Invictus here. The U.S. economy is expanding, but the sources of that growth look narrower than they did during the late 1990s. That distinction matters: if a larger share of output growth depends on one investment cycle, a change in that cycle can have broader consequences.
Real GDP grew at a 3.4% seasonally adjusted annual rate in the period described here. Yet the average growth rate over the latest eight quarters through September 2026 was 2.34%, well below the 4.17% average recorded from 1995 through 1999. The headline numbers show an economy still moving ahead. The composition of growth raises a harder question about how much support comes from the rest of the economy.
Technology investment carries more weight
Information-processing equipment and software made a larger contribution to growth in the recent period than during the late-1990s expansion, despite slower overall GDP growth. On average, those categories added 0.81 percentage points to annualized quarterly growth over the latest eight quarters, compared with 0.68 percentage points from 1995 to 1999.
Relative to average total growth, that contribution rose from about 16% in 1995–1999 to 35% from 2024 Q4 through 2026 Q3. These ratios are descriptive comparisons, not precise measures of cumulative GDP growth. And they don’t mean artificial intelligence itself accounts for 35% of growth. The Bureau of Economic Analysis categories cover a wider range of computing equipment and software, though they do capture important parts of the investment in computing infrastructure.
The comparison draws on BEA Table 1.5.2, which reports contributions to changes in real GDP, and Table 1.5.5, which provides expanded GDP detail. The source data cover Q1 1990 through Q3 2026. The figures point to a different balance of growth engines: technology investment is contributing more, while the economy as a whole is growing at a slower pace.
The rest of the economy adds less
A simple subtraction helps show the scale of the shift. Remove the contribution from information-processing equipment and software, and the remaining components added about 3.49 percentage points to annualized growth in 1995–1999. In the latest eight quarters, the comparable contribution was 1.53 percentage points, roughly 56% lower.
That calculation doesn’t show that the economy outside technology is shrinking. It shows that its contribution to aggregate growth is considerably smaller than it was during the late-1990s expansion. Other components reinforce the difference. Residential investment contributed positively in the earlier period and negatively in the recent one. Consumer spending remains a major source of support, but its contribution is also smaller in this comparison.
Housing adds context to the residential investment figures. Mortgage rates are comparable to late-1990s levels, but home prices have risen much faster than household incomes, making ownership less affordable. Many existing homeowners also have little incentive to give up the unusually low mortgage rates they secured in earlier years. That has constrained both the supply of homes for sale and transaction activity. Nonresidential structures show a similar reversal in their contribution to growth.
My read is that the expansion has fewer strong supports than it did in the late 1990s. That doesn’t make a downturn inevitable, and contribution data alone can’t tell us whether a narrower growth mix will become systemic fragility. Households, banks, and companies may have financial capacity to absorb a setback; consumer spending has remained resilient. Still, a weaker contribution from several areas leaves technology investment carrying more of the load.
The investment cycle is the key uncertainty
Companies are spending on computing capacity, data centers, networking equipment, and software in anticipation of future demand and productivity gains. Those outlays add to GDP as they happen. Whether they ultimately produce financial returns that justify the spending is a separate matter.
If businesses decide expected returns fall short, they could reduce investment. That would subtract directly from growth and could also ripple through construction, suppliers, hiring, and consumer demand. The effects could extend beyond technology companies because investment spending supports activity across connected industries.
There are reasons to avoid treating this exposure as proof of an impending collapse. The figures don’t establish that the AI investment cycle is irrational, or that it will fail to deliver. Transformative technologies can require substantial capital spending before their broader gains appear. And the BEA categories include investment beyond AI alone.
There’s also a limitation to the subtraction exercise: some technology spending may have displaced investment that would otherwise have gone to other sectors. The data can show how reported contributions compare; they can’t reveal exactly what would have happened in a world without the technology boom.
History offers a useful, imperfect comparison
The late 1990s were not a model of steady growth without risk. That expansion ended in the dot-com bust and was followed by a recession in 2001. The early part of the decade had its own recession and sluggish recovery. The comparison here focuses specifically on the strong 1995–1999 period, rather than treating the whole decade as one uniform run.
Even with that qualification, the contrast is informative. During the late-1990s technology boom, the rest of the economy contributed more to growth than it has in the recent eight-quarter period. Today’s expansion is slower and more concentrated in information-processing equipment and software. That concentration creates sensitivity to whether companies keep investing at the current pace.
The figures are based on U.S. Bureau of Economic Analysis National Income and Product Accounts, Tables 1.5.2 and 1.5.5. The BEA data were revised October 29, 2026; the comparisons use averages of quarterly annualized GDP growth contributions. Information-processing equipment is line 31 and software is line 38. The reported percentages are ratios between contributions and total growth, not exact cumulative shares.
The central issue is not whether growth is positive. It is whether growth can broaden enough to rely on more than one major investment cycle. If technology spending holds up, it remains a meaningful source of expansion. If it falters, the smaller contributions from other parts of the economy could make the slowdown more noticeable.
The post An Overly AI-Dependant Economy? appeared first on The Big Picture.