Databricks
Consumption/usage-based SaaS (DBUs) sold across the three hyperscalers; land-and-expand with >140% net retention, layered AI/agent monetization (Mosaic AI, Agent Bricks, Genie, Lakebase).
Priced primary rounds only; Series A-D valuations were never disclosed (omitted). Jun 2026 $165-175B is an in-talks target (not a closed round) — see note, excluded as a point.
The thesis on this name
State of AI Compute
Databricks is the leading data + AI 'lakehouse' platform — unifying data engineering, warehousing, ML and generative-AI agents on open formats (Delta/Iceberg, Unity Catalog, Mosaic AI) across AWS, Azure and GCP.
State of Enterprise AI SaaS
Private reference for the AI-native consumption data platform — $5.4B run-rate +65% YoY, AI products $1.4B annualized, FCF-positive, raised at $134B (Feb-2026), reportedly now raising at $165-175B with an IPO expected ~Q4 2026 (Databricks newsroom + The Information, Jun 2026). The purest expression of the consumption thesis: open-format Lakehouse + agentic tooling that grows with AI workloads, not seats. Not directly investable on public markets yet; play it via the IPO calendar and the listed read-throughs (SNOW competitive set, MSFT/GOOGL/NVDA partners).
State of Frontier AI
The strongest fundamental profile in the entire private cohort and the cleanest consumption-software compounder: $5.4B revenue run-rate growing >65% YoY, AI product revenue >$1.4B (~26% of total), and — uniquely — FREE-CASH-FLOW POSITIVE while growing. The data+AI platform sits underneath every enterprise AI deployment regardless of which model wins, making it a value-chain toll rather than a model bet. Conviction is high because the durability is proven by the cash flows, not promised.
Earnings, margins, COGS & capex
Private, pre-IPO. Run-rate ARR ~$6.9B growing >80% YoY (Jun 2026), accelerating from 55-65% — a rare re-acceleration at this scale. AI products ~$1.7B run-rate (~25% of revenue) and SQL/data-warehousing ~$1.5B. FCF-positive FY25 with >140% net retention, but gross margin slipped to ~74% as AI agents drive far more compute than human queries (each point of margin ≈ ~$69M/yr to cloud providers per the CEO).
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~26¢ is cost of goods and ~62¢ operating expense, leaving ~12¢ of operating profit (~140¢ net).
Revenue trend
Margins
Declining — was >80%; CEO signals further decline as agentic compute grows
Stable, elite for scale
Positive — the only profitable name in the AI IPO pipeline
COGS structure
Dominated by cloud-infrastructure compute purchased from AWS/Azure/GCP (pass-through). Rising agent-driven query volume is the direct cause of gross-margin compression to ~74%.
Capex
Minimal owned-infra capex (asset-light, rides hyperscalers). Spend is concentrated in R&D, sales, and AI acquisitions (e.g. MosaicML, Tabular, Neon, BladeBridge).
Latest earnings
N/A (private); growth re-accelerated vs prior >65% disclosure — directionally a 'beat' vs expectations
No formal guidance issued; CEO Ghodsi guides margins to 'decline further' on agent compute, and reiterates 'IPO-ready' with S-1 reportedly targeted ~Q3-Q4 2026.
- Customers >$1M ARR
- >800 (Feb 2026)
- Customers >$10M ARR
- >70 (Feb 2026)
- Total customers
- >20,000 organizations; ~70% of Fortune 500
- Net dollar retention
- >140%
- AI revenue run-rate
- ~$1.7B (Jun 2026)
Growth drivers
- AI/agent workloads (Mosaic AI, Agent Bricks, Genie) — ~$1.7B run-rate, fastest-growing
- Data warehousing (Databricks SQL) taking share from Snowflake — ~$1.5B run-rate
- Open lakehouse + Unity Catalog governance as enterprise standard — Iceberg/Delta interoperability (Tabular acquisition)
- Lakebase (Postgres/Neon) extending into transactional/operational databases
- Land-and-expand at >140% NDR; >70% of Fortune 500 as customers
Bull & bear
The default data+AI platform of the agentic era — re-accelerating to >80% growth at $6.9B, FCF-positive, open, multi-cloud, and taking the warehouse from Snowflake. The IPO is a re-rating catalyst, not a risk.
- Growth re-accelerated (55%→65%→80%) at a scale where peers decelerate — agentic AI is a genuine new demand wave, not a one-off
- Only profitable AI-pipeline IPO candidate; >140% NDR and ~70% of F500 give durable, expanding revenue
- Open lakehouse (Delta+Iceberg, Unity Catalog) is winning the standards war — lower lock-in than Snowflake/proprietary stacks
- Warehouse + AI + Lakebase widen the TAM well beyond core analytics; AI already ~25% of revenue and growing faster
- Secondary marks ($170B) and a $165-175B primary round show deep, oversubscribed private demand ahead of a 2026/27 IPO
A great business priced for perfection (~25x run-rate revenue) with a structurally compressing margin, paying its biggest competitors (the hyperscalers) for the very compute that drives its growth.
- Gross margin already fell to ~74% and the CEO says it goes lower — agentic volume may not outrun the COGS pass-through to cloud providers
- ~$170B on ~$6.9B run-rate is ~25x; SNOW (a profitable public comp) trades ~14x EV/S and grows 34% — Databricks must hold >60% growth for years to justify it
- Hyperscalers are both supplier and competitor; AWS/Azure/Google can bundle a 'good-enough' lakehouse + AI and squeeze share and margin
- Microsoft Fabric and Snowflake's AI push attack from both flanks; the warehouse land-grab could stall
- No public financials, undisclosed operating margin/SBC; IPO could reset the multiple if the market re-prices growth-at-any-margin
What it is worth
Last-priced round + secondary marks + reverse-multiple sanity check vs public comps (no DCF on undisclosed margins).
$90-120B
growth decelerates sharply or AI capex cools, gross margin keeps sliding, and the IPO re-prices to ~13-15x EV/S in line with Snowflake; a multiple reset, not a business failure.
$165-175B
the round/secondary range holds; growth moderates to ~50-60% post-IPO, ~74% GM steady-ish; multiple compresses toward ~20x as it seasons as a public comp between SNOW (14x) and PLTR (40x+).
$200B+
IPO opens on sustained >70% growth, AI mix >30%, margins stabilize via DBU volume + price; market pays a scarcity premium for the only profitable AI platform at scale.
Last priced ~$134B (Series L, Dec 2025); new round in talks $165-175B (Jun 2026); secondaries imply ~$170B (Forge $242/sh). At ~$170B / ~$6.9B run-rate ≈ ~25x run-rate revenue — a premium to profitable public peer Snowflake (~14x EV/S, 34% growth) justified by the ~80% growth + FCF positivity, but pricing in years of >50-60% growth with margins not collapsing. Reverse read: the price implies the market expects Databricks to roughly double revenue toward ~$12-14B within ~2 years while staying FCF-positive — i.e. growth must stay well above peer rates AND the agent-compute margin drag must be offset by volume.
SWOT
Strengths
- Re-accelerating to >80% YoY at ~$6.9B run-rate — extremely rare at scale
- Only profitable (FCF-positive) name in the AI IPO pipeline
- Open lakehouse standard (Delta + Iceberg via Tabular, Unity Catalog) reduces lock-in fear vs proprietary rivals
- Multi-cloud (AWS/Azure/GCP) — not hostage to one hyperscaler
- Strongest model/data-gravity flywheel: train, govern, serve, and run agents on the same governed data
- Founder-led, deep research roots (Spark, MLflow, Delta); >140% NDR; ~70% of Fortune 500
Weaknesses
- Gross margin compressing (~74%, down from >80%) as agents consume compute — structurally capped by hyperscaler COGS pass-through
- Profitability is thin/undisclosed; operating margin not public
- Heavy dependence on the same hyperscalers it pays for compute and competes with
- Aggressive M&A and stock-comp cadence; private-company opacity on true GAAP profitability
- Valuation (~$170B / ~25x run-rate revenue) prices in years of flawless execution
Opportunities
- Agentic AI as a new compute super-cycle — more queries, more DBUs (volume offsets margin)
- Displacing legacy EDWs and Snowflake in the data-warehouse layer ($1.5B and climbing)
- Lakebase/operational DB expansion (Neon) into the OLTP + transactional market
- IPO window late-2026/2027 unlocking liquidity and a public currency for M&A
- International + regulated-industry expansion (gov, finance, healthcare) on governed AI
Threats
- Snowflake counter-attacking in AI/ML and on price; Microsoft Fabric bundling against both
- Hyperscalers (AWS, Azure, Google) building competing first-party lakehouse/AI stacks while collecting Databricks' compute spend
- Margin erosion if agent compute outpaces price/efficiency gains
- A cooling enterprise-AI spend cycle or IPO-market shutdown could compress the multiple sharply
- OpenAI/Anthropic/foundation-model platforms moving down-stack into data + agents
Moats, dependencies & bottlenecks
Moats
Once governed data, pipelines, Unity Catalog policies and ML/agents live on the lakehouse, migration is painful — the >140% NDR is the proof.
Owns/co-owns the open formats enterprises standardize on; Tabular acquisition co-opted the Iceberg threat.
Runs on all three hyperscalers — buyers avoid single-cloud lock-in; but it also depends on those same clouds.
Mosaic AI, Agent Bricks, Genie, Lakebase shipped fast; founder/academic roots in Spark sustain credibility.
>800 customers >$1M and ~70% of F500 fund a large enterprise salesforce competitors can't easily match.
Dependencies
Infrastructure supplier AND competitor Buys all compute/storage from them (the COGS driving margin compression) while they build competing first-party lakehouse/AI products. Frenemy concentration is the core structural risk.
GPU/compute supplier AI/agent and Mosaic training workloads ride NVIDIA accelerators (via the clouds); GPU cost/availability shapes AI gross margin.
Foundation-model partner Multi-year deals (Anthropic ~$100M Mar 2025; Gemini Jun 2025; OpenAI ~$100M Sep 2025) embed third-party LLMs; model economics/terms affect AI-product margin and differentiation.
The >80% re-acceleration leans on an agentic-AI spend wave; a budget pullback would hit growth and the multiple hardest.
Funding/liquidity Pre-IPO; relies on continued private rounds (and an open IPO window) for liquidity and employee/investor exit.
Advantages
- Re-accelerating growth at scale (>80% at ~$6.9B)
- FCF-positive — unique among AI-IPO peers
- Open lakehouse standard lowers lock-in objection
- Multi-cloud reach across the entire hyperscaler base
- Full-stack: ingest → warehouse → ML → agents → operational DB on one governed platform
Weaknesses
- Structurally compressing gross margin (~74% and falling)
- Undisclosed GAAP profitability / heavy SBC + M&A
- Deep dependence on competitor-suppliers
- Valuation prices in years of flawless execution
Bottlenecks
- Gross-margin pass-through to hyperscalers — agent compute grows COGS faster than revenue can re-price
- GPU/compute cost and availability for AI workloads
- Enterprise sales-cycle length for large migrations
- Competing directly with its own infrastructure suppliers (AWS/Azure/Google)
Top signals & trends
Top signals
Re-acceleration at $6.9B scale is the single strongest fundamental signal — agentic demand is real.
Confirms the agent-compute COGS drag; watch whether DBU volume growth offsets it.
Deep oversubscribed private demand; secondaries (~$170B) corroborate.
Liquidity catalyst, but also the first true public re-pricing test of the ~25x multiple.
Monetization of the AI thesis is showing up in the mix, not just narrative.
Trends
Agents generate far more queries than humans — a volume tailwind for consumption revenue (and the margin headwind).
Interoperability reduces lock-in fear; Databricks co-opted Iceberg via Tabular to stay the neutral standard.
Enterprises collapsing separate warehouse + lake + ML stacks onto one platform — Databricks' core pitch vs Snowflake.
Suppliers bundling competing 'good-enough' offerings can pressure share and price.
Profitability differentiates Databricks in a pipeline of cash-burning AI listings.
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 compute/storage supplier (and reseller + competitor) — large share of COGS.
Azure Databricks first-party integration; compute supplier, reseller, and Fabric competitor.
GCP compute supplier + Gemini LLM partner (Jun 2025); also a competitor.
GPU/accelerator supplier (via clouds) powering Mosaic AI training + agent inference.
Foundation-model partner (~$100M, 5-yr, Mar 2025) — Claude embedded in the platform. Private.
>20,000 organizations; >800 customers >$1M ARR, >70 >$10M. Named users span finance, retail, healthcare, tech.
Shell / Comcast / Block / Rivian (representative) Public reference customers across energy, media, fintech, auto; broad cross-industry base.
Expanding governed-AI footprint in gov, finance, healthcare.
Closest public comp — data cloud/warehouse. ~$83B mkt cap (Jun 2026); FY27 product-rev guide ~$5.84B at ~34% growth, ~126% NRR, ~14x EV/S. Databricks now out-grows it 80% vs 34% and is larger by run-rate.
Both a hyperscaler partner (Azure Databricks) and a competitor via Microsoft Fabric, which bundles lake+warehouse+BI+AI into the Azure stack.
BigQuery + Vertex AI is a competing first-party data+AI stack; also a Databricks cloud partner and Gemini supplier.
AWS first-party analytics+AI competes; AWS is simultaneously Databricks' largest infra supplier and reseller.
Adjacent enterprise-AI/ops platform; $272B cap, ~71% growth, >40x EV/S — the richly-valued public AI comp investors anchor Databricks against.
Operational/transactional DB — overlaps as Databricks pushes into OLTP via Lakebase/Neon. $25.7B cap, ~25% growth, ~10.7x EV/S.