in short
TKT generates monthly yield through three independent mechanisms: income-producing real estate (75% of capital), stablecoin staking protocols (20%), and non-custodial proprietary trading (segregated from investor capital entirely). The result is a structured 2.5%–5% monthly distribution, backed by a rolling three-month buffer and a 7.5% capital reserve. No investor capital is ever exposed to trading risk.
National accounts have absorbed new categories before. Railroads, then telephones, then software each went from a rounding error to a line item serious enough to move the headline growth number. Compute is next — and the strange thing is that a year ago almost no one was measuring it that way, and today everyone is measuring it differently.
The clearest single data point came from Harvard economist Jason Furman in late 2025. Investment in information-processing equipment and software, he calculated, was about 4% of U.S. GDP in the first half of 2025 — but it accounted for roughly 92% of GDP growth over that period (Jason Furman on X; Fortune). Strip those categories out, and the economy grew at an annualized rate of 0.1%. A near-standstill everywhere except the part building the machines that run artificial intelligence.
That is the kind of statistic that should make an investor sit up. Not because the number is precise — it isn’t, and we’ll get to why — but because it means the single most important variable in the U.S. growth picture is now one that no statistical agency reports as a standard series, and that no two analysts define the same way.
Why Compute Is Becoming a Macroeconomic Indicator
For decades, the macro dashboard has treated technology spending as a subplot inside business investment. That framing has quietly broken. AI infrastructure capital expenditure — the data centers, the chips inside them, the power and networking that make them run — has grown large enough, fast enough, and concentrated enough that it now behaves like its own macro force. It moves quarterly GDP prints. It shapes electricity demand and interest-rate expectations. It pulls capital away from other sectors.
The Infrastructure-Buildout Precedent
When a category does that, it stops being a subplot and becomes an indicator in its own right. The useful comparison isn’t to past tech cycles narrowly, but to the great infrastructure build-outs that economists have always read as macro events: the railways of the nineteenth century, the fiber-optic boom of the late 1990s. Each of those was, at its peak, measured as a share of national output — because that ratio was the cleanest way to see how much of an economy was being bet on a single kind of buildout.
The Case for a Compute-to-GDP Gauge
That’s the case for a compute-to-GDP ratio. Not as a curiosity, but as a gauge: how much of a country’s output is now flowing into the physical substrate of AI — the data centers, chips, and power that run it — and how fast that share is changing. Read it well and you understand where the growth is coming from, how durable it is, and what happens to the rest of the economy when it slows. Ignore it, and the headline GDP number becomes genuinely misleading — a figure that looks like broad expansion but is really one buildout wearing the whole economy as a trench coat.
The Evidence: AI Capex Is Already Moving National Output
This is no longer a projection. The spending shows up in the national accounts right now.
How big the buildout actually is
The scale has compounded with startling speed. The largest U.S. hyperscalers spent roughly $126 billion on AI-related infrastructure in 2023, about $197 billion in 2024, and — by most counts — north of $320 billion among the top four in 2025, pushing toward half a trillion dollars once the next tier is included. For 2026, Big Tech has disclosed plans in the range of $700 billion, and consensus estimates for the six leading hyperscalers point toward roughly $0.9 trillion over a forward twelve-month window, then past $1.1 trillion in the year after that. Capital spending has climbed above 20% of these companies’ revenue — roughly double the share of a decade ago. The most asset-light business model in modern capitalism has gone heavily asset-heavy in about three years.
Set against national output, that translates to a ratio rising almost vertically. By Apollo chief economist Torsten Slok’s accounting, data-center capex ran near 0.3% of U.S. GDP in 2019, 0.6% in 2023, and about 1.4% in 2025, with consensus implying roughly 3.1% by 2027 (Apollo — The Daily Spark; Benzinga). Independent estimates from Epoch AI land in the same neighborhood for the narrow measure — around 1.4% of GDP for data centers, hardware, and networking combined in early 2026, up from 0.7% a year earlier.
A capex cycle larger and faster than telecom or housing
Put in historical context, the numbers are hard to overstate. The late-1990s fiber-optic and telecom buildout — the one that ended in the dot-com bust — peaked at roughly 1.0% to 1.2% of U.S. GDP around 2000 (IEEE ComSoc). The AI buildout has already passed that. The 2000s housing boom peaked higher, at about 6.6% of GDP in 2005, so in absolute level the compute cycle still sits below it. But what matters for growth, as Slok emphasizes, is not the level of the share — it’s how fast the share is changing, because that change is what adds to or subtracts from GDP. On that measure the AI cycle is rising at roughly 0.85 percentage points of GDP per year, close to twice the pace of the housing boom at its fastest and many times the pace of the telecom build. In cumulative present-value terms, one estimate puts the internet build-out of 1996–2001 at about $3.3 trillion in today’s dollars; the AI infrastructure cycle is on track to dwarf it.
When Compute Passed Consumer Spending
There’s one more piece of evidence, and it’s the one with teeth: in August 2025, Renaissance Macro Research estimated that the dollar amount the AI data-center buildout was adding to GDP growth had, for the first time, surpassed the contribution from consumer spending (Yahoo Finance / Veritas News synthesis) — a category that is normally about two-thirds of the entire economy. That is the moment compute stopped being a sector and started being the story.
The Measurement Problem: Why Nobody Agrees on the Number
Here is the catch, and it’s the reason this deserves an investor’s skepticism rather than their awe: depending on who is counting, AI-related capital spending is somewhere between about 1.4% and 5% of U.S. GDP. That is not a rounding difference. It is the gap between “large and growing” and “the biggest capital cycle in modern history,” and it exists because there is no agreed definition of what to count.
The Definitions Don't Match
The narrow measures — Slok’s and Epoch’s ~1.4% — count data-center structures, hardware, and networking. Broader ones from firms like KKR reach roughly 5% by folding in the full sweep of AI-related capital formation (KKR). Furman’s 4% figure measures a different thing again: all information-processing equipment and software in the national accounts, only some of which is AI. Others, grossing up chipmaker revenue and applying an investment multiplier, land around 2%. Every one of these is defensible. None of them is comparing the same quantity.
One Country, No Standard Measure
The problem compounds across borders. The evidence base is overwhelmingly American — by one long-range projection the United States alone accounts for nearly half of all global AI infrastructure capex through 2050 (PwC) — and comparable national-accounts breakdowns barely exist for most other economies. So the honest state of play is that the single most important macro indicator of the AI era is measured ad hoc, in incompatible ways, in essentially one country. “The ratio nobody tracks” has become “the ratio everybody argues about and no one standardizes.”
Why the Ambiguity Matters
That ambiguity is not harmless. If growth is being carried by a share that must keep rising to keep contributing — and if that share is concentrated in a handful of firms whose payback timelines are still unproven — then getting the measurement wrong means misreading the entire economy’s health. Some analysts already frame it starkly: Deutsche Bank has argued the AI build-out is, for now, most of what is holding U.S. growth up. Whether that reassures or alarms you depends entirely on a number nobody yet computes consistently. One meaningful comfort, at least relative to the 1990s: this cycle has so far been funded largely by the cash flows of highly profitable companies rather than the speculative debt that sank the telecom carriers — though the recent drift toward off-balance-sheet financing is worth watching closely.
What a Compute-to-GDP Lens Reveals
The instinct to measure compute as a share of output is the right one, and it’s early enough to still be an edge. An investor — or a finance ministry — who builds a consistent compute-to-GDP series, and sticks to one definition long enough to read its trend, will see three things others miss: how much of reported growth is really one buildout, how exposed the economy is if that buildout plateaus, and how the ratio compares across the few countries where it can be estimated at all.
The Edge in Measuring It First
That last point is where the strategic advantage sits. Headline GDP will keep flattering economies that are pouring output into compute and keep understating the fragility underneath. The country that gets serious about measuring this first will understand its own economy better than its peers — and the investor who reads the compute-to-GDP ratio, definitional messiness and all, will understand the growth number better than the market that still takes it at face value.
The category is new, the accounting is unsettled, and the honest answer to “what share of GDP is compute?” is still a range rather than a number. But the direction is not in doubt, and neither is the importance. What used to be a line item is becoming the line.
source
covers
Original 4%/92% GDP figures
0.1% ex-AI growth figure
Torsten Slok's capex-to-GDP series
Slok housing/telecom comparison
Renaissance Macro consumer-spending crossover
Broader ~5%-of-GDP estimate
US share of global AI capex through 2050
Telecom-boom GDP-share comparison
