WillItSurvive

Cloud data warehouse selling sub-second query speed for analytics workloads.

Losing its protection
Assessed 21 July 2026 · Evidence as of 21 July 2026
The structural protection is eroding.
On the recordSince 21 July 2026 · Checkable from August 2026 · Every call on the record Checkpoint: Firebolt OSS general availability
The call

Firebolt has cash left but its speed pitch is now standard across Snowflake, Databricks, and ClickHouse, and it is shrinking while it waits to find out if open source changes that.

Firebolt built software that runs database queries very fast, but Snowflake, Databricks, Google's BigQuery, and ClickHouse have all caught up or nearly matched that speed in 2025 and 2026. Open-source tools and artificial intelligence are also making the engineering tricks behind speed less special and harder to sell. The company has shrunk its team by nearly a third, hasn't raised new funding in over four years, and has only about 12 large customers, which suggests it is struggling to grow. The one thing that could turn this around is if Firebolt's free open-source version, launching in August 2026, attracts a lot of developers and converts them into paying customers over the next six months, but if that launch flops, the company will likely face a down-round deal, be acquired for parts, or shut down within a year.

Why
01
AI pressure: AI-assisted query optimizers in Snowflake, Databricks, and open-source engines like DuckDB. AI-driven query tuning is turning Firebolt's speed edge into a commodity feature everyone else can now ship too. Could AI do the core job? Yes, likely.
02
Bigger players: Snowflake, Databricks, BigQuery, ClickHouse. All four have shipped fast-query features in 2025-26 that match Firebolt's core pitch, with far bigger sales forces behind them. Could a giant copy it? Yes, easily.
03
What protects it. Over $100M in cash gives Firebolt runway to try Firebolt OSS before it runs out of options.
What would change this call
Firebolt OSS reaches general availability in August 2026. Tests whether open-source adoption converts into paid customers or fails to gain traction.
Further headcount reductions or executive departures after Feb 2026 layoffs. Would confirm continued business contraction and instability.
A priced funding round or acquisition announcement within 12 months. Would show whether investors value Firebolt near, above, or below its 2022 $1.4B mark.
Key numbers
  • No priced round since $1.4B Series C (2022-01-26)
  • Headcount down 31% YoY to ~150 (2026-06-20)
  • Only 12 customers at ~$40M ARR (2025-11-27)
Named risks
  • No priced funding round since January 2022; implied valuation may have fallen ~71% to $0.4B. Severity: high
  • Headcount down 31.5% YoY with February 2026 layoffs despite large cash reserves. Severity: high
  • Sole differentiator, query performance, is being matched by Snowflake, Databricks, and ClickHouse. Severity: high
  • Firebolt OSS GA (August 2026) may fail to convert developer adoption into paid revenue. Severity: medium
  • GenAI-assisted development explicitly acknowledged by Firebolt as eroding its engineering moat. Severity: medium
Rivals, as read in July 2026
RivalDirectionEvidence
Snowflakelosing toShipped Optima indexing/performance features narrowing Firebolt's latency edge, 2025-2026
Databrickslosing toShipped Photon performance optimizations closing the gap, 2025-2026
ClickHouselosing toClickHouse Cloud shipped compute-compute separation 'Warehouses' feature in 2025
Google BigQuerycontestingListed as direct incumbent competitor in Firebolt's own comparison pages, 2025-2026
Questions people ask
Is Firebolt profitable in 2026?
No public profitability data exists. Firebolt is estimated at ~$40.4M ARR for 2024 per Getlatka (2025-11-27), with only 12 customers, while headcount fell 31% YoY by June 2026 and no priced funding round has closed since January 2022, suggesting cash conservation rather than profit generation.
Has Firebolt had layoffs?
Yes. In February 2026 Firebolt laid off dozens of employees as part of a stated 'AI-first' strategic shift, despite holding over $100M in cash on hand, per unifygtm.com. Headcount fell from a peak of ~174 (Feb 2024) to ~150 by June 2026, a 31.5% YoY decline.
Could AI replace what Firebolt does?
Firebolt itself warns of this. Its June 2025 blog announcing Firebolt Core states that 'GenAI accelerating software development' makes defensibility 'more difficult than ever,' and the company kept Core closed-source specifically to slow AI-driven cloning of its query engine.
Who is Firebolt's biggest competitor?
Firebolt names Snowflake, Databricks, Amazon Redshift, Google BigQuery, and ClickHouse Cloud as direct rivals in its own comparison pages. Snowflake and Databricks shipped comparable performance features (Optima, Photon) in 2025-2026, narrowing Firebolt's core latency advantage.
Who owns Firebolt and who runs it?
Firebolt is privately venture-backed, founded in 2019 by Eldad Farkash and Saar Bitner. Hemanth Vedagarbha, ex-Oracle/Confluent, became President in February 2025 and was promoted to CEO in September 2025, with founder Farkash moving to Chairman of the Board.
What was Firebolt's last funding round and valuation?
Firebolt's last priced round was a $100M Series C in January 2022, valuing the company at $1.4B per Barron's (2022-01-26). No priced round has occurred since; total funding across three rounds is roughly $264-270M.
Timeline
  • 2026-02Laid off dozens of employees amid 'AI-first' pivot despite $100M+ cash
  • 2025-09-02Hemanth Vedagarbha promoted to CEO; founder Farkash moves to Chairman
  • 2025-06-23Launched Firebolt Core, a free but closed-source self-hosted edition
  • 2025-03-23Launched FireScale benchmark claiming ~120ms latency at 4,000 QPS
  • 2025-02-06Appointed Hemanth Vedagarbha (ex-Oracle/Confluent) as President
  • 2022-01-26Closed $100M Series C at $1.4B valuation, led by Alkeon Capital
  • 2021-06Closed $127M Series B funding round
  • 2020-12Closed $37M Series A funding round
  • 2019Firebolt founded
Revision history
21 July 2026
First assessed. Losing its protection. The text above is fixed as of this date and is not edited afterwards.
In the same band
Sources