other side’s talk was too dreamy, whether this person looked reliable, whether the industry was a good story. Now it has all changed. An AI Agent’s job is: catch errors, check numbers, timelines, inconsistencies, data sources and the reasonableness of assumptions, and it will not change its standards because of who you are.
05 | Conclusion: When AI Becomes the First-Layer Investor, You Must Speak a Different Language
What you need to learn is not “how to tell a moving story” but how to leave AI unable to refute you, because stories expire, emotions ebb, traffic evaporates and trends end. But the chain of evidence does not (nor do assumption logic, milestones or traceability). In the future, those who get capital will not be those who speak most movingly, but those who can get AI to say: **“Verifiable. Traceable. Investable.”**
