Data-native capital is coming

1. Key Themes

Theme: Corporate data is a hidden, unaccounted-for asset class

Data has been invisible on balance sheets

Rechtman argues that accounting conventions have systematically hidden the value of data: expensed at creation, ignored during operations, and absorbed into goodwill at acquisition.

"It's expensed when created (doing the normal work of running a business), ignored on continuous operations (as it's accumulated), and buried in goodwill (when acquired). Data has always just kind of faded away on balance sheets and in M&A pricing."

Model training turned theoretical value into practical value

The unlock is that data now has a clear, monetizable use case, which gives it real economic worth even if individual companies' data varies widely in value.

"We now know what it's all always been for (training models!), which means it's now practically, not just theoretically, valuable. The value of the data of any one individual company is highly varied, but in aggregate there is a huge amount of economic power locked away inside businesses/not accounted for on their balance sheets."

Data should be treated as a capital good

"It's newly monetizable as an asset (and potentially as an ongoing income stream), which means it needs to be thought of more like a capital good. That is, data is or could be more like equipment or a treasury asset than brand."

Theme: Financialization of data will follow the historical pattern of accounting-driven financial engineering

Market-clearing prices will feed through to valuations

As data sales establish prices, those prices will flow into book values and asset prices, creating both near-term arbitrage and long-term structural opportunities.

"First, in being sold, we're starting to establish market-clearing prices for data and its byproducts in various forms. Eventually, that will filter into book values/valuations and asset prices, which has a ton of near term arbs and long term opportunities for new capital structures that can take advantage."

Historical analogies: junk bonds and securitization

"There are lessons from history about accounting driving financial engineering and capital structure. This might be like the invention of junk bonds, which spurred LBOs. Or securitization, which transmuted piles of illiquid, one-off assets into liquid, legible assets. We are at the start of a sea change in technofinance."

Capital markets haven't caught on yet

"Capital markets haven't yet woken up to the opportunity to financialize the data sitting unaccounted for on balance sheets."

Emerging financial products are already visible

"Cash advances against data sales through an emerging field of private credit vehicles... Hard assets that generate data which you can sell to invest in more hard assets... PE firms pre-seling l a target's data before close to cover part of the price... Data contracts big enough to replace a funding round... Prediction markets and staked tokens for decentralized QA or cleaning"

Theme: Data-native business models and operating leverage

Data exhaust as a new revenue stream akin to advertising

The thesis extends beyond financial leverage to operating leverage: data becomes a driver of business models, not just a source of capital.

"it's not just financial leverage (data-native capital structure) BUT ALSO operating leverage (business/earnings transformation). Data exhaust could be a new source of capital AND a new driver of business models, akin to advertising."

Companies run as long-horizon RL environments

"You can even imagine companies being run as long horizon RL environments, continuously generating data/tasks/evals through their regular operations, creating a secondary revenue stream while they improve themselves through the AI they're training/building."

Definition of "data-native" companies

"data-native companies: business built around data as a valuable asset and fully incorporating that belief into their operating assumptions, like internet- or AI-native companies before them."

Theme: Data allocation as a new CEO capital-allocation decision

Keep internal vs. expose externally

"It's a new dimension for capital allocation: data allocation. Making the choice between keeping data internal vs. exposing it externally (and how) is akin to the choice between returning capital to shareholders and letting it compound (and how)." "The best CEOs will always be brilliantly creative allocators."

2. Contrarian Perspectives

Data may be a durable, growing market, not just a bubble

Against the obvious skepticism, Rechtman argues data selling could persist and expand in both volume and sophistication, with a broadening buyer base beyond frontier labs. Supporting evidence cited: the definition of data selling is broad and includes tasks, human data collection, and environments, and buyers will diversify.

"there's some good arguments that data selling (broadly defined as selling raw data, selling tasks, collecting human data, building environments, etc.) for model training will be around forever at increasing volumes AND levels of sophistication selling to more buyers (enterprises training their own models outside the frontier labs, etc)."

Even if the mania ends, the window is open now

He concedes the bear case (budgets, non-rivalry, price competition, synthetic data) but argues that opportunity exists regardless of long-term outcomes.

"The bear case is obvious: the budgets can't last, data is non rivalrous, prices will get competed down as more sellers wake up. Maybe it'll be synthetic data and simulation. All of which might be right. But whether or not the current mania continues, data is this is definitely valuable RIGHT NOW. There is undoubtedly a moment open for those who can seize it."

Data as the next junk-bond-scale financial innovation

A bold claim: data financialization could spawn an LBO-like wave.

"If I'm right, this is the new junk bonds that will spur the new LBO."

3. Companies Identified

CAC

Description: An emerging player in the data-financing space (the article links to it without elaboration).
Why mentioned: Cited as one of the most important early versions of data-native financial products.
Quote: "CAC is among the most important early versions of this and I'm starting to see a lot of exciting ideas"

Slow Ventures

Description: Venture capital firm where the author is a partner.
Why mentioned: The author's firm; he leads pre-seed/seed rounds from a ≈$325M fund.
Quote: "I'm a partner at Slow Ventures, where I lead pre/seed rounds from a ≈$325M fund."

4. People Identified

Yoni Rechtman

Description: Author of 99d; partner at Slow Ventures.
Why mentioned: Article author, articulating the data-native capital thesis.
Quote: "I'm a generalist investor looking for weird takes on important stories: N-of-1 companies taking non-obvious approaches to markets that matter."

Leeor Mushin

Description: Co-host of an upcoming NY dinner for people in the data/model-training supply chain.
Why mentioned: Co-organizer of the event for researchers, data leaders, and quants.
Quote: "Next month Leeor Mushin and I are hosting a dinner in NY for people building in the supply chain (researchers, data leaders, quants, etc.)."

5. Operating Insights

Treat data creation as a potential product, not just exhaust

Design operations so that normal work generates sellable data, tasks, or evals.

"You can even imagine companies being run as long horizon RL environments, continuously generating data/tasks/evals through their regular operations, creating a secondary revenue stream while they improve themselves through the AI they're training/building."

Make deliberate data-allocation decisions

Treat the internal-vs-external data decision with the same rigor as the dividend-vs-reinvest decision.

"Making the choice between keeping data internal vs. exposing it externally (and how) is akin to the choice between returning capital to shareholders and letting it compound (and how)."

Use data sales as financing

Founders and acquirers can use data contracts and pre-sales as non-dilutive or deal-structuring capital.

"Data contracts big enough to replace a funding round" and "PE firms pre-seling l a target's data before close to cover part of the price"

6. Overlooked Insights

Hard assets as data-generating, self-funding flywheels

A brief mention that physical-asset businesses could sell the data they generate to fund further asset purchases, suggesting a new model for capex-heavy industries.

"Hard assets that generate data which you can sell to invest in more hard assets"

Decentralized QA/cleaning via prediction markets and staked tokens

A passing idea that data quality work, often a bottleneck, could be crowdsourced with market-based incentive mechanisms.

"Prediction markets and staked tokens for decentralized QA or cleaning"