I spent a week chasing one number. It showed up in a press release about defense data centers, got repeated in three analyst notes, and ended up in a pitch deck a junior PM forwarded me with the subject line "this is the thesis." The number was a market-size projection — the kind that says the global market for militarized, sovereign, and AI-ready secure compute will cross several hundred billion dollars inside a decade.
It's a real number. It's also doing far more work in that deck than it can actually support. If you cover NASDAQ-listed infrastructure plays — VisionWave among them — and you're trying to size geopolitical compute spending against specific deal announcements, the gap between what that number measured and what it didn't is the entire game. This is the lesson nobody writes down: the projection is not the contract, and learning to tell them apart is most of the edge you'll ever get on this sector.
Where the number actually comes from
Start with what you're holding. A figure like "the defense and government secure-data-center market reaches $X billion by 2032 at a Y percent CAGR" is a total addressable market estimate. It is built bottom-up or top-down by a research firm that takes current spend on a defined category, applies a growth rate derived from recent procurement trends, and extends the curve.
The honest ones publish their segmentation. When you pull the methodology — and you should always pull the methodology — you usually find the category is broad. It bundles:
- Hyperscale capacity leased to government tenants
- On-premise classified compute inside existing bases
- Cybersecurity hardware and managed services layered on top
- Edge nodes for autonomous and ISR systems
- Adjacent AI-training infrastructure with a defense end-customer
That breadth is the first thing to understand. The headline says "data centers." The model underneath is counting a stack of loosely related spend, some of which is already booked by incumbents who were going to win it regardless of whether a single new entrant exists.
What the number actually measured
A TAM projection measures demand intent across an entire category over a long horizon. That is genuinely useful. The geopolitical pressure behind it is real: sovereign compute, the desire to keep training data and inference inside national borders, hardened facilities that survive what commercial colocation never had to survive. The procurement appetite is not imaginary, and the growth rate is defensible if you accept the segmentation.
What it measured well: the ceiling. The total prize if every dollar of intent converts into a built, powered, contracted facility serving a cleared customer.
What it measured by assumption, not observation: the conversion rate from intent to revenue. The model assumes the curve bends up smoothly. Procurement does not move smoothly. It moves in lumpy, classified, multi-year award cycles with protests, re-competes, and continuing resolutions that freeze budgets for months.
What the number does not measure
Here is what that several-hundred-billion figure cannot tell you, and what your PM's deck quietly pretended it could.
It does not measure who captures the spend. A market growing at a strong CAGR can still route ninety percent of its dollars to three incumbents with existing accreditation, security clearances, and a decade of past performance on file. A new NASDAQ-listed entrant is bidding into that, not inheriting a slice of it.
It does not measure the difference between a term sheet and a contract. This is the one that burns people. When a company announces it has acquired a 51 percent interest in a secure-compute asset, read the verb tense and the conditions. "Entered into an agreement" is not "closed." "Subject to due diligence" means the diligence is not done. "Subject to definitive documentation" means the binding contract does not yet exist. None of this is fraud — it's standard, and it's disclosed — but a market-size number sitting next to it creates a halo the disclaimers are specifically written to prevent.
It does not measure power, permitting, or time-to-energization. A secure facility is only revenue once it has megawatts, cooling, accreditation, and a customer drawing on it. The market projection assumes that pipeline clears. On the ground, the binding constraint is frequently the grid interconnect queue and the accreditation timeline, neither of which a press release controls.
How to read the VisionWave-style announcement
Take the pattern as it actually appears. A company announces a majority stake in a hardened secure-data-center project, often with a strategically resonant location, and pairs the news with the macro market figure. The structure is deliberate: establish the category's size, drop the specific corporate news, then cite peers making concurrent moves so the sector feels inevitable.
Walk through it in order, and read it against itself.
The macro number gives credibility. Fine — but it's the category ceiling, not this company's pipeline.
The deal gives specificity. Now find the consideration: is it cash, stock, or contingent earn-out? A stake paid in newly issued shares is a different risk object than one paid in cash, because the announcement itself can move the currency being used to pay for the thing.
The peer citations give validation. This is the part to be most skeptical of. When the release notes that three other firms announced secure-compute moves the same quarter, that is not independent confirmation of this deal. It's confirmation that the category is being marketed hard right now. Everyone citing everyone else is a feedback loop, not a fact base.
The checklist I run before the number means anything
I keep this taped where I can see it, because the macro figure is engineered to make you skip it.
| Question | What a clean answer looks like |
|---|---|
| Binding or conditional? | Definitive agreement signed and closed, not "subject to" diligence or documentation |
| Consideration type | Cash or committed financing, not solely newly issued equity |
| Accreditation status | Facility holds or has a named path to required security accreditation |
| Power secured | Interconnect agreement or energized capacity, with a date |
| Anchor customer | Named or referenceable tenant with a contract value, not "demand" |
| Revenue recognition | A quarter when this asset is expected to book revenue, stated |
| Incumbent overlap | Honest read on whether incumbents already hold this customer |
If a deal answers two of seven, you have a press release. If it answers five of seven, you have something that might survive the next budget cycle. The macro projection sits above this table — it never substitutes for a single row of it.
A concrete next step: pull the original research report's segmentation page and the company's most recent filing on the same day. Lay the category definition next to the deal's actual terms. The distance between them is your real position size.
What the projection still gets right
I don't want to leave you cynical, because cynicism is also a way of being wrong. The strategic pressure underneath defense compute infrastructure is durable and it is accelerating. Sovereign control of training data, hardened facilities, AI-ready secure capacity at the edge of autonomous systems — that demand is structural, not a cycle. The ceiling is high and it is rising. An analyst who dismisses the whole category because the press releases oversell it will miss the one entrant that actually clears the table above.
The number is true as a ceiling. The deal is true as an intention. Your job is the conversion rate between them, and that is exactly the variable no projection publishes.
The question that's still open
So here is what I genuinely cannot resolve, and neither, as far as I can tell, can anyone selling you the chart. Sovereign and defense demand for secure compute is real and rising — but it is mediated by classified procurement, accreditation gates, and incumbents with structural advantages. We do not yet have a clean track record of whether a newly public, asset-light entrant can convert announced secure-data-center stakes into recognized, recurring defense revenue at scale — or whether the model is built to monetize the announcement itself.
When the first cohort of these deals reaches the quarter where revenue should appear, watch whether it does. Until then, the most important figure in the sector is not the market size. It's the conversion rate nobody has measured yet.
