← Full Report
1 / 15

DVC presents

STATE OF AI

The Operating Manual for the AI Revolution

A 20-minute compressed walk through the full stack — where money flows, who wins, what changes.

Read the full report state-of-ai.dvc.ai Scan or visit · May 2026
→ Full report

This runs today.

A day in the life — already automated

→ Dive deeper

Tech cycles

The price of intelligence is collapsing

Every platform made one scarce input cheap. AI makes custom software cheap enough to give every workflow its own brain.

2023 was ignition. · 2026 is still the beginning.

Same software arc. · New cost curve. · Very beginning.

→ Dive deeper

The Stack · 1 of 3

Five layers. One economy.

$60B at the top generates ~$700B of buildout below (Big 4 2026 midpoint; up to $725B at top end).

Revenue vs Margin · 2 of 3

Same stack. Two lenses.

Revenue grew everywhere. Margins didn't follow. The top earns software economics. The bottom earns utility.

Forces on the Stack · 3 of 3

Two forces. One stack.

Down: commoditization. Up: personalization. The moat moved up.

$60Bapps revenue · ~$700B2026 capex (mid · up to $725B) · 12 : 1infra : app revenue
+400%app revenue growth · 50%app gross margin · 12%energy margin (utility)
99.6% ↓inference cost / 24 mo · moat ↑value migrates up · $60B → ~$700B12 : 1 stack ratio

Tap a state to switch lenses

→ Dive deeper

Application layer

The user never sees the model.

Whoever owns the application picks the model. Distribution > intelligence.

→ Dive deeper

Model layer · 15 frontier models

Frontier broke into three races.

Commodity. Specialist. Restricted.

→ Dive deeper

Capability vs. cost

Smarter AND cheaper.

Every prior cycle: more power = more cost. AI inverted that. Embed intelligence everywhere.

→ Dive deeper

Infrastructure & Energy

Software demand. Industrial buildout.

Hyperscaler CapEx tripled in two years. Microsoft signed a 20-yr PPA for a targeted Three Mile Island restart. Google builds SMRs. NVIDIA uses equity to anchor preferred infrastructure partners (DSX as the reference architecture).

$224B2024 CapEx
$700B2026 GUIDANCE
3.1×in 2 years

Hyperscaler Capital Expenditure

~$700B2026 GUIDANCE (mid · up to $725B)
Show AI Revenue vs CapEx gap
Source: SEC filings, company earnings, Goldman Sachs, Epoch AI — 2026 figures are guidance midpoints
→ Dive deeper

Agent anatomy

It's not one product. It's a new stack.

47 companies across 7 layers, plus an emerging control plane above them: governance, agent registries, permissions, observability, and evals. $1B+ raised. Most didn't exist 18 months ago.

1 UI / Frontend Layer How agents meet users
CopilotKit 29.3K ★ DVC Vercel AI SDK 22.6K ★ AG-UI 12.4K ★ DVC Streamlit Gradio
12.4K GitHub ★ in <1 year — AG-UI is becoming the standard event protocol for agent-to-user interaction.
2 Orchestration Planning, routing, multi-agent
LangGraph 129K ★ CrewAI AutoGen Semantic Kernel Pydantic AI Sixtyfour Kapso
ReAct loops → multi-agent systems with planners, workers, verifiers. Most serious startups eventually build proprietary orchestration.
3 Memory State beyond the prompt window
mem0 49.6K ★ DVC Letta / MemGPT Zep Kite
+26% accuracy over OpenAI Memory, 90% less tokens. mem0 externalizes memory — works with any stack.
4 Tool / Action How agents act on the world
MCP 82.2K ★ A2A 22.5K ★ Function Calling Browserbase Composio Firecrawl Browser Use Exa Orthogonal
"MCP is becoming the REST of the AI era" — MCP for tools, A2A for agent-to-agent, AG-UI for agent-to-user.
5 Foundation Models The reasoning engine(s)
OpenAI Anthropic Google xAI Meta / Llama DeepSeek Mistral
37% of enterprises use 5+ models in production. Multi-model routing is standard — Harvey uses 6+ providers.
6 Execution Where code actually runs
E2B Daytona WebContainers Modal Dynamo NVIDIA OSS
3 patterns: local, cloud sandbox, browser-native. Cursor = local, Devin = cloud, Bolt = browser. Dynamo: 30× inference throughput.
7 Eval, Voice & Comms Observability, quality, output
ElevenLabs Vapi LangSmith Phoenix / Arize Braintrust AgentMail
Best startups treat eval as product, not afterthought. Harvey: BigLaw Bench. Perplexity: search_evals.
Sources: Crunchbase, TechCrunch, GitHub, npm, PyPI · 2025–2026
→ Dive deeper

Vibe coding

Code is becoming a throwaway artifact.

42% of committed code is AI-generated or significantly AI-assisted. 72% of developers who tried AI use it every day. Lovable hit $400M ARR. The roofing company runs AI agents now.

Source: SonarSource 2026 State of Code Developer Survey (42% AI-authored code, 72% daily use among triers). Stack Overflow 2025 — 51% daily — as secondary signal.

→ Dive deeper

Business models · usage is the engine

Per-token beats per-seat when AI does the work.

Anthropic 3× in 4 months on per-token API. Perplexity +50% in a month after Computer. Usage is where revenue scales with value.

👤
Per-seat
Per-token (API)
Per-subscription
Per-ad
Per-outcome / per-task

Charge for the work. Anthropic: $30B+ run-rate, in 4 months, ARPU $16.20/mo — per-token. Perplexity: ~$420M ARR after Computer (per-task). OpenAI: 900M WAU, only 5.6% pay — subscription distributes; usage captures.

Anthropic (per-token)
~$194/yr
Netflix (sub)
$138/yr
Meta (ads)
$58/yr
Google (ads)
$51/yr
OpenAI (blended sub)
~$28/yr

Outcome is the direction. Hybrid usage is the reality. Subs are a floor. Anthropic ARPU $16.20/mo vs OpenAI $2.20, Google $1.10. Coatue: only 3.8% of traditional SaaS spend is consumption-based so far. Orb: most agent companies are on hybrid pricing today.

Source: Counterpoint Research, Anthropic, Perplexity, OpenAI, Coatue C:\Takes, Orb 2026 State of AI Agent Pricing
→ Dive deeper

Services disruption

$1.5T of services. Mapped.

Sequoia's framework. Outsourced × Intelligence = autopilot territory. DVC plays across nearly every quadrant.

OUTSOURCED
JUDGEMENT
INTELLIGENCE
COPILOT TERRITORY ~$370B

Outsourced × Judgement — AI augments, humans decide

Management consulting $300B+ Graphic / UX design $30B+ Executive search $20B+ PR & comms $20B+
AUTOPILOT TERRITORY ~$700B

Outsourced × Intelligence — ripe for full automation

Insurance brokerage $140–200B IT managed services $100B+ Payroll & compliance $50–70B Claims adjusting $50–80B Accounting & audit $50–80B Healthcare rev cycle $50–80B Mortgage origination $30–50B KYC / AML $30–50B Paralegal / LPO $36B Tax advisory $30–35B Legal transactional $20–25B Real estate closing $20–25B Cost estimation $16B
WATCH ~$900B

Insourced × Judgement — hardest to automate

Recruitment $200B+ Advertising $100B+ Freight brokerage $100B+ Admin assistants $80B+ Clinical trials / CRO $80B+ SEO / SEM $50B+ ERP implementation $50B+ Corporate training $50B+ Market research $45B Cybersecurity $30B+ Architecture $25B+ Patent / IP $15–20B Travel mgmt $15B+
NEXT WAVE ~$315B

Insourced × Intelligence — automation coming fast

Supply chain & procurement $200B+ Pharmacy back-office $30B+ Wealth mgmt ops $30B+ Medical admin $20B+ Fund administration $15–20B
INSOURCED
Framework: Sequoia Capital, 2026 — contenders filtered to $10M+ raised — DVC portfolio highlighted
→ Dive deeper

How founders attack the $1.5T services map

Three angles. Pick one.

SaaS to incumbents. Vertical agentic. Agentic rollup. The market structure picks the strategy — not the founder.

→ Dive deeper

Physical AI

It already left the screen.

$38 trillion labor market in play. Slower than software people think. Faster than industrial incumbents hope.

→ Dive deeper

Key learnings

Six things we now believe.

After mapping the stack: where value is shifting, what's structural, and where today's leaders are most exposed.

  1. 01

    Intelligence got cheap. The stack didn't.

    Inference fell 99.6% in 36 months. The application surface around the model — distribution, memory, tools, eval — is still being invented.

  2. 02

    The bottleneck moved to atoms.

    Compute, power, memory, distribution. ~$700B of 2026 hyperscaler CapEx for ~$60B of app revenue. Whoever owns the scarce input prices the cycle.

  3. 03

    Pricing is unbundling from seats.

    Per-token, per-task, per-work-unit beat flat seats when the agent does the work. Outcome pricing is where AI margin actually shows up.

  4. 04

    Agents and vibe coding shipped.

    42% of committed code is AI-authored or AI-assisted. Cursor, Claude Code, Lovable, Devin are revenue lines, not demos. Software creation is a new workflow.

  5. 05

    Coalitions, not empires.

    Three stacks now define the frontier. OpenAI / Microsoft / Oracle / AWS: exclusivity reset, multi-cloud assembly, no single protector. Anthropic / Amazon / Google / SpaceX / Cursor: diversified compute and distribution, with a competitor (xAI) already inside the tent. Meta: separate, vertically controlled. Everyone gets stronger; everyone gets more exposed.

  6. 06

    Physical AI is next — and slower.

    $38T labor market in play, but latency, safety, and regulation gate the timeline. Mobility and logistics first; humanoids in controlled environments before open-world.

You are not late to a trend. You are early to a restructuring of the global economy.

→ Dive deeper