America’s service economy is expanding, but the growth is concentrating in the systems that make artificial intelligence possible. New Census Bureau data show revenue at data processing, hosting and related services rose 20.3% from a year earlier in the second quarter. Software publishers grew 14.7%. Both far outpaced the 7.6% increase across selected services.
The same data reveal the other side of the shift. Wired telecommunications revenue fell 2.5% from a year earlier, wireless carriers fell 1.1%, and computing infrastructure and data-processing employers cut 8,000 jobs in August. The AI buildout is generating demand, but it is doing so through a capital-heavy model that rewards servers, software, electricity and scale more quickly than payrolls.
For investors, this is no longer just a technology story. It is a reallocation of revenue, power demand and financing across the economy. The opportunity is real. So are the risks of overbuilding, grid constraints and assuming that fast nominal growth will automatically become durable returns.
The big picture
The United States remains a service economy, and its newest growth engine is increasingly physical. Cloud software may feel intangible, but it rests on data centers, power generation, transmission equipment, cooling systems, semiconductors and large pools of capital.
The Census Bureau’s September 9 Quarterly Services Survey put a fresh number on that shift. Selected services generated $6.419 trillion of revenue in the second quarter, up 3.0% from the first quarter and 7.6% from a year earlier. The figures are seasonally adjusted where available but are not adjusted for price changes, so they measure nominal revenue rather than real output.
Inside the total, the digital infrastructure categories were much stronger. Data processing, hosting and related services produced $124.1 billion of second-quarter revenue, up 5.5% in three months and 20.3% over the year. Other information services rose 18.9% year over year, while software publishers increased 14.7%.
This is a different pattern from a broad consumer-services boom. The fastest growth is appearing around the infrastructure and software layers that store, process and monetize information. It creates a powerful investment cycle, but also a narrow one: the gains depend on capacity utilization, electricity availability and the ability to turn expensive computing assets into recurring cash flow.
What happened
The service economy strengthened in the second quarter. Transportation and warehousing revenue rose 12.3% from a year earlier, information revenue increased 9.9%, finance and insurance grew 7.6%, and health care and social assistance advanced 7.4%. The breadth matters because it shows the economy was not running on data centers alone.
But the dispersion inside sectors was striking. Within information, software publishers reached $177.0 billion in quarterly revenue, while data processing and hosting reached $124.1 billion. Traditional connectivity businesses moved in the opposite direction: wired telecommunications revenue declined 2.5% year over year and wireless carriers declined 1.1%.
The labor data add a warning against equating revenue growth with hiring. Total U.S. payrolls rose by 162,000 in August, yet the information industry lost 23,000 jobs. Computing infrastructure providers, data processing, web hosting and related services accounted for 8,000 of those losses, after a period in which the industry was posting rapid revenue growth.
That is not proof that AI caused the job cuts, and the revenue and employment reports cover different periods and statistical concepts. It does show that a high-growth industry can expand sales without expanding payrolls at the same pace. Scale economies, automation, pricing and capacity coming online can all lift revenue per worker.
Why it matters
The AI investment case rests on a chain of assumptions. Computing demand must keep rising. Customers must be willing to pay enough to cover depreciation and energy costs. Power projects must arrive on schedule. Software must produce economic value beyond experimentation. Finally, that value must be distributed widely enough to sustain demand.
The latest official data support the first part of the chain: spending on digital services is growing rapidly. The Energy Information Administration supports the physical side. In its September outlook, EIA said U.S. electricity sales are expected to reach 4,135 billion kilowatthours in 2026, almost 2% above 2025, with record consumption driven by data-center development and increased manufacturing activity.
That power demand changes the investable map. The beneficiaries extend from cloud and software vendors to utilities, natural-gas infrastructure, renewable generation, grid equipment, cooling technology and construction. The bottlenecks also widen. A delayed interconnection, transformer shortage or local regulatory constraint can impair the economics of a data-center project even when demand for computing remains strong.
Federal Reserve Governor Christopher Waller described the buildout as a continuing source of rapid business investment, while noting that it is concentrated in a capital-intensive sector. That distinction matters. Capital spending adds to GDP and can lift future productivity, but it does not guarantee broad job creation today. It can also keep demand for financing and scarce equipment elevated before productivity benefits reduce costs elsewhere.
Chain reaction
The first link is computing demand. Businesses buy model access, storage, security, databases and software services. Revenue rises fastest at platforms able to bundle those capabilities and spread fixed costs across many customers.
The second link is capacity. Providers order servers, networking equipment and cooling systems, lease land and sign long-term power agreements. The spending flows into manufacturing, construction, utilities and commodity supply chains.
The third link is electricity. Data centers require reliable round-the-clock power, forcing utilities and grid operators to accelerate generation and transmission plans. Regions with available power, land and permitting capacity attract projects; constrained regions face higher costs or delayed connections.
The fourth link is financing. Large projects require deep balance sheets or outside capital. Higher rates raise the hurdle for speculative capacity, while established operators can use scale and access to funding as a competitive advantage.
The final link is utilization. Infrastructure only earns attractive returns if customer demand fills the capacity at healthy prices. If computing efficiency improves faster than use expands, or if too many projects arrive together, revenue growth can remain strong while returns on new capital disappoint.
Market implications
Equities
Revenue momentum favors data-center operators, software publishers and suppliers of scarce infrastructure. Yet the most important distinction is between demand growth and shareholder returns. Investors should track utilization, pricing, depreciation, power costs and free cash flow—not only capital expenditure or contracted megawatts.
Traditional telecommunications offer a useful warning. Owning essential infrastructure does not guarantee revenue growth when competition is intense and capacity becomes commoditized. The same outcome is possible in parts of cloud computing if supply expands faster than differentiated demand.
Utilities and equipment makers may have a longer runway because the grid buildout reaches beyond any single model provider. Their risks are regulatory lag, construction inflation and the possibility that projected loads arrive later than expected.
Treasuries
The buildout can pull rates in opposite directions. Heavy investment and electricity demand support capital spending and may keep the economy’s neutral rate higher. Successful deployment can later raise productivity and reduce unit costs. For bonds, the timing matters: investment demand arrives before many of the disinflationary benefits.
Credit
Balance-sheet quality is a central divider. Companies funding long-lived capacity with short-term or floating-rate debt are more exposed to delays and price competition. Borrowers with contracted customers, staggered maturities and access to power have greater resilience. Credit investors should stress-test lower utilization and later completion dates rather than assuming every announced project starts on time.
Energy and commodities
EIA expects record electricity consumption, making power availability an economic asset. Natural gas, solar generation, storage, transmission equipment and uranium-related supply chains can all benefit, but not evenly. Local grid rules and connection queues can matter more than national demand forecasts.
The dollar and global markets
The United States’ advantage in cloud infrastructure, capital markets and energy supply can continue attracting global investment. It can also deepen dependence on U.S. platforms and raise the cost of competing for power and equipment abroad. Countries with reliable low-cost electricity and permitting capacity gain strategic value; import-dependent systems face a harder financing tradeoff.
Decision brief
- Follow revenue dispersion. Data processing and hosting at +20.3% year over year is more informative than the 7.6% selected-services headline.
- Separate nominal growth from real output. Census revenue figures are not adjusted for price changes.
- Test the power constraint. A computing project without a credible connection date is not operating capacity.
- Track jobs and revenue together. Rapid sales growth alongside falling employment signals scalability, but also a narrower transmission to household income.
- Demand return evidence. Utilization, pricing and free cash flow should confirm that capital spending is creating value.
Three horizons
Three months
Watch the September 30 BEA annual update, utility load forecasts and company guidance on data-center utilization. The next Quarterly Services Survey arrives in November and will show whether digital-service revenue kept outrunning the broader economy.
One year
The key question is whether capacity constraints ease without destroying pricing. Monitor grid connections, equipment lead times, power contracts and the spread between revenue growth and capital expenditure. A healthy cycle should produce rising utilization and cash returns, not just a larger construction pipeline.
Three years
The durable winners will own either scarce infrastructure or software that customers cannot easily replace. The broader economy wins only if the buildout lifts productivity outside the technology sector. If it remains concentrated, returns may be strong for a few firms while hiring, wages and downstream adoption lag.
Bottom line
The service economy is not simply growing; it is being reorganized around computing infrastructure. Data processing, hosting and software revenue are expanding far faster than the aggregate, while legacy connectivity categories and information-sector employment lag.
For investors, the message is to follow the entire chain from software demand to servers, power, financing and utilization. The AI boom has moved beyond narrative. It is producing measurable revenue and record electricity demand. The next test is whether the physical buildout produces durable cash returns before capacity, costs or constraints catch up.
Sources
- U.S. Census Bureau — Quarterly Selected Services Estimates, Second Quarter 2026
- U.S. Bureau of Labor Statistics — Employment Situation, August 2026
- U.S. Energy Information Administration — September 2026 Short-Term Energy Outlook: Electricity, Coal and Renewables
- Federal Reserve — Governor Christopher Waller on the Economic Outlook
- U.S. Bureau of Economic Analysis — GDP and Corporate Profits, Second Quarter 2026
Research cutoff: September 10, 2026, 10:15 CEST. Census service-revenue figures are nominal and are not adjusted for price changes. Monthly employment and quarterly revenue data cover different periods. This article provides general economic analysis, not individualized investment advice.
