Anysphere (Cursor)
Bottom-up SaaS: freemium individual subscriptions (Pro seat) plus usage-metered enterprise/team seats; revenue is model-inference-intensive, with COGS dominated by pass-through LLM inference (Anthropic/OpenAI) increasingly offset by Anysphere's own in-house models.
Now part of SpaceX — ~$60B all-stock (announced Jun 16 2026, closing ~Q3 2026) · public-market exposure via SPCX
Chronological priced-round trail. Seed (2023) and Series A (~$400M, Aug 2024) valuations are approximate; Series C ($9.9B) and Series D ($29.3B) are confirmed. The Apr 2026 ~$2B at $50B pre-money (~$52B post) was in advanced talks but halted/superseded. Not plotted as a round: the announced Jun 16 2026 $60B all-stock SpaceX/xAI acquisition (pending ~Q3 2026 regulatory close) — an M&A valuation, not a closed primary/secondary/SPAC financing.
Earnings, margins, COGS & capex
Private company, no audited financials or S-1. The verifiable spine is priced venture rounds and widely-reported ARR milestones. Growth is exceptional and real; profitability is the open question. The core financial tension: revenue scales with heavy LLM usage, and inference COGS (paid to Anthropic/OpenAI) has historically kept gross margin far below the 80%+ software norm. The 2025-2026 strategic pivot to first-party models (Composer) is the margin-repair thesis. The announced $60B SpaceX acquisition, if it closes, supersedes the private-round path.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~70¢ is cost of goods and ~30¢ operating expense, leaving ~0¢ of operating profit.
Revenue trend
Margins
improving on enterprise as in-house inference displaces third-party API COGS
declining where Composer/first-party models serve requests
unclear; scale + first-party models are the levers
COGS structure
Dominated by LLM inference fees paid to Anthropic (Claude) and OpenAI (GPT), plus cloud compute (reported heavy AWS spend). Every power user's agentic/autocomplete request is a direct variable inference cost. Mitigation: Anysphere's own Composer model (launched Oct 29 2025 with Cursor 2.0, iterated through Composer 2.5 by 2026) reportedly serves a growing share of in-editor completions on par with frontier models for common tasks, cutting pass-through COGS.
Capex
Not separately disclosed. Rising from near-zero (pure API reseller) toward model training/serving infrastructure as the company builds first-party models; still far below hyperscaler capex intensity.
Latest earnings
N/A
N/A
- Annualized revenue (Jun 2026)
- ~$4B
- Enterprise share of revenue
- ~$2.6B (~65%)
- Fortune 500 penetration (company claim)
- 64%
- Enterprise code written/day (company claim)
- 100M+ lines
Growth drivers
- Enterprise land-and-expand — company reports 64% of Fortune 500 use Cursor, 50,000+ enterprises, 100M+ enterprise lines/day
- Product velocity in agentic coding (multi-file edits, background agents) keeping it ahead of incumbents
- Bottom-up developer adoption converting to paid seats and then enterprise contracts
- Secular tailwind: AI-assisted/agentic software development becoming default developer workflow
- First-party models (Composer) improving unit economics, enabling profitable enterprise pricing
Bull & bear
Cursor is the emergent standard for how software gets written, monetizing the single highest-ROI enterprise AI use case with unmatched growth, and its first-party-model pivot fixes the one real flaw (margins).
- Fastest software company to ~$4B annualized revenue ever, still ~doubling every few months
- AI coding is the most proven enterprise LLM use case; Cursor is the leading pure-play
- Composer and in-house inference are already improving enterprise gross margins, reframing the bear thesis
- Enterprise moat compounding: 64%-of-Fortune-500 footprint, switching costs, and workflow lock-in
- An announced $60B strategic acquisition validates the asset and grants hyperscaler-grade compute (xAI Colossus) and capital
Cursor is a thin-margin, thin-moat application layer renting its core intelligence from companies that can and will build the same product, priced for perfection at a valuation that ignores negative unit economics.
- Gross margins are structurally low (~30-50% est.) and the business reportedly loses money — a wrong-side-of-software cost structure
- The moat is a UX veneer over other people's models; Anthropic (Claude Code), OpenAI (Codex), and Microsoft (Copilot) can replicate and bundle
- Model-provider dependency is existential: pricing, rate limits, or a competing product from Anthropic/OpenAI can compress margins overnight
- $29.3B (and announced $60B) valuations imply flawless execution; any growth deceleration or margin stall re-rates it hard
- First-party models are unproven at sustained frontier quality and add training capex/execution risk to a company that was asset-light
What it is worth
Private, revenue-multiple anchored to priced rounds and the announced acquisition; no public price or DCF-grade disclosure.
Sub-$20B if incumbent bundling (Microsoft/Google) and forward integration by Anthropic/OpenAI commoditize the IDE layer, growth decelerates, and losses persist — a re-rate toward the underlying (low) margin structure.
~$29-50B range reflects a dominant-but-thin-margin leader dependent on frontier-model suppliers; multiple compresses from Series D levels as revenue outgrows valuation and unit economics remain the swing factor.
$60B+ justified
if Cursor sustains hyper-growth, Composer drives durable software-like gross margins, and enterprise lock-in compounds — a category-defining platform, not a wrapper.
Last CLOSED priced round: $29.3B post-money Series D (Nov 13 2025) on ~$1B ARR (~29x ARR). Announced $60B all-stock SpaceX acquisition (Jun 16 2026) against ~$4B annualized revenue implies ~15x forward ARR (CNBC's own framing) — a lower multiple than the Series D, reflecting both the revenue surge and the discount for a low-margin, high-dependency application-layer business. A ~$2B round at a $50B pre-money valuation in talks (Apr 2026) was halted/superseded. All non-round figures are third-party estimates; margins/burn are undisclosed.
SWOT
Strengths
- Category-defining product with best-in-class agentic coding UX and a devoted developer base
- Historic revenue velocity: ~$100M to ~$4B ARR in ~18 months
- Deep enterprise penetration (64% of Fortune 500 reported) with expanding contracts
- In-house model (Composer) reducing dependence on Anthropic/OpenAI and repairing gross margin
- Elite cap table (Thrive, Accel, Coatue, a16z, Nvidia, Google) and abundant capital
Weaknesses
- Structurally low gross margin vs software peers due to inference COGS; reportedly loss-making, especially on individual subscriptions
- Heavy dependence on third-party frontier models (Anthropic, OpenAI) that are also competitors
- Thin technical moat at the model layer — the editor UX is replicable by well-capitalized incumbents
- Concentration/key-person and integration risk if the announced SpaceX/xAI acquisition closes
- Private, opaque financials — margins, burn, and net loss not disclosed
Opportunities
- Own more of the stack with first-party models to convert negative/thin margins to durable software margins
- Expand enterprise ACVs: security, compliance, admin, and agent orchestration up-sell
- International and non-code adjacencies (agentic knowledge work) leveraging the same agent runtime
- Become the default enterprise agentic-coding platform before incumbents catch up
- Leverage xAI Colossus compute (if the acquisition closes) to train frontier-grade first-party coding models at hyperscaler scale
Threats
- Microsoft/GitHub Copilot bundling into the world's dominant developer ecosystem
- Model providers (Anthropic Claude Code, OpenAI Codex) integrating forward into the IDE and disintermediating Cursor
- Google (Gemini Code Assist, ex-Windsurf talent) and Amazon (Q Developer/Kiro) at hyperscaler scale/price
- Inference-cost economics of the whole AI-coding category ('everybody losing money on AI')
- Valuation/expectation risk — $29.3B-$60B on a loss-making, low-moat business if growth decelerates
- Regulatory/antitrust risk to the announced SpaceX/xAI acquisition and integration disruption if it clears
Moats, dependencies & bottlenecks
Moats
Real today, but replicable by well-funded incumbents; must keep out-shipping.
Loved brand and habit among developers; converts to enterprise seats.
Moderate-Strong 64%-of-Fortune-500 penetration, admin/security integration, workflow lock-in.
if it compounds Composer trained on real coding telemetry; the strongest potential durable moat and margin lever.
Abundant funding and strategic investors, but capital is not itself a defensible moat.
Dependencies
Supplier / model provider Historically the dominant back-end model and a direct COGS driver; also a forward competitor via Claude Code.
Supplier / model provider + early investor Model supplier and competitor (Codex); OpenAI Startup Fund led the 2023 seed.
Supplier / compute Reported heavy AWS spend; inference and serving compute.
Supplier + investor Underpins inference/training economics; Nvidia is on the cap table.
Product quality tracks the underlying models' capability curve.
Supplier / platform Cursor is a VS Code fork dependent on the open-source editor + extension ecosystem it also competes with (Copilot).
Advantages
- First-mover and brand leader in AI-native coding
- Fastest-scaling revenue in the category with genuine enterprise pull
- Vertically integrating into first-party models to own margin and reduce competitor dependency
- Capital-rich with a marquee strategic cap table (Nvidia, Google, Thrive, a16z, Accel, Coatue)
Weaknesses
- Loss-making with structurally low gross margins vs software norms
- Core intelligence historically rented from direct competitors
- Replicable application-layer moat exposed to incumbent bundling
- Opaque private financials; valuation priced for perfection
Bottlenecks
- Inference cost per request vs willingness-to-pay — the core unit-economics constraint
- Frontier-model quality/availability for its own Composer models vs Anthropic/OpenAI
- Enterprise security/compliance/governance requirements gating large-seat expansion
- Talent and compute to sustain in-house model training against much larger rivals
Top signals & trends
Top signals
Validates the asset and grants xAI compute/capital, but is announced/pending regulatory approval (CNBC/Quartz/Yahoo), not a closed venture round, and carries integration/key-person risk.
Demand at $50B pre-money (a16z/Thrive/Nvidia/Battery) shows investor appetite; superseded by the acquisition path.
Direct margin-repair and dependency-reduction proof point; iterated to Composer 2.5 by 2026.
Sustained hyper-growth, now enterprise-led (~65% of revenue).
Category-wide 'losing money on AI' economics remain the key overhang.
Trends
Expands Cursor's TAM and seat count.
Anthropic Claude Code, OpenAI Codex threaten disintermediation.
Cursor's Composer follows the pattern; owns margin and data.
Pressures valuations across AI-coding until margins prove out.
Microsoft/Amazon/Google can undercut on price via bundling.
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
Primary Claude model provider; largest COGS input and a competitor. Private.
GPT model supplier and 2023 seed lead (OpenAI Startup Fund); also a competitor. Private.
Cloud compute/serving infrastructure; reported heavy spend.
GPU supplier underpinning inference/training; also an investor.
Cursor is a VS Code fork; depends on the open-source VS Code/Open VSX ecosystem while competing with Copilot.
Enterprise software teams (64% of Fortune 500, per company) ~65% of revenue is enterprise B2B; 50,000+ enterprises reported.
Bottom-up freemium + Pro subscriptions; the top-of-funnel, reportedly lower-margin.
The incumbent AI pair-programmer bundled into GitHub/VS Code and enterprise agreements; largest distribution threat.
Hyperscaler AI coding assistant; absorbed key Windsurf talent in 2025; scale + model ownership.
AWS-native agentic coding tools with enterprise cloud distribution and pricing leverage.
Cursor's key model supplier and a forward competitor via its own terminal/agent coding product; private.
Model supplier, early Cursor investor, and competitor via Codex/coding agents; private.
AI-native browser IDE / agent focused on build-and-deploy; private (backed by a16z, Coatue).
Enterprise code AI and code search; private.
Enterprise/self-hosted AI code assistant emphasizing privacy; private.