Fireworks AI Raises $1.5B Series D at $17.5B Valuation: The Inference Layer Reaches $1B ARR

Fireworks AI, a San Francisco-based AI inference and model serving platform, raised $1.505 billion in Series D financing at a $17.5 billion post-money valuation. The round was led by Atreides Management, Index Ventures, and TCV, with participation from existing investors Lightspeed Venture Partners, Evantic, and NVIDIA. The financing closed in July 2026 and was tracked by AngelLinx Intelligence in the August window. Fireworks AI has surpassed $1 billion in annualised revenue and serves more than 40 trillion tokens per day across its platform.

What Fireworks AI Builds

Fireworks AI provides API-based inference infrastructure that allows companies to deploy and customise large language models at production scale. Unlike OpenAI's API, which serves a fixed set of proprietary models, Fireworks' platform allows customers to fine-tune and deploy specialised models trained on their own proprietary data. The company's core technical differentiation is the ability to serve custom fine-tuned models at costs and latencies that are competitive with serving foundational models directly, a capability that matters enormously for enterprise customers whose use cases require domain-specific model behaviour rather than general-purpose outputs.

Fireworks processes more than 40 trillion tokens per day, with over 95% of that traffic coming from models customised on proprietary customer data. That ratio is the company's most telling performance metric: it signals that the demand driving Fireworks' growth is not generic AI API usage but enterprise customers building differentiated AI products on top of the inference layer.

Named Entity: Key Investors and Their Thesis

Atreides Management is a Boston-based investment firm that manages both public and private equity, known for concentrated bets on technology companies it believes have durable structural advantages. Index Ventures has backed Notion, Figma, Roblox, and dozens of other infrastructure and product companies. TCV has a long history of backing late-stage technology leaders including Netflix, Spotify, Facebook, and Airbnb. NVIDIA's participation signals that Fireworks' growth in custom model serving aligns with NVIDIA's compute infrastructure roadmap: more custom models means more NVIDIA GPU consumption per token. See enterprise AI and infrastructure investors on AngelLinx for the broader landscape of funds deploying at Fireworks' scale and stage.

The Customer Base: Inference as Critical Infrastructure

Fireworks' named customer list includes Cursor, Perplexity, Notion, Sourcegraph, Uber, DoorDash, Shopify, and Upwork. The breadth of this list across developer tools (Cursor, Sourcegraph), consumer AI (Perplexity), enterprise productivity (Notion, Upwork), and consumer platforms (Uber, DoorDash, Shopify) confirms that AI inference is becoming a horizontal infrastructure layer rather than a narrow developer tool. When Uber and DoorDash are running customised AI models for operational use cases through a third-party inference platform, inference infrastructure has crossed into mission-critical territory.

We covered enterprise AI infrastructure in earlier reports on Cohere's $240M ARR trajectory and on the broader AI capital deployment patterns of the week. Fireworks sits at the infrastructure layer beneath companies like Cohere: it provides the serving layer that makes running customised LLMs at scale economically viable.

What This Means for Founders

Fireworks AI's $17.5 billion valuation at $1 billion in annual recurring revenue implies a 17.5x revenue multiple, significantly higher than typical SaaS multiples but consistent with AI infrastructure companies that are growing 5x year-over-year. Founders building AI infrastructure products can use Fireworks as a revenue multiple benchmark when contextualising their own valuations in investor conversations. The more important insight is the growth rate: Fireworks grew 5x year-over-year. At that pace, the business will approach $5 billion in ARR within two to three years if growth sustains, which justifies the current valuation on a forward multiple basis.

Founders can see what other AI infrastructure companies are raising on the AngelLinx live listing. Find investors backing AI infrastructure and enterprise AI startups on AngelLinx @ angellinx.ai/register.

AngelLinx Intelligence | angellinx.ai