Capitalizing invention.

On April 20, 2026, the Bureau of Economic Analysis published a blog post with a deceptively plain title: When Are Data and Tractors Treated the Same in Economic Accounting? The answer, starting with the 2028 comprehensive update of the National Income and Product Accounts, is: increasingly, both.

BEA is aligning the U.S. national accounts with the 2025 System of National Accounts revision, which for the first time classifies certain data — long-lived, directly producing economic benefits — as a produced fixed asset. The tractor comparison is not a metaphor. It is the point. A manufacturer's sensor data, collected over years, used repeatedly across production decisions, will appear in the national capital stock the way a tractor appears there today. It is the latest move in a thirteen-year accounting migration that has been quietly reshaping the valuation of every company, fund, and individual whose balance sheet rests on intangibles.

I'm writing this as an inventor, not as a tax accountant. Five patent families filed between 2014 and 2018 describe a coherent stack: identity resolution, scarcity-weighted valuation, control surfaces, trigger-event systems, and rule-governed settlement. The work was funded as R&D and filed as intellectual property. What it produced, on the BEA's emerging definitions, is not just IP. It is also the method for producing long-lived data assets — a detail that becomes substantial once the accounting frameworks, the tax code, and the marketplace infrastructure all converge.

Three frameworks, three treatments

When you write code that becomes a patent, three separate accounting systems see that work and they do not agree.

National accounts (BEA / SNA).Since 2013, BEA has capitalized R&D in the NIPAs as an Intellectual Property Product. Gross domestic product added roughly 2.5% overnight when that change landed; nothing about the economy changed that morning except the measurement. Starting 2028, data produced by an organization's own operations — the "own-account" category — joins R&D as a recognized capital good at the national level, closing a long-standing inconsistency in which BEA capitalized the database but excluded the data it held.

Corporate GAAP.GAAP is the outlier. U.S. GAAP (ASC 730) expenses R&D as incurred with narrow exceptions; internally-developed intangibles generally do not appear on the balance sheet at all. IFRS (IAS 38) allows capitalization of development-phase costs when six specific criteria are met — probable future benefits, intent to complete, adequate resources, and so on — but most U.S.-domiciled filers operate under the stricter GAAP regime. A startup that spends five years inventing a patent portfolio shows five years of operating losses and, at the end, a patent grant booked at roughly zero cost basis. The licensing market does not share this view.

Tax (IRC §174).Until 2022, U.S. firms could immediately deduct R&E expenditures. The Tax Cuts and Jobs Act changed this for tax years beginning after December 31, 2021 — domestic R&E had to be capitalized and amortized over five years; foreign R&E over fifteen. That produced an unusual three-year window (2022-2024) in which tax accounting converged with national accounts and, arguably, with economic reality against the lone holdout of GAAP. The One Big Beautiful Bill Act reversed the rule for domestic R&E effective 2025, restoring immediate expensing at home while leaving foreign R&E on the fifteen-year treatment.

Three of the five cells in the R&D column now agree with the economically intuitive position — that invention builds capital — and yet the cell that matters most for corporate enterprise value, GAAP, still dissents. This is where the mechanical valuation problem hides.

Damodaran's bridge

Aswath Damodaran has been writing about this for twenty years. His R&D-capitalization adjustment is straightforward arithmetic. Choose an amortizable life appropriate to the industry (typically five years for software, longer for pharmaceuticals). Build a "research asset" by summing the unamortized portion of past R&D. Add back current-year R&D to operating income; subtract the year's amortization of the research asset. Adjust the tax line for the difference. Restate book equity and invested capital to include the new asset.

The effects are mechanical, which is what makes them persuasive. Earnings stop being systematically understated. Book value rises by the full unamortized research asset. The P/E ratio compresses toward peers. Enterprise value's composition shifts: where GAAP presents a company as mostly future growth options, the restated balance sheet shows a company with an actual recorded asset base.

For intangible-heavy firms the size of this adjustment is substantial. Empirical work published in the Review of Quantitative Finance and Accountingin 2024 estimates that roughly 22% of annual R&D expenditures end up effectively capitalized in samples where capitalization is permitted or measurable, and about 17% of firm-wide depreciation in those samples is attributable to long-lived R&D assets. Studies across IFRS jurisdictions consistently find that capitalized development costs are positively and significantly correlated with market value, and that the capitalization rate is correlated with future patent applications and citations — a leading indicator of innovation output, not just a trailing bookkeeping entry.

What this does to a patent portfolio

Bring it back to an actual portfolio. The five families I filed between 2014 and 2018 cost what they cost to invent. That cost — salaries, prior-art research, prosecution fees, engineering, prototyping — was expensed under GAAP as it was incurred. On any corporate balance sheet, the portfolio exists at its capitalized legal prosecution cost, which is a tiny fraction of the invention cost, which is itself a tiny fraction of the value the market ascribes when the patents are licensed, litigated, or acquired.

This has practical consequences. When you license IP, you are negotiating against a counterparty whose financial view of the asset's cost basis is zero. When you collateralize IP, the lender's willingness to extend credit hinges on a recognized cost basis, not a residual. When you sell IP in a distressed situation, the seller without recognized basis is mechanically disadvantaged.

The 2028 BEA change does something new on top of this. It recognizes a second capital asset that patents like these produce: the data. Family 1's identity-resolution patent describes how to build a graph of equivalences between identifiers. The graph itself — updated continuously, long-lived, directly monetized — is what SNA 2025 calls a produced data asset. Family 2's qualitative index is a database. Family 5's rule sets are a database. Under the coming national-accounts treatment, the outputs of these patents are not just IP. They are also the method by which specific long-lived data assets come to exist. That is an accounting distinction with real valuation consequences, because a patent's licensing market prices what the patent enables, and the frameworks are now aligning to recognize both sides of that.

The missing fourth framework

Accounting recognition is necessary but not sufficient. A research asset shown on a restated balance sheet still has no market. A data asset recognized in the national accounts still has no standardized instrument in which to trade. GAAP recognition, when it eventually comes, will move the book value; it will not by itself create a venue where the book value prices.

That gap is the fourth framework, and it is the one I've spent the last several years working on directly. The public home for the argument is americaninnovationeconomy.com. The specific proposal is an instrument class called Innovation Credits: zero-coupon sovereign instruments denominated in U.S. dollars, backed by verified and insured domestic innovation yield, that foreign importers purchase to access the American market — the same economic function a tariff performs, but producing a tradeable asset on the Treasury's balance sheet instead of a one-way revenue line.

The mechanics borrow directly from the BEA's logic. Every transaction flowing through the system generates an SKU-level, supply-chain-verified, consumer-linked commercial dataset. That dataset is actuarially valued and placed on the Treasury balance sheet — "a new category of sovereign asset analogous to the intellectual property protected by the USPTO," as the framework puts it. The parallel is explicit: the USPTO transforms ideas into roughly $7.8 trillion of IP assets; an Innovation Credits market transforms commercial activity into a sovereign data asset of similar scale. The credits and the dataset collateralize each other in a recursive loop — more credits attract more participation, more participation generates more data, more data increases the collateral base, stronger collateral supports more credit issuance.

The infrastructure already exists. GS1 Application Identifier 8112 is the standard for attaching digital coupon-like values to individual SKUs at point of sale; it is already deployed at CVS and other major U.S. retailers. BNY Mellon / Pershing clears institutional securities at the scale Treasury instruments require. Stablecoin rails provide the dollar-denominated settlement layer described in Patent Family 5. Tokenization for fractional trading is standard infrastructure now. The framework assembles existing components — it does not ask for speculative technology or greenfield rails.

What the convergence looks like from here

Four frameworks converging produces a different kind of valuation than any one alone. National accounts recognize the invention as capital. Tax code has done the same, intermittently, in one direction for domestic R&E and continuously for foreign. Practitioner methodology restates corporate financials so the research asset shows up where GAAP has obscured it. And a marketplace — one that securitizes innovation yield and hardens commercial data into a sovereign asset class — gives those recognized assets a venue in which to price.

The personal version of this story: when I filed the patents, the industry framed them as ad-tech, the accounting framed the invention as expense, and the cost basis recorded on any balance sheet I ever touched was approximately zero. In the decade since, the ad-tech framing has given way to a broader reading of what the portfolio enables. The accounting frameworks are starting to catch up. And the specific piece of infrastructure that lets capitalized innovation trade — rather than just sit on a restated balance sheet — is the work I'm doing now.

Five patents on my side of the ledger. Roughly fifteen years of R&D behind them. Sensor data, databases, rule sets — long-lived, directly beneficial, not yet on anyone's GAAP balance sheet. A tractor and a patent and a database walk into the national accounts, and starting 2028, all three check the same box. What comes next is the market that lets the third one trade.