Qdrant
Open-source core (Apache 2.0) + commercial managed cloud (Qdrant Cloud, usage-based) + enterprise self-hosted licenses (hybrid/private cloud, SSO/RBAC, tiered multitenancy); land via OSS developers, expand to paid enterprise deployments
Round dates, amounts, and leads are verified facts. NO Qdrant post-money valuation has ever been disclosed - every valuationB above is a stage-typical inference (seed ~4-5x raise, Series A ~4-5x raise, Series B midpoint of the dossier's $250-500M working range), anchored loosely to verified comps Pinecone ($750M, 2023) and Weaviate ($200M, 2023). Treat the trail shape (steady step-ups) as directional, not the absolute levels.
Earnings, margins, COGS & capex
Qdrant is private and discloses no audited financials. Verified facts: $50M Series B (announced 2026-03-12) led by AVP with Bosch Ventures, Unusual Ventures, Spark Capital and 42CAP participating, bringing total funding to $87.8M (after a $7.5M seed led by Unusual Ventures, Apr 2023, and a $28M Series A led by Spark Capital, Jan 2024). Adoption proxies are strong: 250M+ downloads across packages and 29K GitHub stars (both per the Series B announcement), ~89 employees (2024-2025 third-party estimate). Revenue is not disclosed; a third-party estimate pegged ~$9.2M ARR in 2024 (Getlatka, unverified and low-confidence). Monetization is early relative to adoption - classic OSS-infrastructure conversion-funnel economics.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~35¢ is cost of goods and ~0¢ operating expense, leaving ~65¢ of operating profit.
Revenue trend
Margins
n/a
n/a
COGS structure
Primarily cloud compute/storage for Qdrant Cloud (rented from AWS/GCP/Azure), plus support engineering. Rust single-binary architecture and aggressive quantization (binary/1.5-bit/2-bit) are explicitly aimed at lowering per-vector serving cost versus JVM-based or managed-only rivals.
Capex
Minimal owned capex; infrastructure is hyperscaler-rented. Series B proceeds earmarked for engineering/product expansion, enterprise offerings, and global go-to-market - opex, not capex.
Latest earnings
n/a
n/a; stated Series B use of funds: expand engineering and product teams, strengthen enterprise offerings, scale global operations and open-source adoption
- Series B (Mar 2026)
- $50M led by AVP; Bosch Ventures, Unusual Ventures, Spark Capital, 42CAP participating
- Total funding
- $87.8M (pre-seed 2021 + $7.5M seed Apr 2023 + $28M Series A Jan 2024 + $50M Series B Mar 2026)
- Downloads
- 250M+ across packages (per Series B announcement, 2026-03)
- GitHub stars
- 29K (per Series B announcement, 2026-03)
- Headcount
- ~89 (third-party estimate, ZoomInfo/Getlatka; not company-disclosed)
- Named customers
- Canva, HubSpot, Tripadvisor, OpenTable, Bazaarvoice, Roche, Bosch, Lyzr, Deutsche Telekom, Flipkart
Growth drivers
- Agentic AI and RAG moving from prototypes to production, requiring low-latency retrieval at billion-vector scale
- Composable vector search strategy — dense + sparse vectors, metadata filtering, multi-vector, custom scoring as query-time primitives
- Enterprise feature buildout (SSO, RBAC, tiered multitenancy) unlocking regulated buyers (Roche, Bosch, Deutsche Telekom)
- New deployment surfaces: Qdrant Cloud Inference, Qdrant Edge (beta), GPU-accelerated HNSW indexing
- Marquee logo flywheel — Canva, HubSpot (Breeze AI), Tripadvisor (AI features over 1B+ reviews and opinions, with a Qdrant case study citing 2-3x revenue uplift on AI-powered experiences), OpenTable
Bull & bear
Qdrant is the performance-and-cost leader in open-source vector search at the exact moment agentic AI makes retrieval a core production workload; its Rust architecture, edge story, and enterprise logos position it to become the default independent vector layer - a credible acquisition target or IPO candidate if ARR compounds.
- Category tailwind is real and durable: production RAG and agent memory require dedicated, filterable, billion-scale retrieval that bolt-on database features handle poorly at scale
- Proof of enterprise pull: Tripadvisor (case study cites 2-3x revenue uplift on AI-powered features), HubSpot Breeze AI, OpenTable, Canva, and Deutsche Telekom chose Qdrant after technical evaluation - production wins, not free-tier vanity logos
- Rust + quantization gives a structural TCO advantage - in infra, the lowest-cost-per-query engine tends to win the workload over time
- Qdrant Edge opens an underserved segment (on-device/embedded vector search for automotive, robotics, consumer hardware) where hyperscaler managed services cannot follow; Bosch Ventures is a strategic validator
- $87.8M raised with a disciplined burn profile (~89 people, third-party estimate) means multi-year runway; strategic acquirer interest (Databricks bought Neon, Snowflake bought Crunchy Data, both 2025) sets active M&A comps for data-infra assets
Qdrant is a small, revenue-opaque vendor in a rapidly commoditizing category: vector search is becoming a feature of every database and cloud, OSS conversion is unproven, and the standalone vector-DB category may not sustain multiple independent winners.
- Vector search is a feature, not a product: pgvector, MongoDB, Elastic, Redis, and all three hyperscalers ship it natively - the default buyer path is 'use what you already have'
- 250M+ downloads against an estimated single-digit-millions ARR (2024, unverified) implies weak monetization of the funnel so far; enterprise sales motion is young
- Pinecone raised $138M at a $750M valuation (2023) and Zilliz raised $113M - deeper war chests; a price war in managed vector search compresses everyone's margins, and Pinecone reportedly weighing a sale (2025) signals category-wide monetization strain
- Apache 2.0 licensing means AWS/GCP/Azure could offer managed Qdrant-compatible services without paying Qdrant - the exact dynamic that forced Elastic and MongoDB into license changes
- Long-context LLMs and integrated retrieval inside model providers (OpenAI, Anthropic file/retrieval APIs) could absorb a meaningful share of simple RAG use cases, shrinking the independent retrieval layer
- No disclosed valuation, revenue, or margins - private-market pricing risk is unquantifiable; a flat/down Series C is plausible if ARR lags the AI-infra hype cycle
What it is worth
Stage/round-based triangulation + private SaaS comps (no disclosed post-money; no public financials)
Commoditization outruns monetization: ARR stalls below $20M, flat/down Series C in a tighter 2027 funding market, endgame is an acqui-hire or sub-$300M trade sale as vector search folds into general-purpose databases
Solid niche winner: $25-40M ARR by 2028, Series C at a modest step-up ($400-700M), category consolidates to 2-3 independents with Qdrant among them
ARR compounds toward $50M+ by 2028 on agentic-AI retrieval demand and edge wins; strategic acquisition (hyperscaler, Databricks-class platform, or Bosch-adjacent industrial) or late-stage round at $1B+; OSS funnel converts at improving rates
Series B post-money undisclosed. Working-range inference only: a $50M Series B with full insider participation typically prices at roughly $250M-$500M post-money for AI-infra assets of July 2026; treat as an assumption, not a fact. Comps (verified): Pinecone $750M post-money (Apr 2023, $100M Series B led by a16z), Weaviate $200M (Apr 2023, $50M Series B led by Index Ventures). If the unverified ~$9.2M ARR (2024, Getlatka) grew 2-3x by 2026, an implied 10-25x forward-ARR multiple would be consistent with 2026 AI-infra private pricing. All revenue-linked math here is explicitly estimate-on-estimate.
SWOT
Strengths
- Rust-native engine with credible performance edge — predictable low tail latency at billion-vector scale, from edge devices to the Aurora supercomputer at Argonne National Laboratory (deployment cited in the Series B announcement)
- Strong OSS distribution — 250M+ downloads, 29K GitHub stars, active Discord community, 35+ new integrations added in 2025 (incl. official n8n node, per Qdrant 2025 recap)
- Enterprise logo validation across verticals — Canva, HubSpot, Tripadvisor, OpenTable, Roche, Bosch, Deutsche Telekom (2M+ conversations via Frag Magenta), Flipkart (real-time fraud detection)
- Fresh $50M balance sheet (Mar 2026) with strategic investor Bosch Ventures signaling industrial/edge use cases
- Cost-efficiency features (binary/1.5-bit/2-bit quantization, inline storage, GPU-accelerated indexing) as a differentiated TCO pitch
Weaknesses
- Revenue undisclosed and likely small (~$9.2M ARR 2024 per unverified Getlatka estimate) relative to 250M+ downloads - OSS-to-paid conversion still unproven at scale
- Sub-100 headcount competing against hyperscalers and incumbents with thousands of database engineers
- Single-product company — vector search only, while buyers increasingly want one platform for OLTP + search + vectors
- Apache 2.0 license permits hyperscaler re-hosting (the Elasticsearch/MongoDB playbook risk in reverse - no license moat)
- Valuation and unit economics opaque; no audited financials
Opportunities
- Agentic AI memory layer — every production agent needs persistent, filterable semantic retrieval - a structurally growing workload
- Qdrant Edge: on-device vector search for robotics/automotive/IoT (Bosch Ventures participation points here) - a segment hyperscalers serve poorly
- Cloud Inference bundling (embedding + retrieval in one platform) raises ASP and stickiness
- Enterprise displacement of DIY pgvector deployments as scale/latency requirements outgrow Postgres
- European data-sovereignty demand favors an EU-headquartered, self-hostable vendor
Threats
- Feature convergence — Postgres pgvector, MongoDB Atlas Vector Search, Elastic, Redis, OpenSearch, and every hyperscaler now ship 'good enough' vector search inside databases customers already run - IDC and Omdia analysts note the market increasingly favors integrated vector capabilities
- Pinecone, Weaviate, Zilliz/Milvus, Chroma, and Turbopuffer compete for the same dedicated-vector-DB budget; category may consolidate (Pinecone reportedly weighed a sale, Calcalist 2025)
- Embedding-model and retrieval-paradigm shifts (e.g., late-interaction, long-context LLMs reducing retrieval need) could shrink the standalone vector-DB TAM
- AI infra funding cycle risk — if the 2026-2027 window tightens, sub-scale infra vendors face down-rounds or forced M&A
- Hyperscalers can bundle vector search at zero marginal price into existing cloud commitments
Moats, dependencies & bottlenecks
Moats
moderate-strong Benchmark leadership is real but must be continuously re-earned; rivals re-platform (e.g., Pinecone serverless rewrite)
durable while maintained 29K GitHub stars, 250M+ downloads, 35+ new integrations in 2025; default choice in many LangChain/LlamaIndex/n8n tutorials
grows with enterprise penetration Re-indexing billions of vectors plus rewriting filtering/scoring logic is painful; tiered multitenancy deepens lock-in
Qdrant Edge (beta) is differentiated vs cloud-only rivals but unproven commercially
Canva/HubSpot/Tripadvisor/Deutsche Telekom references matter in enterprise procurement
Dependencies
Google Cloud GOOGL, Microsoft Azure MSFT) infrastructure + channel Qdrant Cloud runs on and sells through the same clouds that compete with it via native vector services
Anthropic, Cohere, open-source models) complementary technology Vector DB demand is derivative of embedding-based architectures remaining the dominant retrieval paradigm
Large share of developer acquisition flows through framework integrations Qdrant does not control
Loss-making growth model depends on Series C availability circa 2027-2028 absent profitability
GPU-accelerated HNSW indexing benefits from GPU availability; vendor-agnostic (Vulkan-based) design mitigates
Advantages
- Rust single-binary architecture — low memory footprint, predictable tail latency, no JVM/garbage-collection overhead
- Aggressive cost-reduction features (binary/1.5-bit/2-bit quantization, inline storage) that directly lower customer cloud bills
- Deployment breadth rivals lack: managed cloud, hybrid, private cloud, self-hosted OSS, and edge devices
- EU roots (Berlin HQ alongside New York) + self-hosting = data-sovereignty story for European and regulated enterprises
- Strategic investor set (Bosch Ventures) opening industrial/automotive edge channels
Weaknesses
- No disclosed revenue, margins, or valuation - opaque fundamentals
- Sub-scale headcount versus incumbent database vendors and hyperscalers
- Single-category product exposure to vector-search commoditization
- Permissive Apache 2.0 license enables competitive re-hosting without revenue share
- Brand recognition still trails Pinecone in US enterprise mindshare
Bottlenecks
- OSS-to-paid conversion rate - the core economic engine, still unproven at disclosed scale
- Enterprise go-to-market capacity — ~89 employees (est.) limits concurrent enterprise sales cycles; Series B explicitly targets team and go-to-market expansion
- Category education — convincing buyers a dedicated vector DB beats the 'free' feature inside their existing database
- Multi-cloud managed-service operational maturity versus hyperscaler-native reliability expectations
Top signals & trends
Top signals
Insider re-participation plus a new strategic (Bosch Ventures) is a quality signal
Published case studies with quantified outcomes (Tripadvisor 2-3x revenue uplift; Flipkart fraud detection from 9 hours to under 1 minute), not free-tier vanity logos
Common for EU rounds, but prevents mark-to-market; any post-money range is inference, not fact
Category commoditization pressure is the central bear signal
Shipping cadence consistent with a technically strong, focused team
Sets acquirer appetite and comps for independent data-infra assets, but also signals standalone vector-DB monetization strain
Trends
strong positive · Agent memory and tool-augmented retrieval are recurring, latency-sensitive vector workloads - Qdrant's core pitch
pgvector/MongoDB/Elastic absorb the low end; dedicated engines must win on scale, latency, and cost
Quantization and Rust efficiency align with 2026's enterprise focus on AI unit economics
Qdrant Edge is early but aligned with automotive/robotics investment (Bosch)
Shrinks trivial use cases but production systems still need retrieval for cost, freshness, and permissioning
Favors self-hostable, EU-rooted infrastructure vendors
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 Qdrant Cloud; also a competitor
Qdrant Cloud region provider and marketplace channel
Cloud provider and marketplace channel
GPUs for GPU-accelerated HNSW index building (vendor-agnostic Vulkan implementation)
Breeze AI assistant retrieval layer (published Qdrant case study)
AI-powered travel features; case study cites 2-3x revenue uplift
Named customer in Series B announcement; AI dining discovery
AI-powered features across the design platform
Enterprise/regulated deployment (named in Series B announcement)
Customer and strategic investor via Bosch Ventures
Frag Magenta multi-agent platform, 2M+ conversations across 10 European subsidiaries (case study)
Real-time multimodal fraud/similarity search in Trust & Safety; detection time cut from 9 hours to under 1 minute (case study)
UGC/review intelligence workloads (named in Series B announcement)
Best-funded dedicated vector DB ($138M raised; $750M valuation, Apr 2023); managed-only, strong US enterprise brand; reportedly weighed a sale in 2025 (Calcalist)
Open-source vector DB peer (Amsterdam), ~$68M raised, $200M valuation at Series B (Apr 2023); similar OSS+cloud model
Commercial steward of Milvus (LF AI project); $113M raised through 2022; strong at very large scale. China-origin roots, US-headquartered (Silicon Valley) - context only, no buy/own framing
Developer-first OSS vector store popular in prototyping; moving upmarket
Object-storage-native serverless vector search; wins cost-sensitive high-scale workloads (used by Cursor, Notion)
Atlas Vector Search bundles vectors into the operational DB enterprises already run
ESRE/vector search on the largest installed search base; hybrid BM25+vector strength
OpenSearch vector engine, Aurora/RDS pgvector, Bedrock Knowledge Bases - default for AWS-committed buyers
Azure AI Search + Cosmos DB vector - bundled into Azure OpenAI deployments
Vertex AI Vector Search (ScaNN lineage), AlloyDB vectors
Redis vector similarity on ubiquitous caching layer; strong for low-latency hybrid workloads
Database 23ai AI Vector Search targeting its enterprise installed base
Mosaic AI Vector Search bundled with the lakehouse; acquired Neon (Postgres/pgvector) in 2025
The 'good enough free' option absorbing the long tail of vector workloads