Supabase
Open-source core (Postgres) + usage-based managed cloud subscriptions (free/pro/team/enterprise tiers); land via free developer tier, expand with usage and enterprise contracts
Earlier rounds (YC S20; $30M Series A, Sep 2021; $80M Series B led by Felicis, May 2022; $80M Series C co-led by Craft Ventures and Peak XV, Sep 2024) did not carry officially disclosed valuations and are omitted from the plotted trail. All plotted marks are primary priced rounds, not secondary trades.
Earnings, margins, COGS & capex
No audited financials are public. Third-party (Sacra) estimates put ARR at ~$30M exiting 2024, ~$70M (Aug 2025), ~$101M (end-2025), and ~$170M (May 2026) - roughly 3x growth in 2025 and continued triple-digit pace into 2026, driven by AI codegen tools (Lovable, Bolt, Replit, and coding agents like Claude Code and Codex) defaulting to Supabase as the backend. Company-disclosed operating metrics: nearly 10M developers (Jun 2026, doubled in 8 months), database launches +600% YoY, >60% of new databases created by AI tools, and 250,000+ customers (Series F release). Monetization is usage-based cloud (compute hours, storage, bandwidth, MAUs) with an enterprise tier; conversion of the massive free tier is the central financial question.
Revenue trend
Margins
Structurally capped vs pure SaaS: COGS = cloud compute/storage resold as managed Postgres; free tier is a real cost center
Presumed negative at this growth stage; the $500M raise funds continued investment
COGS structure
Primarily cloud infrastructure (AWS compute/storage/egress) underlying every hosted Postgres instance, plus the free-tier subsidy (millions of free projects), support, and managed-service SRE labor. Multigres (Vitess-style Postgres sharding/proxy layer, announced Jun 2025, v0.1 alpha released alongside the Series F in Jun 2026) is partly a COGS/scale play - denser multi-tenancy and horizontal scaling improve unit economics.
Capex
Minimal owned capex; infrastructure is rented from hyperscalers (shows up in COGS/opex, not capex). No disclosed plans for owned data centers.
Latest earnings
Not applicable
None issued; the company discloses usage metrics (developers, database launches, customers), not revenue or margins
- Developers
- Nearly 10M (Jun 2026), 2x in 8 months
- Database launches
- +600% YoY; >60% created by AI tools
- Customers
- 250,000+ (Jun 2026, company release)
- ARR (third-party est.)
- ~$170M (May 2026, Sacra)
- Total equity raised
- Over $1B cumulative incl. $500M Series F
- Valuation path
- $2B (Apr 2025) -> $5B (Oct 2025) -> $10.5B (Jun 2026)
Growth drivers
- AI codegen / vibe-coding boom — >60% of new Supabase databases are created by AI tools (TechCrunch, Jun 2026); default backend for Lovable, Bolt, Replit, Figma-class app builders
- Developer-count flywheel — nearly 10M developers (Jun 2026), doubled in 8 months; bottoms-up free-to-paid conversion
- Postgres consolidation — one platform replaces database + auth + storage + realtime + edge functions + pgvector for AI apps
- Enterprise upsell — SOC 2/HIPAA compliance, SSO, dedicated infra as AI prototypes graduate to production; 250,000+ customers per the Series F release
- Multigres: Postgres horizontal sharding ('Vitess for Postgres', led by Vitess co-creator Sugu Sougoumarane; v0.1 alpha Jun 2026) opens large-scale enterprise workloads
- Agentic infrastructure positioning — company says agents now deploy the majority of new databases; CEO credits Claude Code and Codex for expanding who can build
Bull & bear
Supabase is the picks-and-shovels winner of AI-generated software: every AI-built app needs a database, auth, and storage, and Supabase is the default choice wired into the leading codegen tools. If even a modest fraction of nearly 10M developers and 600%-growing database launches convert to paid production workloads, ARR compounds at triple digits for years and the company becomes the next-generation MongoDB with a better open-source moat.
- Structural tailwind: >60% of new databases on the platform are created by AI tools, and agents now deploy the majority - Supabase is upstream of the entire vibe-coding wave, not a bet on any single tool
- Developer count doubled to nearly 10M in 8 months while estimated ARR grew from ~$101M (end-2025) to ~$170M (May 2026) - both curves still steep
- Postgres-native strategy means it rides, rather than fights, the industry's consolidation onto Postgres; Multigres removes the scale ceiling that pushed big workloads elsewhere
- Bundle economics: replacing Firebase + Auth0 + S3 + Pusher + Pinecone with one usage-based bill drives high net revenue retention as apps grow
- Cap table doubles as distribution: Stripe (second investment), Salesforce Ventures, Figma Ventures, GIC - partners that embed Supabase into their own ecosystems
- MongoDB (MDB) built a ~$2B-revenue public company on a restrictive license (SSPL) and a non-standard API; Supabase attacks a larger surface with a standard (Postgres) and a stronger community motion
Supabase is priced at roughly 60x third-party-estimated ARR on the assumption that AI-prototype traffic converts into durable production revenue. Most vibe-coded apps are disposable, the gross-margin structure (cloud-infra resale plus a huge free tier) is worse than pure software, and the deepest-pocketed competitors in software - AWS, Google, Microsoft, and now Databricks via Neon - are all aiming at the same workload.
- Valuation risk: $10.5B against ~$170M estimated ARR (~62x) leaves no room for deceleration; the 2x markup in 8 months was momentum-priced during peak AI-infra enthusiasm
- Funnel quality: database launches +600% but many AI-generated projects are demos and weekend apps; paid conversion and 12-month retention of these cohorts are undisclosed
- Channel dependence: a large share of growth is intermediated by AI app builders (Lovable, Bolt, Replit) that could vertically integrate their own backends or be acquired by competitors
- Hyperscaler squeeze: Firebase (GOOGL) and AWS (AMZN) can bundle equivalent services at cost inside existing enterprise agreements; Databricks-Neon gets enterprise distribution Supabase lacks
- Open-source escape hatch: the largest, most sophisticated customers are exactly the ones most capable of self-hosting the Apache 2.0-licensed stack
- Margin ceiling: managed-infra COGS plus free-tier subsidy means even at scale this is unlikely to reach pure-SaaS gross margins, compressing the justified revenue multiple
- No disclosed financials: burn rate, gross margin, and NRR are all unknown - the company leads with developer counts, not dollars
What it is worth
Last priced round + ARR-multiple cross-check vs public DBaaS comps
AI-builder funnel cools and hyperscaler/Neon competition compresses pricing; growth falls under 50% and the multiple compresses to ~10x on ~$300M ARR - a $3-4B outcome, materially below the last round
Growth decelerates to ~70-90% as prototype churn bites; ~$400-500M ARR by 2028 at 15-20x supports roughly the current $8-11B mark - flat-to-modest appreciation from Series F
ARR compounds >100%/yr to ~$800M-$1B by 2028; IPO at 20-30x forward revenue supports $20-30B - the 'next MongoDB with a better funnel' case
Series F (Jun 4 2026): $500M at $10.5B post-money, led by GIC - roughly 62x Sacra-estimated ~$170M ARR, vs MongoDB (MDB) trading at high-single-digit to low-teens EV/revenue. The premium prices sustained triple-digit growth, successful free-to-paid conversion of the AI-app wave, and margin improvement via Multigres-driven density. Private mark, not a traded price; no liquidity assumed. ARR is a third-party estimate, not company-disclosed.
SWOT
Strengths
- Default backend of the AI codegen ecosystem (Bolt, Lovable, Replit, Figma integrations) - distribution it did not have to pay for
- Built on Postgres, the most trusted open-source database - no proprietary lock-in objection at adoption time
- Full backend bundle (auth, storage, realtime, edge functions, vector) with one API - replaces 4-5 point tools
- Nearly 10M developers and a strong open-source brand (top-tier GitHub traction); bottoms-up motion with near-zero CAC
- Deep-pocketed cap table — GIC, Accel, Peak XV, Coatue, Felicis, Y Combinator, Craft, Georgian, plus strategics Stripe, Salesforce Ventures, and Figma Ventures
- Over $1B raised gives a multi-year runway to outspend independent rivals
Weaknesses
- Revenue (~$170M ARR, third-party estimate) is small relative to a $10.5B valuation (~60x est. ARR) - priced for sustained hypergrowth
- Gross margin structurally pressured: reselling cloud infrastructure plus a large free-tier subsidy
- Top-of-funnel concentration in the AI app-builder channel — if Lovable/Bolt-class traffic cools, new-project growth cools with it
- Many AI-generated projects are throwaway prototypes — free-to-paid conversion and retention of 'vibe-coded' apps is unproven at scale
- Enterprise sales motion younger than MongoDB's or the hyperscalers' - large-account go-to-market still maturing
- Postgres single-node scaling ceiling for the largest workloads until Multigres (v0.1 alpha, Jun 2026) matures
Opportunities
- Agentic infrastructure — AI agents provisioning ephemeral per-task databases could expand instance counts by orders of magnitude - agents already deploy the majority of new Supabase databases
- Multigres opens sharded, planet-scale Postgres - the workloads that today default to Vitess/Spanner/Cosmos
- Enterprise graduation wave — thousands of AI-built prototypes maturing into compliant production apps needing paid tiers
- pgvector + AI-native features make Supabase a credible RAG/embedding store, taking budget from dedicated vector DBs (Pinecone, Weaviate)
- International expansion and regulated verticals (HIPAA today; further compliance certifications)
- Eventual IPO candidate: category leader with a recognizable developer brand
Threats
- Hyperscalers bundling equivalents — Google Firebase/AlloyDB (GOOGL), AWS Amplify/Aurora (AMZN), Microsoft Azure (MSFT) can cross-subsidize and undercut
- Databricks-owned Neon (acquired ~$1B, May 2025) attacks the same AI-agent/serverless-Postgres niche with a far larger balance sheet
- AI coding-tool churn — if the app-builder layer consolidates or shifts default backends (e.g., to a house database), Supabase loses part of its funnel
- Open-source self-hosting — the core is Apache 2.0-licensed - large customers can run it themselves and pay nothing
- Late-stage private-market repricing — a 2x-in-8-months markup is vulnerable if growth decelerates before an IPO window
- Managed-Postgres commoditization — Neon (Databricks), Snowflake Postgres (Crunchy Data), PlanetScale-for-Postgres, Render, Railway, Fly.io all pressure pricing
Moats, dependencies & bottlenecks
Moats
Being the pre-wired backend of Bolt/Lovable/Replit/Figma-class tools is today's strongest moat, but it is convention-based, not structural - defaults can be switched
Nearly 10M developers, top-ranked GitHub repo; community trust is slow to build and slow to erode
Apps using auth + storage + realtime + RLS policies + edge functions together face meaningful migration friction, though Postgres portability caps lock-in by design
Operating millions of Postgres instances informs Multigres and fleet automation; density improvements compound into a cost advantage over smaller managed-Postgres rivals
Dependencies
Infrastructure supplier Primary hosting substrate for Supabase cloud; AWS is simultaneously supplier and competitor (Aurora, Amplify, RDS)
Supabase does not control Postgres's roadmap, but the project is healthy, neutral, and Supabase employs contributors
A large share of new-project top-of-funnel arrives through third-party codegen tools' default integrations
CEO Paul Copplestone credits Claude Code and Codex for expanding who can build; the Series F release calls Claude Code the largest usage contributor since the start of 2026
Undisclosed burn funded by over $1B raised; a funding-window freeze before profitability would force a hard pivot to efficiency
Advantages
- Standard-Postgres portability removes the adoption objection every proprietary rival (Firebase, DynamoDB, Cosmos) must argue around
- Whole-backend bundle means one integration for an AI tool to wire in - exactly why codegen tools chose it as default
- Open-source transparency wins developer trust that closed hyperscaler services cannot replicate
- Community-led growth yields near-zero customer acquisition cost at the entry tier
- Speed of shipping (launch-week cadence) against slow-moving hyperscaler product orgs
- War chest (over $1B raised) exceeds every independent DBaaS rival's
Weaknesses
- No disclosed audited financials; all revenue figures are third-party (Sacra) estimates
- Margin structure inferior to pure software (infra resale + free-tier subsidy)
- Dependence on third-party AI tools for top-of-funnel distribution
- Self-hostable open core caps pricing power over the most sophisticated customers
- Enterprise go-to-market still maturing relative to MongoDB/hyperscalers
- Valuation embeds hypergrowth persistence - execution missteps get repriced violently
Bottlenecks
- Free-to-paid conversion of AI-generated projects - the funnel is enormous but of unproven quality
- Postgres vertical-scaling ceiling for the largest tenants until Multigres reaches production maturity (v0.1 alpha as of Jun 2026)
- Enterprise sales capacity and compliance certifications versus hyperscaler incumbency
- Support and reliability at near-10M-developer scale with a lean team - fleet operations are the hidden constraint
- Gross-margin headroom bounded by cloud list pricing and egress economics
Top signals & trends
Top signals
Sovereign-wealth lead plus strategics (Salesforce Ventures, Stripe doubling down, Georgian joining) validate category leadership
Usage growth is still accelerating, not just riding a 2024-25 spike
Proves AI-native positioning but concentrates funnel-quality risk in disposable prototypes
Credible attack on the horizontal-scale ceiling, the historical reason big workloads left Postgres-as-a-service; still alpha-stage
Data-platform giants bought their way into serverless/enterprise Postgres for AI agents - the competitive field got much richer
Signals both extraordinary demand and momentum pricing typical of late-cycle AI-infra markups
Company leads with developer and database counts, not dollars; ~62x estimated ARR requires strong conversion
Trends
Strongly positive · Every AI-built app needs a backend; Supabase is the wired-in default for the leading tools
Postgres is eating the relational market; Supabase is its most visible commercial standard-bearer alongside Neon
Positive but early · Per-agent ephemeral databases could multiply instance counts; also the explicit thesis behind Databricks buying Neon
AWS/Google/Microsoft can price equivalent services at cost within committed-spend agreements
Supabase captures RAG workloads without customers buying a dedicated vector DB
Marks doubled in 8 months across the category; a sentiment turn reprices the whole cohort
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 cloud substrate for Supabase's managed fleet
Core engine; Supabase employs contributors and builds on ecosystem projects (PostgREST, pgvector)
Edge/network layer relevant to edge-function and CDN-adjacent workloads
AI app builder; Supabase is its default backend
AI codegen tool with native Supabase integration
AI coding platform routing generated apps to Supabase
Named Supabase user/integrator (Figma Make ecosystem); Figma Ventures invested in the Series E
Long tail of ~10M developers and 250,000+ customers Bottoms-up usage-based revenue base
The incumbent Supabase was built as an open alternative to; proprietary NoSQL model vs Supabase's open Postgres
Supplier and competitor; can bundle at cost inside enterprise agreements
Enterprise distribution via Azure commitments and GitHub/Copilot adjacency
Serverless Postgres built for AI agents (>80% of its databases agent-created), acquired by Databricks for ~$1B in May 2025 - the most direct architectural rival, now with enterprise distribution
The reference public DBaaS comp; document model vs Postgres, stronger enterprise sales machine
Vitess-based MySQL (and since 2025, Postgres) at scale; competes for the high-end workloads Multigres targets
Smaller open-source BaaS rivals competing for the same AI-tool default slots
Acquired Crunchy Data for ~$250M (Jun 2025) to enter managed Postgres; targets enterprise data-platform buyers