Hi {{first_name}},
Codesmith's revenue fell from $18.5 million in 2022 to $5 million in 2025. Anyone pricing the business off those numbers would have marked it down hard. Once AI started doing the work junior developers used to do, the things a coding academy sells, like seats in a cohort and placements into entry-level jobs, lost some of their value.
I wasn't a buyer though. I've been an investor in Codesmith since 2015, and Build Wealth's SPV, BuildForce I, is investing in its current round. So I watched the revenue evaporate from the inside, and I didn't like it. But I stayed optimistic anyway.
For a while, that came down to my belief in the founder, Will Sentance, and what he'd built. He had spent years developing a world-class curriculum and a network of graduates now working inside the agencies and companies that need them. The curriculum improved every quarter, and the network grew one employer at a time. Codesmith had built a raving fanbase by changing people’s lives. None of that shows up on financial statements, and all of it was still fully intact.
In January 2025, with AI tearing through the coding academy industry, Codesmith launched an AI intensive. This year, Forbes Advisor named its Software Engineering + AI/ML Immersive the IT bootcamp with the #1 best outcomes, citing a $110,000 median graduate salary. The proof was showing up again.
That stretch reminded me of something I learned in my 20 years buying businesses. The assets that drive the value are often completely missing from the balance sheet.
Last week was about the Deployment Constraint, the gap between AI spending and AI working inside a business. This week is about who gets paid for closing it, and why it’s historically the most valuable solution. This week:
Nvidia is co-signing up to $105 billion of OpenAI's rent on an Ohio data center, but that’s not the layer that keeps the value long-term.
Two buildouts, telecom and enterprise IT, where the premium held, and one firm that failed anyway.
The Conversion Screen, five questions, two of which a deal must pass, to find out which layer a deal sits on.
— Walker Deibel
WSJ & USA Today Bestselling Author of Buy Then Build
Founder, Build Wealth
P.S. My first check into Codesmith cleared in 2015. Now Build Wealth is joining its current round through BuildForce I, a small investment to help the company scale to meet demand.
If you're curious about joining us, check out the full deal room here.
Please see disclaimers at the bottom of this email and in the presentation. This email is for educational purposes only, shares personal investing decisions that I've made and why, and includes forward-looking statements. Potential investors should always conduct their own diligence and read the full PPM.
SHIFT YOUR STACK
From the Constraint to the Premium
Last week’s issue called the gap between AI spending and AI working inside a business the Deployment Constraint. This week is about learning who gets paid to close it, and why that payment tends to last.
One deal this summer shows how AI capacity is getting funded.
In July, Bloomberg and the Wall Street Journal reported that Nvidia was in talks to guarantee up to $250 billion so OpenAI could lease computing capacity at a 10-gigawatt campus in Ohio. By August, Nvidia had signed guarantees on that campus, and its latest SEC filing caps them at $105 billion. They now back roughly 4.25 gigawatts under 20-year leases to OpenAI, and the exposure shrinks as OpenAI pays its rent. The same filing indicates each generation of Nvidia hardware at the site could mean about 1.5 million GPUs, or $150 to $200 billion in Nvidia revenue.
Nvidia says that many AI clouds and model makers are growing faster than their balance sheets and credit can support. So Nvidia lends its credit to the customer, in this case OpenAI, and the customer's site houses the chipmaker's chips. In fact, the filing says the site will exclusively host Nvidia infrastructure, with only limited exceptions.
All of it is disclosed. It still means part of the demand behind Nvidia's future sales, like the $150 billion to $200 billion per hardware generation the filing ties to this site, rests on credit Nvidia itself extends.
Capacity earns a premium while it is scarce
As a buildout matures, anyone with financing can get the same capacity, and the premium fades. Nvidia's guarantee gets the data center built and the chips plugged in. A room full of chips doesn't earn anything on its own, though. Someone has to put that capacity to work inside a hospital, an agency, or a company that keeps paying for the results.
That work is valuable. The payment for it is called the Conversion Premium, and it goes to the firms that solve last week's Deployment Constraint. The pattern shows up in every infrastructure boom, and it points to where the money ends up.
The balance sheet has its limits
If you've ever looked at a business for sale, you know the balance sheet is the easy part. Equipment, inventory, buildings and cash are simple to price, because anyone can buy the same things. That's also why they rarely explain what a business is worth.
The value usually sits in things that never make it onto paper:
Customer relationships built over a decade
A process refined through years of trial and error
Know-how that lives in a few people's heads
Nobody can buy any of that. It has to be built inside the business, which makes it hard to price and hard for a competitor to copy.
To tell whether those assets are real, follow the cash. Intangibles that already bring in steady revenue are worth paying for. The ones that haven't proven they can earn yet deserve a discount.
Codesmith is a good example. Its revenue fell by more than two-thirds in three years. Those numbers left out the curriculum it had developed over nearly a decade and the network of graduates now working inside the agencies that need them. No buyer could have priced either one from the statements alone.
Those assets are now turning into signed work. Codesmith was selected as a subcontractor under a U.S. Treasury purchase agreement with a ceiling of $118 million, meaning the most it could be worth. New York City also recently signed a seven-figure engagement with the firm.
This is the thesis underneath Buy Then Build. An acquirer who prices only the balance sheet is paying for the part of the business any competitor can buy, and missing the part that earns the premium.
Where the premium sits in a buildout
Every infrastructure buildout has two layers.
Capacity is tangible. It’s the fiber, software licenses, servers, and now GPUs. Conversion is intangible. It’s the work of getting that capacity to function inside one particular agency, hospital, or enterprise.
Conversion gets built and proven inside that institution's walls, and then it starts from zero at the next site. A competitor with more capital has no way to buy in, and a cheaper model release leaves it untouched.
When conversion firms fail anyway
You may be thinking of the consulting and services firms that collapsed in the dot-com bust. They did implementation work too, and it didn't save them. Many of their clients were startups running on investor money, and when that money dried up, the contracts disappeared with it.
Conversion work holds its value only when the client does. Governments, hospitals, utilities, and established companies keep paying after a boom ends, because they still need the work done. Clients funded by speculative capital can stop paying when the funding stops.
The Ohio campus deal raises the same question. Nvidia's guarantee ends once OpenAI earns a satisfactory credit rating. Until then, the rent leans partly on Nvidia's balance sheet. That doesn't mean the leases will fail, but it does mean anyone building a business in and around that campus should know where the money behind it comes from.
Next time the size of a deal impresses you, ask which layer it sits on and who's paying for it.
BEHIND THE NUMBERS
Who Kept Growing After the Buildout
Most big technology buildouts follow a similar arc. Money floods into new equipment, and the companies selling it do well until the spending stops. The firms that put that equipment to work for steady clients often keep growing. Telecom and enterprise IT both show the pattern, and one web consultancy shows its limits.
The counterexample is marchFIRST
Web consultancy, bankrupt 2001
marchFIRST did implementation work too, and it still went bankrupt during the dot-com bust. Doing the work of making technology useful didn't protect it.
Go deeper on implementation
This week's report, The Value Was Never in the Infrastructure Alone, follows the pattern through telecom and enterprise IT, including where it can break down. One example is Equinix. Revenue grew 22-fold between 2000 and 2006, and it still came close to losing its Nasdaq listing in 2002.
Check out this week’s report.
THE PLAYBOOK
The Conversion Screen
Many AI companies now pitch themselves as "deployment" or "implementation" businesses, because that's where investor interest is heading. These five questions help you check whether the label fits, and each one can be asked in a single phone call. Ask them before you read the projections.
Last week's Constraint Test asked whether a company can deliver on its AI story. The Conversion Screen asks whether what it delivers will hold its value after the buildout slows. It works best on services and AI-adjacent deals that claim to turn capacity into results. Power and land deals need a different test. There's only so much power and land to go around, so owning them can stay valuable on its own.
1. Which layer does the revenue come from?
Ask for the revenue split between selling capacity (chips, space, licenses, seats) and making capacity work inside a client (implementation, integration, operation, training).
Why it matters: Many companies do both, and the split tells you what the valuation is paying for.
Look closer if a pitch calls itself "AI infrastructure" or "deployment" and can't produce the split.
2. Who is the end client, and where does their money come from?
Ask for the top five clients and how each one funds its spending.
Why it matters: Governments, utilities, hospital systems, and established enterprises keep paying after a boom ends.
Walk away if the client list leans on companies whose own funding is still an open question.
3. Is anyone financing the demand?
Ask whether the vendor, an investor, or an affiliate is funding, guaranteeing, or discounting for the customer. Then ask what happens to demand if that support goes away.
Why it matters: Nvidia's latest SEC filing shows what this looks like, with guarantees capped at $105 billion behind OpenAI's leases at the Ohio campus. A sponsor who has run the number will have the answer right away.
Look closer if there's no answer, or the answer treats the support as permanent.
4. Where has the conversion already worked, and what happened next?
Ask for the first client's results and whether the work expanded.
Why it matters: A conversion company sells a record of making technology work somewhere specific. Proof is a named institution, a signed scope, and a second purchase.
Walk away if the proof is a pipeline of pilots and letters of intent with no named institution or signed scope.
5. How long would a funded competitor need to copy this?
Ask how many years a rival would need to match the offering. Then ask which client relationship took longest to build and what it required. Names, dates, and people make the answer checkable.
Why it matters: If a better-funded competitor could copy the offering within a year, the company's only real advantage is money, and money can be raised. What takes years to copy is trust with clients, people already working inside client organizations, and a track record built one client at a time.
Look closer if the edge disappears when someone writes a bigger check or a cheaper model ships.
How to Score Your Answers
Mark each answer with one of three grades:
Pass: the sponsor shows evidence.
Partial: a credible claim, with no evidence yet.
Fail: no answer, or a flag applies.
Then read the results:
A fail on question 2 or 4 should end the screening. The premium depends on a durable client and a record of proof.
If 2 and 4 pass and nothing else fails, you're looking at conversion, and the next question is on price.
Any other combination is a judgment call. The partials show you where to spend your diligence.
A big number tells you how much capacity is being built. These five answers tell you who gets paid once it's running. Run the screen on the next deal in your inbox, and you'll know which layer your money is buying.
A Note On Our Deal
Codesmith is the company closest to this week's thesis, so you should know where I stand. I've been an investor since 2015, and Build Wealth's SPV, BuildForce I, is investing in its current round.
Run the Conversion Screen on it the same way you'd run it on any other deal. The full deal drop is in the portal.
Please see disclaimers at the bottom of this email and in the presentation. This email is for educational purposes only, shares personal investing decisions that I've made and why, and includes forward-looking statements. Potential investors should always conduct their own diligence and read the full PPM.
WEALTH STACK REBELLION
"Knowing is not enough; we must apply." — Johann Wolfgang von Goethe, Wilhelm Meister's Journeyman Years.
The chips, the campuses, and the credit are infrastructure. The lasting value sits in the applying. That gets earned inside one agency, one hospital, one company at a time. The firms that earned it for clients who stayed kept growing after the buildouts broke.
The next time a headline gives you a number, go find the people doing the applying.
Did this land in your inbox via a friend or colleague? Subscribe to get your own weekly insight into private markets.
Already a reader? Help other investors open their eyes to the private markets and forward this to someone who would value the insight. Or hit reply to tell us what you think.
By continuing to receive the Wealth Stack newsletter, you acknowledge and accept our Privacy Policy, which includes the sharing of subscriber engagement data with select advertising partners.
To unsubscribe, click here.
This is not financial advice. Illustrative output of a reasoned thought experiment. Not a backtest, guarantee, or prospectus. Actual results vary based on market conditions, fund selection, timing, fees, taxes, and factors not modeled. Private credit, CRE, and leveraged strategies involve significant risk including loss of principal. Consult a qualified financial advisor.
1
