Pinecone
Usage-based cloud DBaaS: serverless consumption pricing (reads/writes/storage), Builder tier at $20/mo (launched May 2026), Standard/Enterprise tiers, Dedicated Read Nodes for high-throughput workloads (public preview Dec 2025, GA Apr 2026), plus a marketplace of knowledge applications; sold self-serve and via enterprise sales and AWS/GCP/Azure marketplaces
Only the Apr 2023 Series B ($750M) is a disclosed priced mark. Seed and Series A valuations were never disclosed - those points are rough unverified estimates included for chart shape only. The 2025 ~$2B figure is press speculation around a reported sale exploration, not a transaction.
Earnings, margins, COGS & capex
Pinecone has raised $138M across three rounds ($10M seed led by Wing VC in 2021, $28M Series A led by Menlo Ventures with Tiger Global in 2022, $100M Series B in Apr 2023 led by Andreessen Horowitz with ICONIQ Growth, Menlo Ventures and Wing VC at a $750M valuation). It has never disclosed revenue; third-party estimates put ARR in the mid-$20Ms as of 2024, up sharply from 2023 but small relative to the valuation and to the AI-infra hype cycle it rode. The Jan 2024 shift from pod-based to serverless pricing improved unit economics and win rates but compressed revenue per customer, and flagship customer Notion migrated its 10B+ vector workload to Turbopuffer in Oct 2024. In Aug 2025 The Information and Calcalist reported Pinecone had held early talks with bankers about a sale or strategic options, naming Oracle, IBM, MongoDB and Snowflake as possible buyers; the company then installed Ash Ashutosh as CEO (Sep 2025) and pivoted upmarket toward agent infrastructure (Nexus, KnowQL, Marketplace, Dedicated Read Nodes) in late 2025-2026. No new priced round or acquisition had been announced as of Jul 2026.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~40¢ is cost of goods and ~60¢ operating expense, leaving ~0¢ of operating profit.
Revenue trend
Margins
Likely improved after the Jan 2024 serverless re-architecture (usage-metered, multi-tenant, storage-compute separation); Dedicated Read Nodes shift high-throughput reads to dedicated capacity with company-claimed cost cuts of up to 97% vs on-demand for suitable workloads
Presumed negative; ~130 employees (third-party est., 2026), enterprise GTM build-out under new CEO
COGS structure
Cloud compute, storage and network rented from AWS, GCP and Azure to run customers' indexes; serverless architecture separates storage from compute to cut idle-capacity cost. Pinecone carries the infrastructure bill inside its usage pricing, so its gross margin is a direct spread over hyperscaler list/committed pricing.
Capex
Minimal owned capex - asset-light SaaS on rented hyperscaler infrastructure; investment flows to R&D (core engine work, Nexus knowledge engine, KnowQL) and GTM headcount rather than physical assets.
Latest earnings
Not applicable
None public. Management framing (Sep 2025-2026): independent path, growth and category expansion under new CEO, focus on enterprise scale-out and the Nexus agent-knowledge platform
- Last priced valuation
- $750M (Series B, Apr 2023, a16z lead)
- Total raised
- $138M (Seed $10M + Series A $28M + Series B $100M)
- ARR (third-party est.)
- ~$26.6M (2024, getlatka; unverified - company does not disclose)
- Customers
- ~4,000 (third-party estimate, 2025-2026); 1,500 paying customers company-disclosed at Series B (Apr 2023)
- Headcount
- ~130 (2026, third-party est.)
- Founded / HQ
- 2019, Edo Liberty (ex-Amazon AI Labs / SageMaker); New York HQ with Tel Aviv and San Francisco offices
Growth drivers
- RAG and AI-agent adoption expanding the addressable retrieval-infrastructure market
- Serverless + $20/mo Builder tier (May 2026) widening the top of funnel; ~4,000 customers per third-party estimate
- Move upmarket — Dedicated Read Nodes and enterprise GTM under CEO Ash Ashutosh (three-time infrastructure founder incl. Actifio, acquired by Google; ex-Google)
- Platform expansion beyond raw vector search — Pinecone Nexus knowledge engine (early access May 2026, public preview Jul 2026), KnowQL agentic query language, Marketplace of 90+ knowledge applications
- Geographic expansion — first Asia serverless region (AWS Singapore, May 2026) and Frankfurt region for Central Europe (2026)
- Native full-text/hybrid search closing the feature gap versus Elastic/OpenSearch-style incumbents
Bull & bear
Pinecone owns the best-known brand in a layer every AI agent needs, has re-architected to competitive serverless economics, and is moving up-stack into agent knowledge infrastructure just as agentic AI spend inflects; even the bear exit is a strategic sale, potentially above the last round.
- Agent retrieval/memory is a structural, growing workload - Nexus and KnowQL position Pinecone above commodity ANN search where bundlers cannot easily follow
- Serverless + $20/mo Builder tier funnel gives cheap distribution into tomorrow's enterprise accounts; Dedicated Read Nodes give an upsell path with company-claimed cost cuts of up to 97% vs on-demand for high-throughput workloads
- New CEO Ash Ashutosh (three-time founder; built and sold Actifio to Google) is a proven enterprise operator - the exact gap in Pinecone's founder-led GTM
- Named strategic interest (Oracle, IBM, MongoDB, Snowflake per The Information/Calcalist) supports the equity: sale speculation referenced north of $2B, roughly 3x the last mark, though never priced
- Purpose-built engines historically win the performance tier of a data category (Snowflake vs bundled warehouses; MongoDB vs bundled document stores) - the same pattern can hold for vector-scale retrieval
Vector search is a feature, not a company: every database Pinecone's customers already use now ships it, ARR is small and reportedly decelerating, the flagship customer walked to a cheaper rival, and the board explored a sale - the classic profile of a squeezed mid-stage asset marked above its fundamentals.
- Estimated ~$27M ARR against a $750M mark is ~28x estimated ARR - a multiple that assumes durable hypergrowth precisely as growth reportedly slowed
- pgvector, MongoDB, Elastic, Redis, Snowflake, Databricks and AWS/GCP/Azure bundle vector search at marginal cost zero to them, collapsing standalone pricing power
- Notion's Oct 2024 migration to Turbopuffer shows the sophisticated-customer failure mode: at scale, users optimize cost by moving to object-storage-native or in-house solutions - and Turbopuffer's estimated revenue (~$100M annualized, Sacra) now reportedly exceeds Pinecone's
- The Jan 2024 serverless repricing was defensively necessary but cut revenue per existing customer - growth must outrun self-inflicted ARPU compression
- Sale exploration plus the founder stepping back within one year signals investors managing toward exit, not conviction in standalone escape velocity; if strategics only bid near or below the last round, common and late-preference holders are squeezed
- Long-context models and model-resident memory could structurally shrink the external retrieval layer over a 3-5 year horizon
What it is worth
Last priced round anchored, cross-checked against ARR multiples and reported strategic-interest pricing; no public comps apply directly since revenue is undisclosed
~$250-500M
down-round or distressed sale if commoditization compresses growth further and strategics bid opportunistically; preference stack would absorb much of the loss ahead of common
~$700M-1.1B
roughly the last round to a modest premium, via strategic sale or flat-ish extension round, reflecting real brand/technology value against decelerating standalone metrics
~$1.5-2.5B in a competitive strategic sale (Oracle/IBM/MDB/SNOW-class buyer paying for the category-leading retrieval brand and Nexus platform), or a marked-up round if agent-platform ARR inflects
Last mark: $750M (Series B, Apr 2023). Against third-party-estimated ~$27M ARR that is roughly 28x estimated ARR - defensible only with reaccelerating growth or a strategic premium. H2 2025 sale chatter referenced speculation north of $2B, which reads as aspirational banker pricing on category scarcity (best-known pure-play managed vector DB), not a struck price. Public bundlers trade at roughly 6-15x forward revenue, implying a fundamentals-only value well below the last round; the gap is scarcity/strategic value.
SWOT
Strengths
- Category-defining brand: 'Pinecone' is near-synonymous with vector database among AI developers
- Purpose-built serverless architecture with strong recall/latency/cost trade-offs at billion-to-multi-billion-vector scale
- Blue-chip investor base (a16z, ICONIQ, Menlo, Wing, Tiger Global) and credible new operator CEO with enterprise GTM and exit experience
- Fully managed, zero-ops posture appeals to teams that do not want to run open-source infra
- 2025-2026 platform layer (Dedicated Read Nodes, Nexus, KnowQL, Marketplace) differentiates above commodity ANN search
Weaknesses
- Estimated ARR (~mid-$20Ms) is small versus the $750M mark and versus platform rivals bundling vector search 'for free'
- Closed-source in a category where open-source (Milvus, Qdrant, Weaviate, Chroma, pgvector) dominates developer mindshare
- Marquee-customer churn (Notion migrated its 10B+ vector workload to Turbopuffer, Oct 2024) and serverless repricing that cut revenue per customer
- Single-product concentration in a feature that is being absorbed into general-purpose databases
- Strategic uncertainty: 2025 banker-led sale exploration and founder-to-CEO handoff signal board-level pressure
Opportunities
- Agentic AI wave: agents need persistent, queryable knowledge/memory - Nexus/KnowQL target exactly this layer
- Enterprise upmarket motion (compliance, dedicated capacity, hybrid search) where managed service beats DIY open source
- APAC expansion (Singapore region) plus Frankfurt/Central Europe and further multi-cloud footprint
- Marketplace / knowledge-application ecosystem (free at launch) could add a take-rate revenue stream if commercialized
- Strategic acquisition by Oracle, IBM, MongoDB or Snowflake (all named in press reports) remains a realistic exit at a premium to the last round
Threats
- Commoditization — pgvector (Postgres), MongoDB Atlas Vector Search, Elastic, Redis, OpenSearch, Snowflake Cortex, Databricks and all three hyperscalers ship vector search inside databases customers already pay for
- Object-storage-native and open-source specialist rivals (Turbopuffer, Qdrant, Weaviate, Zilliz, Chroma) competing hard on price - Turbopuffer reportedly reached ~$100M annualized revenue by Mar 2026 (Sacra est.) on Cursor- and Notion-class workloads
- Retrieval-architecture shift risk — very long context windows and model-native memory could shrink the standalone vector-DB layer
- AI-infra funding sentiment cooling for sub-scale ARR at rich multiples, complicating a next round at or above $750M
- Customer concentration among AI startups whose own mortality is high
Moats, dependencies & bottlenecks
Moats
moderate-strong eroding at the edges Default name developers reach for in 'vector database', but bundlers reduce how often that search happens at all
Storage-compute separation and Dedicated Read Nodes give real cost/latency leverage at billion-vector scale; less relevant at small scale where bundled options suffice, and Turbopuffer contests the object-storage cost curve directly
Re-embedding and re-indexing is friction, but vectors are regenerable from source data - Notion's exit proved migration is feasible even at 10B+ vectors
Early access May 2026, public preview Jul 2026; if agents standardize on KnowQL it becomes a real moat, but adoption is unproven
Classic infrastructure SaaS - little cross-customer network effect; Marketplace is an attempt to create one
Dependencies
infrastructure supplier + channel + competitor Primary hosting substrate (incl. the Singapore ap-southeast-1 region) and marketplace channel, while AWS sells competing OpenSearch/Aurora pgvector/Bedrock Knowledge Bases
infrastructure suppliers + competitors Multi-cloud hosting; both bundle vector search (Vertex AI Vector Search, Azure AI Search)
Pinecone stores what these models embed; model-native retrieval/memory features could disintermediate the external store
Integrations drive developer adoption; frameworks are neutral and list every competitor alongside
Last priced raise Apr 2023; a future raise must clear a $750M bar in a cooler AI-infra market, or the company sells
Advantages
- First-mover brand and largest managed-vector-DB mindshare
- Proven performance at billion-to-multi-billion-vector scale with zero-ops managed service
- Cost-engineered serverless stack plus Dedicated Read Nodes for high-QPS production workloads
- Fresh agent-era product surface (Nexus knowledge engine, KnowQL, 90+ app Marketplace) launched May-Jul 2026
- Experienced operator CEO with a prior strategic exit (Actifio to Google) and deep storage/data-infra background
Weaknesses
- Sub-scale disclosed-nothing financials: estimated ARR in the mid-$20Ms with reported deceleration
- Closed-source positioning against a largely open-source competitive field
- Feature-vs-product risk as vector search commoditizes inside incumbent databases
- Loss of flagship customer Notion (to Turbopuffer, Oct 2024) and ARPU hit from serverless repricing
- Strategic limbo: 2025 sale exploration unresolved as of Jul 2026
Bottlenecks
- Enterprise sales capacity — the pivot upmarket under the new CEO is young and unproven at scale (~20 quota-carrying reps per third-party estimate)
- Differentiation velocity — must ship platform value (Nexus/KnowQL) faster than hyperscalers and incumbents absorb baseline vector features
- Capital runway vs growth spend without a new priced round since Apr 2023
- Talent retention through founder transition and sale speculation
Top signals & trends
Top signals
bearish for standalone thesis, bullish for near-term exit · Oracle, IBM, MongoDB, Snowflake named as potential buyers; no deal announced as of Jul 2026
Classic scale-up handoff to an enterprise GTM operator; also consistent with board grooming the asset for outcome
Signals investment in an independent up-stack strategy, not a wind-down
Largest-scale customers can engineer their way off Pinecone when the bill gets big
Three-plus years without a mark-up during the biggest AI-infra funding boom on record
Category compression from all sides - hyperscalers, Postgres ecosystem, MongoDB, Elastic, Redis, Snowflake, Databricks
Trends
Expands the workload Pinecone serves beyond one-shot RAG; core premise of Nexus/KnowQL
The single biggest structural headwind; standalone vector DB TAM narrows to the high-scale performance tier
Turbopuffer's rise (Cursor, Notion; ~$100M annualized revenue per Sacra est., Mar 2026) attacks exactly the high-scale tier Pinecone retreats to
negative (medium-term) · Reduces some retrieval calls, though cost/latency still favor external retrieval for large corpora
positive for exit · Strategics with public-market AI narratives (ORCL, IBM, MDB, SNOW) have reported appetite for retrieval assets
Pinecone moved early (Jan 2024) - better economics and funnel, but ARPU compression
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 incl. Singapore region; also marketplace channel
Multi-cloud hosting option
Multi-cloud hosting option
Embedding and LLM providers whose outputs Pinecone stores and serves
Cited customer (company Series B disclosure, Apr 2023)
Cited customer (company Series B disclosure)
Cited customer
Cited customer
Former flagship customer; migrated its 10B+ vector workload to Turbopuffer in Oct 2024 - a key bear datapoint
~4,000 customers per third-party estimate (2025-2026); 1,500 paying customers company-disclosed at Series B (2023)
Bundles vector search into the document database many AI apps already use; also a reported potential acquirer
ESRE/vector search on the dominant enterprise search install base; hybrid keyword+vector strength
OpenSearch vector engine, Aurora/RDS pgvector, Bedrock Knowledge Bases - default choice inside AWS accounts
Azure AI Search bundled with the OpenAI enterprise franchise
Vertex AI Vector Search (ScaNN lineage), AlloyDB pgvector
Vector + hybrid search inside the enterprise data cloud; reported potential acquirer
Vector search in the core RDBMS for the Oracle installed base; reported potential acquirer
Enterprise AI stack with vector capabilities; reported potential acquirer
Vector search native to the lakehouse; massive AI budget adjacency
'Just use Postgres' is the highest-volume competitive loss at the low end
Object-storage-native vector search on a radical cost curve; won Notion (Oct 2024) and Cursor; ~$100M annualized revenue per Sacra estimate (Mar 2026)
Open-source Rust vector DB with managed cloud; aggressive on price/performance
Open-source vector DB, ~$68M raised, strong hybrid search and modules
Steward of Milvus, among the most-deployed open-source vector DBs; ~$113M+ raised
Developer-favorite open-source embedded store for prototypes moving upmarket
Vector similarity search on the ubiquitous in-memory store