Where the Tech Economy Hides

The Information sector is 5.4% of American GDP. In 1997, it was 4.6%. That is a gain of 0.8 percentage points over 27 years — one of the smallest changes in the entire BEA dataset. In a series about structural transformation, the Information sector appears to be the industry where nothing happened.

This is deeply misleading.

The flatness of the headline number conceals one of the most violent rotations in the American economy. Inside that stable 5% shell, a complete changing of the guard took place. The sub-industries that dominated the sector in 1997 — broadcasting networks and telecom carriers — lost nearly a full percentage point of GDP. The sub-industries that barely existed in 1997 — data processing, internet publishing, cloud computing — grew to replace them and then some. The Information sector did not stand still. It churned.

And that is only half the story. The “tech economy” as most people understand it — the companies that build software, run data centers, design computer systems, and power the digital infrastructure of every other industry — does not live inside a single NAICS sector. It spills across the BEA’s classification boundaries into professional services, manufacturing, and industries that the North American Industry Classification System was never designed to capture. To find the real tech economy, you have to look in places the official statistics were not built to show you.

NAICS sector 51, Information, contains four major sub-industries. In 1997, the sector was dominated by broadcasting and telecommunications (NAICS 513), which alone accounted for 2.6% of GDP — more than half the entire sector. These were the cable companies, the phone companies, the broadcast networks. They built the pipes and owned the airwaves.

Publishing, including software (NAICS 511), was the second-largest piece at 1.1% of GDP. This was the era of boxed software — Microsoft Office on CD-ROMs, enterprise licenses sold by the seat, newspapers and book publishers still operating at scale.

Motion pictures and sound recording (NAICS 512) contributed 0.5%. And then there was data processing, hosting, and related services (NAICS 514), at 0.4% of GDP. In 1997, this was a rounding error. It included data centers that processed payroll, early web hosting companies, and the first internet service providers. Google would not be founded for another year. Amazon was selling books. Facebook did not exist.

By 2024, the composition had flipped. Broadcasting and telecom fell from 2.6% to 1.7% — a loss of 0.9 percentage points, driven by cord-cutting, the displacement of legacy voice networks by internet protocol, and the collapse of traditional telecom pricing power. Data processing and internet publishing rose from 0.4% to 1.8% — a gain of 1.4 percentage points. The sub-industry that was an afterthought became the largest contributor to the sector.

The chart tells the story with brutal clarity. The teal band at the bottom — data processing and internet services — barely registers in 1997. By 2024, it is the thickest layer in the stack. The gray band at the top — broadcasting and telecom — was once the dominant force. It has been compressed from above and below, losing share to streaming services that show up in other categories and to internet-native companies that replaced its entire business model.

The total stays remarkably stable. The sector hovered near 5% for the entire 27-year period, never exceeding 5.5% and never falling below 4.6%. If you only looked at the sector-level number, you would conclude that the Information economy barely grew. You would be wrong about the magnitude of change by a factor of twenty.

Shares of GDP measure relative size. They can be misleading when the entire economy is growing. To understand the true scale of what happened inside the Information sector, you need to look at real value added — output measured in constant 2017 dollars, stripped of inflation.

In 1997, NAICS 514 — data processing, hosting, and related services — produced $30.9 billion in real value added. By 2024, that figure was $608.5 billion. That is growth of 1,869%. Not 18%. Not 186%. One thousand, eight hundred and sixty-nine percent.

It is the fastest growth of any sub-industry in the entire BEA dataset over this period. No other line item comes close. To put it in perspective: the entire Information sector grew from $290 billion to $1.7 trillion in real terms — a 487% increase — and more than half of that growth came from a single sub-industry that was 10% of the sector when the measurement began.

But the tech economy does not stop at the border of NAICS 51. Across the classification boundary, in Professional, Scientific, and Technical Services (NAICS 54), there is another sub-industry that tells the same story: computer systems design and related services (NAICS 5415). These are the companies that build custom software, design IT architectures, manage enterprise systems, and provide the technical consulting that powers every other industry’s digital transformation.

In 1997, computer systems design produced $49.0 billion in real value added and accounted for 0.9% of GDP. By 2024, it reached $648.2 billion — growth of 1,222% — and its share had doubled to 1.8% of GDP. This is the second-fastest-growing sub-industry in the BEA data, trailing only data processing itself.

The table reveals the hierarchy of transformation. The two highlighted rows — data processing and computer systems design — are the engines of the digital economy. Together, they grew from $80 billion to $1.26 trillion in real terms. They now account for 3.6% of GDP combined, up from 1.3% in 1997. Neither of these sub-industries existed in recognizable form before the internet age.

Add publishing and software (NAICS 511), which includes the operating systems, productivity suites, and enterprise platforms that underpin the entire digital stack, and you get the “hidden tech economy” — roughly 5.1% of GDP in 2024, up from approximately 2.4% in 1997. That is larger than the entire Information sector. The tech economy has outgrown the box the government built to measure it.

Broadcasting and telecom grew too — 266% in real terms is not nothing — but the economy grew faster. When the overall economy expands by 110% and your industry grows by 266%, you gain share. When data processing grows by 1,869%, you get eclipsed. Telecom went from the dominant Information sub-industry to the second-largest, passed by a category that in 1997 was processing payroll checks and hosting GeoCities pages.

The North American Industry Classification System was designed in the 1990s, adopted in 1997, and last substantially revised in 2017. It was built for an economy of clearly delineated industries: factories made things, banks lent money, phone companies provided dial tone. Each company belonged to one sector, and the sector boundaries made intuitive sense.

The digital economy broke those boundaries. Google is classified under Information (NAICS 51), but its advertising revenue competes with traditional media companies that are also in Information and with print and outdoor advertising companies in other sectors. Amazon’s marketplace is classified under Retail Trade (NAICS 44-45), even though its cloud computing division — Amazon Web Services — is the largest infrastructure-as-a-service provider in the world and belongs conceptually in data processing. Apple designs hardware (Manufacturing, NAICS 31-33), runs a services ecosystem (Information), and operates retail stores (Retail Trade).

The result is that the “tech economy” as investors, workers, and policymakers understand it simply does not correspond to any single line in the BEA accounts. It is scattered across at least four NAICS sectors. The Information sector’s flat 5.4% GDP share is an artifact of this classification, not a reflection of technological stagnation. The real tech economy is hiding in plain sight — growing at extraordinary rates, reshaping every industry it touches, but distributed across categories designed for an economy that no longer exists.

This matters beyond academic taxonomy. Policy debates about “Big Tech” often reference the Information sector’s modest GDP share as evidence that technology is less dominant than it appears. Trade negotiations focus on goods sectors because they have clear NAICS codes and measurable trade flows. Workforce development programs target industries as defined by classification systems that put a software engineer at an automotive company in Manufacturing and a software engineer at Google in Information, even though they may do identical work.

The BEA’s data is not wrong. It is precise, rigorous, and internally consistent. But the structure it imposes on the economy was designed for a different era. The Information sector’s apparent flatness is the clearest demonstration of that mismatch — and the most dramatic example in this series of how a single number can conceal a revolution.

In the next episode, we turn to an industry group where the numbers are large and unmistakable: finance, insurance, and real estate — the FIRE empire that crossed 20% of GDP in 2001 and never came back.