Video summary

$15 млрд после «Яндекса»: как Алексей Миловидов построил ClickHouse

Main summary

Key takeaways

Business

Business-focused summary: how ClickHouse (after Yandex) was built and scaled

Origin & strategic positioning (product + distribution)

  • ClickHouse began as internal tooling at Yandex, evolving from:

    • A “primitive” reporting database updated ~daily (numbers + hashed strings)
    • Real-time data aggregation technology
    • The eventual insight: unify these ideas into an analytical “data warehouse” concept for clickstream/analytics workloads

  • Key adoption tactic inside Yandex:

    • Keep the infrastructure “barely noticeable” and avoid forcing org-wide redesign
    • Let it grow through documentation + internal champions
    • Spread via real performance wins (reports in seconds vs hours)

  • External go-to-market idea crystallized later:

    • Open-source as a “funnel” into analytical DB niches with incumbents (e.g., Vertica/Teradata-class players)
    • License strategy: open code + strong brand/trademark control to prevent easy rebranding

Market messaging / product naming framework

  • The “Clickstream Data Warehouse” naming logic:
    • Clickstream = web/ad click events and telemetry
    • Data warehouse = analytics storage (vs transactional databases), aligning expectations with “analytics-first” use cases

How it became a global company (spinoff mechanics + governance)

  • 2019–2020: outside investors noticed repo/activity (GitHub + community interest) and reached out

  • Critical constraint: IP/trademark risk if the founder left without a structured carve-out

  • Funding & control negotiation:

    • Yandex wanted a controlling stake (mentioned as ~51% during negotiations)
    • The parties later agreed on a smaller share structure aligned with investors’ growth expectations
    • Yandex ultimately held around ~28% (later referenced during valuation/share discussions)

  • Team scaling approach:

    • Avoid hiring “too fast” in big batches to preserve startup dynamics and engineering culture
    • The move required convincing the ~12–14 person core team; some refused, including due to 2022 war/time/family logistics concerns

Co-founder & org design (role clarity)

  • Co-founder triad (framed as a complementary system):

    • Alexey Milovidov: core technology + engineering leadership
    • Aron Katz: sales/investor interaction (VPOF) and business building
    • Yuri Izrailevsky: product/company scaling and cloud/product direction
      • Inspired by cloud/startup ops experience
      • Described as managing “product + execution” while engineering stays focused on tech

  • Repeated leadership principle: The founder should focus on the loop they can win in—avoid low-leverage work once direction is clear

Growth playbook / GTM execution (as described)

1) Open-source + community-driven adoption

  • Publish the code and encourage external contributions (features, bug fixes, docs)
  • Treat contributors as co-builders while maintaining:
    • Brand protection via licensing (code usable commercially; trademark not)
  • Organize meetups/conferences early to build developer awareness
    • Marketing through community rather than ads

2) Enterprise product motion: “Cloud” added on top of OSS

  • Open-source = “bare product”
    • Customers handling their own scaling/backups/monitoring/security/integration if self-hosting
  • ClickHouse Cloud sells the operational bundle
    • Managed deployments for environments like AWS to reduce engineering/admin burden
  • Expansion pattern: “land small, then expand”
    • Start with pilots/limited contracts (e.g., a single department)
    • Once proven, neighboring departments adopt (mouse-to-elephant growth)
  • The company emphasized real-time analytics as a core technical advantage versus BI/request-driven approaches

3) Sales packaging & pricing mechanics (subscription/commitments)

  • Revenue model described as largely driven by:
    • Subscriptions/commitments
    • Customer expansion over time in both:
      • paid usage
      • enterprise footprint across departments

Key metrics & KPIs mentioned (and what they imply)

  • Company valuation: ~$15B (target/in-the-future Forbes billionaire mention)
  • Investment amounts:
    • $50M “Series Seed/A”-style funding “at once” around Aug 2021, with later increases around Oct 2021
    • Total disclosed investments discussed as about ~$200M (one round not disclosed)
  • Revenue mentioned:
    • ~$250M revenue in 2025 (“R was 250 million”)
  • Customer base:
    • About 4,000 paying customers
    • Average contract spend calculations were discussed (e.g., $250M / 4,000), with the conclusion that average enterprise bills likely far exceed a naive average
    • Mention of average enterprise spend “on the order of” ~$500k/year as a rough exploration
  • Growth dynamics:
    • Revenue growth “in previous year” described as roughly ~3x
  • Performance claims used for adoption:
    • Reports produced in seconds vs older workflows running hours

Concrete examples / customer archetypes

  • Large AI and tech customers mentioned:
    • OpenAI (client; described as using ClickHouse)
    • Anthropic
    • Tesla (machine telemetry use case)

  • Additional examples (finance/telecom/AI):

    • Monitoring/anti-fraud transaction analytics (review/telemetry)

  • Pricing expansion example pattern:

    • A small department starts with ClickHouse Cloud, sees cost/performance benefits, then expands to neighboring departments

Engineering/process playbooks (operational tactics)

Continuous integration & automated quality gates

  • “Continuous integration” described as:
    • Thousands of tests
    • Millions tested daily
    • Outcomes tracked as red/green to ensure correctness
  • “Virtual employees / AI employees” used internally for workflow support
    • Humor example: “green cat” correcting checks
    • External contributor interactions mediated via AI-style personas

Code quality & safety against AI errors

  • AI-assisted coding described in phases:
    1. Agent drafts code; user confirms steps
    2. Multiple agents run in parallel in isolation
    3. Feedback loops where agents review/check each other
  • Explicit quality/safety stance:
    • Human review still required (“Human review”)
    • AI is not trusted for 100% correctness; AI can make contextual mistakes
    • Test strategy includes classical engineering:
      • CI + tests + fuzzing (“randomized trials”)

Competitive advantage thesis (what wins vs incumbents)

  • Core moat: real-time analytics performance
    • Benefits compound with AI agent workloads that require fast repeated queries
  • Market capture strategy:
    • Win analytical workloads that incumbents can’t serve efficiently (latency + throughput for agent-driven data exploration)
  • Long-term demand driver:
    • More agents / AI systems increase query counts, driving more database utilization (“turbocharging” demand)

High-level discussion on future / IPO (execution emphasis only)

  • Conversation touched IPO readiness/trajectory at a high level:
    • Concern about being “hostage” to quarterly reporting and short-term investor moods
  • Team plans to be “IPO-ready” through financial/accounting preparedness
    • Emphasis on weighing pros/cons and referencing example outcomes (successful and unsuccessful IPOs)

Presenters / sources

  • Presenter / interview subject: Alexey Milovidov (founder of ClickHouse)
  • Interview location / context (host voice): Unnamed interviewer/host (not clearly identified in subtitles)
  • Mentioned third parties (individuals):
    • Arkady Volosh, Arkady Bashkeev, Greg Gobovsky, Eran Katz, Peter F… (Benchmark Capital), Mike W… (Index Ventures)
    • Yuri Izrailevsky, Kevin (recruitment lead), Nicholas II (“mentioned” in context of a book), Feynman (book author), Oshmanova (book author mentioned)

Original video