Replit
Freemium subscriptions (Core $20/mo, Pro $100/mo after the Feb 2026 repricing) plus usage-based agent inference and cloud consumption billing, deployment/hosting fees, and enterprise contracts; marketplace take-rate on bounties
Valuation 3x'd twice in quick succession: $1.16B (Apr 2023) to $3B (Sep 2025) on the Agent-driven ARR ramp, then to $9B (Mar 2026) six months later. All marks are primary priced rounds; no known secondary or SPAC events. The 2021 $800M figure is the widely reported post-money for the Coatue-led Series B.
Earnings, margins, COGS & capex
Replit is one of the fastest revenue-scaling software companies on record: company-stated $2.8M annualized revenue pre-Agent to $150M (Sep 2025) to $300M (end-2025, Sacra est.) to an estimated ~$525M annualized (Apr 2026), powered by consumption-based billing for its autonomous coding agent layered on subscriptions. The trade-off is gross margin: paying Anthropic/Google for frontier-model inference compressed 2025 gross margins to an estimated 36% at best and negative 14% at worst; the Feb 2026 pricing overhaul (Pro $100/mo, Core $25 to $20) was an explicit margin-repair move. Profitability is undisclosed and the $1B end-2026 ARR target assumes continued agent-usage growth.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~64¢ is cost of goods and ~36¢ operating expense, leaving ~0¢ of operating profit.
Revenue trend
Margins
volatile; hostage to LLM token pricing and agent efficiency; Feb 2026 repricing is the repair attempt
presumed deeply negative; funded by venture capital
COGS structure
Dominated by frontier-model inference purchased from Anthropic (Claude) and Google (Gemini 3, 2.5 Flash, Imagen 4 integrated since the Dec 2025 partnership), plus Google Cloud compute/storage/deployment infrastructure serving 50M+ users. Model-API spend scales roughly with agent usage, making COGS consumption-linked; agent efficiency gains (Agent 3's testing loop was claimed ~10x more cost-effective than computer-use approaches) are the main margin lever.
Capex
Minimal traditional capex - Replit rents rather than owns compute (multi-year Google Cloud partnership signed Dec 2025; Google remains primary cloud provider). Investment shows up as opex (inference, cloud) and R&D headcount, not PP&E.
Latest earnings
n/a
Company-stated goal of $1B ARR by end of 2026 (per Forbes/TechCrunch coverage of the Series D)
- Last round
- $400M Series D at $9B, Mar 2026, led by Georgian; G Squared, Prysm, Coatue, a16z, Craft, Y Combinator, Accenture Ventures, Okta Ventures, Databricks Ventures participating
- Prior round
- $250M at $3B, Sep 2025, led by Prysm Capital (Google AI Futures Fund, Amex Ventures participating)
- Users
- 50M+ (Mar 2026)
- ARR run-rate
- ~$525M annualized (Apr 2026, Sacra est.)
- Implied EV/ARR
- ~17x on Apr-2026 run-rate; ~9x if $1B target is hit
Growth drivers
- Agent adoption — Agent 4 (launched Mar 2026; parallel agents on one project, design canvas, mobile app May 2026) converts free users into consumption revenue
- Market expansion from ~30M professional developers to hundreds of millions of non-programmer knowledge workers
- Enterprise push — users at 85% of the Fortune 500 on-platform; multi-year Google Cloud partnership (Dec 2025) with Marketplace co-sell and Gemini Enterprise Agent Gallery distribution
- Pricing mix shift — usage-based agent billing on top of Core/Pro subscriptions raises ARPU; Feb 2026 repricing lifted Pro to $100/mo
- User growth: 40M+ (Sep 2025) to 50M+ (Mar 2026)
Bull & bear
Replit is a category-defining platform for the largest new software market of the decade - non-programmers creating real software - with hypergrowth, a full-stack product rivals lack, and strategic capital and distribution from Google. If it hits $1B ARR in 2026, the $9B mark looks cheap versus AI-coding comps.
- Revenue trajectory is nearly unprecedented: $150M (Sep 2025) to ~$525M annualized (Apr 2026, Sacra est.), with a company-stated path to $1B ARR by end-2026 - at which point the Series D price is ~9x forward ARR, well below the ~15x SpaceX paid for Anysphere (~$60B on ~$4B ARR, Jun 2026)
- Owning the full lifecycle (agent + IDE + database + auth + hosting) creates switching costs point tools like Cursor or v0 lack - the app lives on Replit's infrastructure and keeps paying hosting/usage forever
- The buyer base is structurally bigger than developers: 'everyone with a business problem' versus ~30M professional coders; Replit's decade of education/community footprint and 50M+ users is a head start on that demographic
- Google alignment (AI Futures Fund investor, multi-year cloud partnership signed Dec 2025, Gemini Enterprise Agent Gallery distribution, 2026 Google Cloud Partner of the Year) provides model access, enterprise channel, and an implicit strategic-acquirer floor
- Agent 4 (Mar 2026) runs parallel agents on a single project and adds a design canvas plus mobile creation (May 2026) - application-layer engineering that both widens the non-coder funnel and, with the Feb 2026 repricing, attacks the gross-margin problem
- 85% of the Fortune 500 already have users on the platform - land-and-expand into paid enterprise contracts is a large, mostly unmonetized wedge
Replit is a negative-to-thin gross-margin reseller of other companies' intelligence, in the most crowded category in software, priced at $9B on estimated consumption revenue whose durability is unproven - and its model suppliers are its most dangerous competitors.
- Gross margin is the tell: estimated between +36% and -14% in 2025 (Sacra) means Replit sometimes lost money on every agent run; a software business at these margins is structurally closer to a cloud reseller than a SaaS company, and deserves a far lower revenue multiple
- The intelligence layer is rented from Anthropic and Google - both of whom sell competing products (Claude Code, Gemini/AI Studio app-building) directly; Replit's differentiation must survive its own suppliers' roadmaps
- Consumption ARR from prosumer vibe-coders is not contracted revenue: hobbyist experimentation can churn fast when novelty fades or a cheaper rival (Lovable, Bolt.new, v0) undercuts, making the '$525M annualized' snapshot a fragile extrapolation - and the Feb 2026 repricing itself admits the prior pricing was underwater
- Valuation tripled (Sep 2025 $3B to Mar 2026 $9B) in six months on momentum-market dynamics; any AI-funding cooldown or a missed $1B ARR target leaves late investors underwater and could force a flat/down round
- Trust and safety incidents are existential for the 'non-programmer' promise - the widely covered July 2025 episode of Replit's agent deleting a production database illustrates that autonomous agents plus naive users is a liability tail-risk enterprises price in
- Anysphere/Cursor owns the high-value professional-developer segment at ~$4B ARR - roughly 8x Replit - and is being folded into SpaceX ($60B all-stock deal, Jun 2026), capping Replit's ability to move up-market while low-end clones attack from below
What it is worth
Last priced round anchored, cross-checked with EV/ARR comps for AI-coding platforms (SpaceX-Anysphere at ~15x ARR Jun 2026; Anysphere Series D ~$29.3B Nov 2025; Lovable $6.6B Dec 2025; public high-growth software multiples as a sanity floor)
~$3-5B
consumption churn plus foundation-model competition stalls ARR below ~$700M, gross margin stays sub-30%, AI-funding cooldown compresses multiples toward 5-7x - a flat-to-down round versus Series D
~$9-12B
ARR lands $800M-$1B but margins remain thin; valuation grows roughly with revenue, no multiple expansion
~$15-20B within 12-18 months
$1B+ ARR achieved, gross margin repaired above 50% on cheaper inference plus the Feb 2026 repricing and Agent efficiency, enterprise ACVs ramping - supports 15-20x ARR at IPO-candidate quality, in line with what SpaceX paid for Anysphere
Mar 2026 Series D set $9B, or ~17x the ~$525M Apr-2026 annualized run-rate (Sacra est.) and ~9x the company's $1B end-2026 target. That is inside the range paid for AI-coding leaders in 2025-26 (SpaceX-Cursor ~15x, Lovable ~20x+ at Series B) but assumes both the ARR target lands and gross margin turns durably positive. Private-company figures are self-reported or third-party estimates - treat multiples as indicative. Not financial advice.
SWOT
Strengths
- End-to-end platform — idea to deployed, hosted app in one product (IDE, agent, database, auth, hosting) - unlike point-tool rivals
- Fastest-in-class revenue scaling (~$150M to ~$525M annualized in ~7 months) with consumption pricing aligned to value delivered
- Distribution moat-in-progress — 50M+ users, presence at 85% of Fortune 500, 10+ years of community and education footprint
- Deep-pocketed strategic backers (Google AI Futures Fund, a16z, Coatue, Databricks Ventures, Accenture Ventures) and ~$650M fresh capital
Weaknesses
- Structurally thin/negative gross margins (est. +36% to -14% in 2025, Sacra) - resells frontier-model inference it does not own
- No proprietary foundation model; core intelligence sourced from Anthropic/Google, who also compete downstream
- Consumption revenue from prosumer vibe-coders may be spiky and churn-prone versus contracted enterprise seats
- Reliability/trust scar tissue — the July 2025 incident where Replit's agent deleted a customer's production database went viral
Opportunities
- TAM expansion from developer tools (~$30B) to citizen-developer app creation across the broader knowledge workforce
- Enterprise tier — multi-year Google Cloud partnership, Marketplace co-sell, and Gemini Enterprise Agent Gallery distribution open regulated, high-ACV accounts
- Agent efficiency gains plus the Feb 2026 repricing can structurally repair gross margin as token prices fall
- Marketplace and hosting attach — monetizing the apps users build (deployment, databases, payments rails) beyond creation
Threats
- Foundation-model vendors moving up-stack — Anthropic's Claude Code, OpenAI's Codex, Google's own Gemini coding surfaces commoditize the agent layer
- Anysphere (Cursor) at ~$4B ARR — roughly 8x Replit's run-rate - dominates the professional-developer segment and gains SpaceX's balance sheet via the pending $60B acquisition (announced Jun 2026)
- Cheap fast-followers (Lovable at ~$500M ARR, StackBlitz Bolt.new, Vercel v0, Wix Base44) compress prosumer pricing
- If LLM prices stay high while competition forces retail prices down, the negative-gross-margin trap becomes permanent
- AI-app-security and agent-safety incidents (data loss, insecure generated apps) could trigger enterprise hesitancy or regulation
Moats, dependencies & bottlenecks
Moats
moderate-strong Apps built on Replit run on Replit; migration cost grows with each deployed app. Closest thing to a durable moat here.
A decade of free-tier and classroom presence seeds the funnel; but users are multi-homing across AI tools.
Proprietary agent scaffolding (Agent 4 parallel agents + design canvas; Agent 3's autonomous testing loop) Real engineering edge today (Agent 3 claimed ~10x cost advantage over computer-use approaches), but application-layer techniques diffuse quickly.
Category association is valuable but contested by Lovable, Bolt, v0 and by foundation-model vendors.
Usage telemetry helps tune agents, but the base models it cannot train remain the core intelligence.
Dependencies
core agent intelligence Replit Agent has been powered primarily by Claude; Anthropic's own Claude Code competes with Replit. Pricing and access terms set Replit's COGS.
supplier + partner + investor Multi-year cloud partnership (Dec 2025; Google is primary cloud provider), Gemini 3 / Imagen 4 model integration, Google AI Futures Fund on cap table, Gemini Enterprise Agent Gallery distribution - concentrated but currently aligned.
Gross margin recovery depends on token prices falling faster than Replit's retail pricing.
Undisclosed but presumed cash burn; ~$650M raised in ~6 months implies the model consumes capital until margins normalize.
Browser-native product (plus its own iPhone app since May 2026) avoids dependence on platform gatekeepers.
Advantages
- Only major player owning creation-to-production lifecycle in one product (build, test, deploy, host, monetize)
- Consumption pricing aligned with delivered value — revenue scales automatically with agent capability improvements
- Ten-year head start on community, education, and mobile/browser-based coding versus 2023-vintage rivals
- Strategic investor set (Google, Databricks, Accenture, Okta, Amex) doubling as distribution and integration partners
- Capital cushion (~$650M raised Sep 2025-Mar 2026) to outspend smaller prosumer rivals
Weaknesses
- Structurally thin/negative gross margins (est. +36% to -14% in 2025, Sacra) - resells frontier-model inference it does not own
- No proprietary foundation model; core intelligence sourced from Anthropic/Google, who also compete downstream
- Consumption revenue from prosumer vibe-coders may be spiky and churn-prone versus contracted enterprise seats
- Reliability/trust scar tissue — the July 2025 incident where Replit's agent deleted a customer's production database went viral
Bottlenecks
- Gross margin — cannot scale profitably until agent inference cost per completed app falls well below price charged
- Agent reliability ceiling — autonomous runs still fail or misbehave on complex apps, capping enterprise willingness to pay
- Enterprise sales motion — converting 'Fortune 500 users present' into contracted six-figure ACVs requires a GTM muscle Replit is still building
- Model access concentration — no owned foundation model; capacity/priority during industry-wide compute crunches is negotiated, not controlled
Top signals & trends
Top signals
Top-tier investors underwriting continued hypergrowth; also a froth indicator.
Sacra estimate; consumption momentum intact post-round.
The single most important number to watch; the repricing is the margin-repair attempt and the tell that v2 pricing was underwater.
Enterprise channel and model access secured; deepens supplier concentration.
Suppliers moving up-stack is the classic platform-risk pattern.
The pro-developer leader gains a mega-cap parent; also validates AI-coding multiples (~15x ARR) that flatter Replit's mark.
Hit = re-rating validation; miss = down-round risk for Series D holders.
Funnel breadth; monetization depth still to be proven.
Trends
strongly positive · Replit is a category leader in the defining application-layer AI trend of 2025-2026.
Directly repairs Replit's gross margin if retail pricing holds.
Claude Code, OpenAI Codex, Gemini coding surfaces squeeze the independent agent layer.
Validates category value and Replit's strategic-acquirer floor, but arms rivals with mega-cap balance sheets.
Opens high-ACV contracts but raises the bar on safety, audit, and reliability where Replit has scar tissue.
A cooldown hits negative-gross-margin hypergrowth names first.
Budget migration from labor to tools expands the addressable spend pool Replit bills against.
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 agent model (Claude) - a large COGS line.
Primary cloud provider (Cloud Run, GKE, BigQuery); Gemini 3 / Imagen 4 model integration; multi-year partnership (Dec 2025).
Secondary model option on the platform.
Indirect - GPU layer beneath every inference dollar Replit spends.
Core of consumption revenue - 50M+ registered users building personal and business apps.
Users present at ~85% of the Fortune 500; enterprise contracts are the monetization frontier.
Legacy Teams for Education base feeding the top of funnel.
Use agent-built apps as production internal tools and MVPs.
AI code editor for professional developers; ~$4B ARR by Jun 2026 - roughly 8x Replit - expanding toward non-coders.
Fastest-growing European vibe-coding rival aimed at the same non-programmer buyer; ~$500M ARR, $6.6B valuation (Dec 2025), reportedly in talks at $12B.
Default AI coding layer for the world's largest developer platform; enterprise distribution unmatched.
Autonomous coding agent from a model owner with zero marginal intelligence cost disadvantage.
Replit's own model supplier selling a competing agentic coding product - supplier-turned-competitor risk.
Partner and investor, but also ships its own app-building surfaces; coopetition.
Prompt-to-UI generation plus best-in-class web hosting; overlaps Replit's deploy layer.
Browser-based prompt-to-app rival with viral prosumer traction.
Acquired vibe-coding startup Base44 (Jun 2025); attacks the SMB no-code app segment.
AWS agentic IDE push bundled into the largest cloud's enterprise relationships.