What is Fireworks AI?
Fireworks AI helps teams build and scale ai inference cloud products.
- Category
- AI inference cloud
- Headquarters
- Redwood City, CA
- Founded
- 2022
- Employees
- 115 disclosed in 2025; hiring planned
- Total funding
- $327M+ disclosed equity
- Valuation
- $4B last disclosed valuation
What is Fireworks AI?
Fireworks AI is a ai inference cloud company founded in 2022 and headquartered in Redwood City, CA.
Fireworks AI builds ai inference cloud infrastructure for teams that need production software, AI, or data workflows rather than one-off prototypes. $280M ARR reported in 2025 Its public scale signal is Cursor and AI application teams reported; full customer count not disclosed.
The company sits in a fast-moving market where buyers care about reliability, security, integration depth, and procurement maturity. Processes about 15T AI tokens per day as reported in 2026. Its position is strongest when customers need a managed platform that shortens engineering time while still fitting into existing cloud, data, and developer workflows.
For sellers, Fireworks AI is best treated as a scaled technical buyer. Engineering and product leaders influence architecture, finance and operations shape budget, and security or procurement becomes more important as contract size grows.
What does Fireworks AI offer?
Fireworks AI's product set centers on Serverless inference, Dedicated deployments, Fine-tuning.
- Serverless inference· Core product
- Dedicated deployments· Core product
- Fine-tuning· Core product
- Reinforcement fine-tuning· Expansion product
- Model library· Expansion product
- Enterprise inference platform· Expansion product
How does Fireworks AI make money?
Fireworks AI makes money through usage, subscription, committed-capacity, and enterprise contracts depending on customer scale.
Fireworks charges serverless models per million input and output tokens, with cached and batch tokens discounted by 50% in many cases. Fine-tuning is priced per million training tokens by model size, and on-demand deployments are priced per GPU-hour: H100 and H200 at $7/hour, B200 at $10/hour, and B300 at $12/hour.
Growth is driven by land-and-expand adoption: individual developers or small teams start with self-serve usage, then production workloads create larger commitments, security requirements, support needs, and procurement events. Enterprise customers typically pay for higher limits, private deployment patterns, governance, support, SLAs, and negotiated usage economics.
The unit economics depend on the underlying product category. Software-heavy products expand through seats and usage, while AI infrastructure and GPU-cloud businesses require disciplined capacity planning, reserved commitments, power and data-center execution, and high utilization of expensive compute assets.
Who leads Fireworks AI?
Fireworks AI is led by Lin Qiao, with technical, product, and go-to-market ownership spread across the leadership team.
- Lin QiaoCo-founder & CEOCo-founder since 2022Former Meta engineer who helped build PyTorch; leads company and product strategy.
- Fireworks founding teamCo-foundersFormer Meta/PyTorch teamBuilt the systems background behind Fireworks inference and model tooling.
- AI infrastructure leadershipEngineering leadersScale-up phaseOwns GPU operations, runtime optimization, and model serving reliability.
- Enterprise GTM leadershipGo-to-market leadersScale-up phaseBuilds enterprise adoption around inference migration and cost control.
How do you contact Fireworks AI's leadership?
Use published company channels first. The personal addresses below are format-following examples using fireworks.ai; they should be verified before outreach and are not presented as confirmed personal inboxes.
first@fireworks.ai (format-following example, not a verified personal mailbox)How much funding has Fireworks AI raised?
Fireworks AI has $327M+ disclosed equity; its latest disclosed valuation/status is $4B last disclosed valuation.
Fireworks AI's disclosed financing history is concentrated in these major events: 2022-2023 Seed/Series A - about $25M; Jul 2024 Series B - $52M; Oct 2025 Series C - $250M at $4B valuation. The latest disclosed valuation or market status is $4B last disclosed valuation.
2022-2023: Seed/Series A - about $25M. Benchmark and other investors back the former Meta team building inference infrastructure. Jul 2024: Series B - $52M. Sequoia-led financing scales the model-serving platform and go-to-market. Oct 2025: Series C - $250M at $4B valuation. Led by Lightspeed, Index Ventures, and Evantic, with Sequoia participating; split between primary and secondary capital.
The funding signal matters because it defines buying capacity and operating pressure. Late-stage capital usually means new hiring, platform expansion, security upgrades, finance-process maturity, and larger procurement reviews; earlier-stage profiles require tighter ROI and founder-led evaluation.
How did Fireworks AI get here?
Fireworks AI's path runs from founding in 2022 through product expansion and its latest financing or public-market milestone.
- 2022Company foundedLin Qiao and former Meta/PyTorch colleagues start Fireworks AI.
- 2023Platform launchesFireworks begins serving open models through APIs for application developers.
- Jul 2024Series B raisedThe company raises $52M to expand inference capacity.
- Oct 2025Series C raisedFireworks raises $250M at a $4B valuation and reports $280M ARR.
- Late 2025Token volume acceleratesDaily processed tokens reach roughly 10T by late 2025.
- Apr 202615T daily tokens reportedBusiness Insider reports Fireworks processing about 15T tokens per day.
Who are Fireworks AI's competitors?
Fireworks AI competes with category specialists, open-source alternatives, and larger platform vendors.
- Together AICloud for open and custom models with training and inference products.
- BasetenInference platform focused on deploying custom models into production.
- ReplicateDeveloper marketplace and API for running open-source models.
- CerebrasAI inference and training using wafer-scale systems.
- GroqSpecialized LPU inference cloud optimized for low latency.
Fireworks AI — frequently asked questions
