a16z Infographic on Trends: AI Costs Halved and Usage Doubled This Year; Major Life Milestones for U.S. 30-Year-Olds Are Being Delayed Across the Board

Author: a16z New Media

Translation & Editing: TechFlow

Original Link: https://www.a16z.news/p/charts-of-the-week-dexit-real-or

TechFlow Commentary: This week’s Charts of the Week from a16z covers four topics—each substantial enough to warrant its own dedicated article: the Jevons effect triggered by falling AI costs; the true scale of Big Tech’s capital expenditures; Kalshi prediction markets outperforming professional forecasting institutions; and the broad-based delay in major life milestones for Americans aged 30. The data sources are robust, the analysis calm and measured—making this a high-quality reference for understanding the intersection of current technology and macroeconomic trends.

DExit… Real Trend or Illusion?

Delaware remains the top choice for U.S. corporate incorporation—but that dominance is quietly eroding:

According to Ramp data, Delaware’s share of new company registrations has declined steadily since 2023, dropping roughly 10% in Q3 2025.

History doesn’t repeat itself—but it often rhymes… maybe.

Delaware wasn’t always the corporate incorporation capital.

About a century ago, Delaware displaced New Jersey—the original “Mother of Trusts”—as the preferred jurisdiction for incorporation. New Jersey lost its edge because Governor Woodrow Wilson sought to rein in “corporate abuses,” significantly worsening its business climate. Delaware’s corporate law had already been modeled on pre-Wilson-era New Jersey statutes, so it was eager to welcome fleeing corporations. Over nearly a century, Delaware—and especially its Court of Chancery—built a reputation as a mature, impartial forum for resolving disputes between corporations and investors.

Yet something built over a century can begin to waver in just a few years. Rightly or wrongly, the Court of Chancery has recently adopted a more permissive stance toward shareholder litigation (especially in several high-profile cases—including, but not limited to, Tesla), prompting real relocations of corporate headquarters. Goodnight, and good luck, Delaware.

That’s at least the mainstream narrative—but other data tell a more complicated story.

First, even Delaware’s foundational myth isn’t entirely accurate.

It wasn’t until the 1980s—roughly 60 years after Governor Wilson’s tenure—that Delaware truly overtook New Jersey as the state with the most corporate registrations:

New Jersey’s reign lasted far longer than the mainstream narrative suggests. Delaware’s eventual overtaking was likely catalyzed by its passage of director liability–related laws that proved especially attractive to public companies—combined with self-reinforcing network effects that generated their own momentum.

Second, regardless of what’s happening with high-profile public companies (and those captured in Ramp’s dataset), Delaware as a whole appears not only healthy—but thriving:

According to data published by Harvard Law School’s Forum on Corporate Governance, Delaware’s share of total U.S. corporations actually surged between late 2024 and 2025.

In fact, if you’re looking for a clear-cut “DExit” case, it’s probably this one—and it has nothing to do with Tesla. Instead, it involves a specific corporate structure:

Wyoming LLCs began surging around 2015.

Why? Likely due to specific asset-protection and privacy provisions embedded in Wyoming’s LLC statutes—a structure the state itself promotes as the “cowboy cocktail.”

In short, the point here isn’t that DExit isn’t happening (because at least some data suggest it is—even a small number of high-profile relocations carries outsized significance). But reality is certainly more nuanced than the mainstream narrative implies.

The truth is, Delaware still enjoys the advantage of being the default option—not to mention all the network effects bundled with that status—which are extraordinarily difficult to dislodge.

We previously published an earlier version of this chart, but as more data accumulate, its impact grows even more striking.

Token costs down, token consumption up:

Since the start of this year, paid-token pricing has dropped from ~$0.90 to $0.50 per million tokens, while token volume processed has nearly doubled—from ~6,000 to 12,000.

This is a textbook Jevons effect: the cheaper AI becomes, the more we use it. A welcome development.

Remember when people claimed older GPUs would become obsolete once newer, better ones hit the market?

That doesn’t seem to be happening either:

According to Silicon Data, rental prices for both NVIDIA’s H100 and A100 chips have risen this year.

The market shows no sign of compute oversupply—in fact, it appears we’ve barely scratched the surface of existing demand.

This comparison isn’t perfect, but if history offers any guide, we may need some time before we truly understand what an “AI-driven” economy looks like:

From Faraday’s and Henry’s earliest explorations of electric current to the industrial productivity boom of the first half of the 20th century, roughly 100 years elapsed.

Since the 1820s, technological iteration cycles have indeed accelerated—but platform-level transitions still involve an extremely complex web of variables.

As Roy Amara famously observed: “We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.”

Capital Expenditures—Placed on the Coordinate Axis

Here’s a set of data that never goes out of style: AI capital expenditures are enormous.

Consider these comparisons:

Projected 2026 AI capex is approaching the total net new loan volume issued by U.S. banks in 2025:

Capex is about 33% higher than total U.S. corporate income tax revenue—and roughly triple the total value of U.S. tariffs:

Capex is approximately six times the annual military budget of any non-U.S. G7 nation:

So yes—the scale of capital expenditure really is enormous.

Kalshi Enters Macro Forecasting

Federal Reserve researchers consider prediction markets quite effective.

At least on one metric, Kalshi’s forecasts for the federal funds rate have already surpassed those of professional forecasting institutions:

For the federal funds rate 150 days ahead (i.e., three FOMC meetings), Kalshi’s mean absolute error closely matches that of professional forecasters. But unlike surveys—which provide only a single modal path snapshot every six weeks—Kalshi delivers continuously updated, full probability distributions. We find Kalshi’s median and mode forecasts achieve perfect accuracy on the day before each FOMC meeting—an improvement over federal funds futures that is statistically significant.

In other words, although all forecasters start from similar positions, Kalshi’s “continuously updated” approach refines predictions over time—reaching “perfect predictive accuracy” the day before the official rate announcement. Moreover, Kalshi outperforms futures-market forecasts.

Kalshi’s edge extends beyond the federal funds rate. As Federal Reserve researchers note, because there are no other options markets for inflation, growth, unemployment, and similar macro indicators, Kalshi stands alone in offering a “high-frequency, continuously updated, rich-probability-distribution benchmark” reflecting “the crowd’s” judgment on where these economic indicators are headed.

That sounds pretty important.

The Delay of Adulthood

This is a thought-provoking chart—with (a little) commentary:

The share of 30-year-olds achieving key life milestones has been declining steeply since at least the 1980s.

Among 30-year-olds, fewer and fewer:

Live independently;

Have ever been married;

Live with children;

Own their own home.

The sole exception is college enrollment—since 1995, the share of 30-year-olds holding bachelor’s degrees has nearly doubled.

So—is college worth it?

Milestones? More like millstones around the neck, right?!

Maybe. Maybe not. But buyer’s remorse seems to be in the air.