Singular Link Intelligence · Issue 01 · July 2026

The State of the AI Economy — In Numbers

As part of our ongoing effort to keep friends of Singular Link informed on the AI investment landscape, we share a few findings from Exponential View’s newly released State of the AI Economy report — the first bottom-up, deduplicated measurement of real customer demand for AI. We found it one of the most rigorous pieces of research published on the sector to date, and several of its findings challenge common narratives.

$175B
Annualized revenue run rate for generative AI
$110B
Banked trailing-12-month revenue
Faster than the internet, mobile or cloud at the same stage
0.42%
Of US GDP — still remarkably early

Three numbers stood out to us

01 · Demand is real, not projected

$110B of actual revenue — a run rate above $175B

Generative AI generated $110 billion in actual revenue over the past twelve months, with the annualized run rate now above $175 billion — growing roughly three times faster than the internet, mobile or cloud waves at the same stage. In 2023, the industry needed 180 days to add $1 billion of revenue; today it needs less than two.

02 · The buildout is beginning to pay for itself

Q4 2025 — revenues first cleared infrastructure depreciation

For the first time, quarterly AI revenues exceeded the depreciation cost of the infrastructure serving them — and did so again in Q1 2026. The margin is thin, but it is the most conservative test of whether $2 trillion of committed capital expenditure is economically rational, and the industry now passes it.

03 · It is still remarkably early

Just 0.42% of US GDP, against 9.4% for IT as a whole

All of this activity amounts to a small fraction of the economy. The deeper insight is structural: value flows to specific layers of the chain — power, chips, compute, networking, models and applications — each with very different economics. Understanding which layers hold pricing power, and which are commoditizing, is in our view the central question for anyone allocating to this sector this decade.

The AI economy value chain — and where Singular Link sits

The stack turns capital and energy into cognitive work: chips convert energy to tokens, hosting converts CapEx to token OpEx, models convert tokens to intelligence, and apps convert intelligence to customer value. Highlighted cells mark Singular Link’s current and pipeline positions.

01 · Power
Power
Utilities · gas turbines · SMR nuclear · fusion
Type One Energy (stellarator fusion)
02 · Chips
Chips
Nvidia · AMD · TSMC
Cerebras (wafer-scale, Nasdaq: CBRS) · Cortical Labs (biological compute)
03 · Compute
Compute
CoreWeave · Nebius · AWS · GCP
Fluidstack (AI GPU cloud) · Starcloud (orbital data centers) · FluidCloud (multicloud infra)
04 · Networking
Networking
Arista · Broadcom · Cisco
Ayar Labs (optical I/O) · Lightmatter (silicon photonics)
05 · Memory
Memory
Micron · SK Hynix · Samsung
Open
06 · Inference / Hosting
Inference / Hosting
RunPod · Together AI · Fireworks
Served via Fluidstack GPU capacity
07 · Foundation Models
Foundation Models
Anthropic · OpenAI · Google DeepMind
Reka AI (multimodal frontier) · Hologen AI (Large Medicine Models)
08 · Evals
Evals
Scale AI · Epoch AI
Open
09 · Agents
Agents
OpenAI · Anthropic
Lucid Capital (autonomous AI trading agents)
10 · Apps & Human Interface
Apps & Human Interface
Cursor · Harvey · OpenRouter
NovaLexi (AI IP management, KSA) · Synchron (endovascular BCI) · Evolve Neuro (neurotech)

Conclusions & risks

Key conclusions

  • Demand is real, big and fast — more revenue-validated than any prior platform shift, driven by external paying customers.
  • The biggest buildout in tech history is paying back, for now: revenues first cleared quarterly depreciation in Q4 2025, again in Q1 2026.
  • Value is moving up the stack toward apps and models; chat → agents is multiplying token use.
  • Tokens are AI’s billing metric but not yet its unit of value.

Biggest risks

  • The depreciation base rises as committed CapEx enters service — growth, utilization and pricing must compound or headroom compresses.
  • Open-weight models commoditize last year’s frontier, squeezing lab margins and generic “wrapper” pricing power.
  • Debt-heavy external financing moves repayment risk to third parties.
  • Falling token prices may not move enough volume to earn a return on $2T of CapEx.

Our read

This is not a bubble — it is a rebuild of the world’s computing infrastructure, and demand is more revenue-validated than any prior platform shift. The investment case comes down to whether falling prices can move enough token volume to earn a return on CapEx. That is precisely the question Singular Link is built to answer: we invest across the layers that convert energy into intelligence, positioned where pricing power is most durable.

The full visual summary. Singular Link’s one-page map of the report’s key figures across the value chain.
Download PDF Mobile PDF

Source: Exponential View — The State of the AI Economy · June 25, 2026 · Azeem Azhar, William Gildea, Hannah Petrovic PhD, Nathan Warren & Marija Gavrilov. Figures: global ex-China, deduplicated (tokens incl. China). Charts and value-chain mapping prepared by Singular Link; all underlying figures are the work of Exponential View and are reproduced here with attribution for commentary.

This material is provided by Singular Link for informational purposes only. It is not investment advice, nor an offer to sell or a solicitation of an offer to buy any security or interest in any fund. Company names shown reflect current and pipeline positions as of the date of publication, are illustrative of Singular Link’s thematic focus, and should not be taken as a recommendation, a complete list of holdings, or an indication of future portfolio composition. Past performance is not indicative of future results.