Groq, which makes chips for data centers and autonomous vehicles, raises $300M Series C, led by Tiger Global and D1 Capital, at a $1B+ valuation
Context & Ripple Effects
This April 2021 round is the entry point of Groq's capital arc: the $640M Series D led by BlackRock three years later explicitly marked itself as up from this $1B-plus valuation, and the escalation continued into 2025's $750M raise at a $6.9B post-money.
What makes the 2021 milestone worth rereading is where the curve went next: after peaking near $7B, Groq signed a $20B Nvidia licensing agreement that took CEO Jonathan Ross and much of its senior team, and its subsequent raise priced well below that peak.
First-order effects
- Tiger Global and D1 Capital gain their position in AI silicon via a crossover-style growth check into a pre-revenue-scale hardware maker, giving Groq the runway to build out its LPU and data-center business.
- Groq crosses the unicorn threshold, which sets the reference price every later investor anchors on — the Series D pitch, the $6.9B round, and even the post-Nvidia-deal markdown all quote this $1B baseline.
Second-order effects
- The round normalizes nine-figure checks for inference-chip startups, pulling BlackRock, Disruptive, and Infinitum into progressively larger follow-ons and shifting Groq's funding from venture funds toward asset managers financing physical data-center capacity.
- Nvidia's eventual response was not a price cut but a talent-and-license play: the $20B non-exclusive licensing agreement absorbed Jonathan Ross's team, showing incumbents competing for startup engineers rather than just market share.
Third-order effects
- If the pattern holds, AI-inference companies are financed like infrastructure projects — successive mega-rounds tied to megawatt and data-center targets (Groq's 12 centers in 2025, a stated path to 200 MW by end-2027) — which exposes them to down rounds when strategy shifts, as the fall from $6.9B to $3.5B shows.
- The trajectory suggests a structural consolidation risk for independent chip challengers: once a hyperscaler incumbent can license the technology and hire the founders, minority investors are left holding equity in a residual operator.
The trend: Inference-compute startups are being funded through ever-larger infrastructure-style mega-rounds whose valuations swing on capacity buildouts and incumbent poaching rather than steady fundamentals — and this 2021 Series C is the baseline of that cycle.