· investment-strategies · 3 min read
What Is Groq’s Valuation in 2026? $3.5B After a $350M Series A
Groq’s current valuation is $3.5 billion as of the August 17, 2026 $350M Series A (Disruptive; planned NVIDIA). That sits on top of $650M in June — not instead of it. $3.5B is not a simple down round vs the old ~$6.9B chip-company mark.
Groq’s current valuation is $3.5 billion as of the August 17, 2026 $350 million Series A, led by Disruptive, with planned participation from NVIDIA. That raise sits on top of $650 million in June 2026 — company: $1 billion of recent capital — not a replacement for the June headline. The $3.5B mark is a new-entity print after the NVIDIA licensing deal, not a simple down round from ~$6.9B. The raise is subject to customary closing conditions.
Groq funding — June vs August
| Date | Round | Amount | Mark |
|---|---|---|---|
| Aug 17, 2026 | Series A (reset entity) | $350M | $3.5B |
| Jun 2026 | Pivot raise | $650M | Not the August close |
| ~Sep 2025 | Prior chip-company mark | — | ~$6.9B (different story) |
Key facts
| Field | Detail |
|---|---|
| Company | Groq LLC — AI inference / neocloud (Mountain View / SF materials) |
| Round | $350M Series A @ $3.5B (restructured entity) |
| Date | August 17, 2026 |
| Lead | Disruptive (Alex Davis: Groq Executive Chairman and Disruptive CEO) |
| Planned | NVIDIA |
| Footprint (company) | 13 data centers, NA / Europe / ME / APAC; 6M+ developers |
| Capacity plan (company) | ~54 MW today → 200+ MW in 2027 |
| Status | NVIDIA Cloud Partner (NCP); closing conditions apply |
Who uses the product — and for what job
Users: model makers, AI-native companies, and Fortune 500 teams that need medium and larger NVIDIA clusters for training and inference without becoming CoreWeave-scale operators themselves.
Job: buy tokens and cluster-hours from a global inference cloud, not bet that Groq’s old LPU roadmap beats NVIDIA silicon.
That is a different company than the 2024–25 chip challenger. NVIDIA hired founder Jonathan Ross and others as part of a ~$20B licensing deal (TechCrunch). Remaining Groq sells NVIDIA-architecture capacity.
Why now
- Inference is the power-and-capex layer. June’s $650M started the pivot; August adds more for cluster demand.
- Neoclouds (CoreWeave, Lambda, Nebius, Firmus) are a financing competition. Groq is raising equity to stay in that race.
- A 50% lower headline valuation vs ~$6.9B is the cost of selling the old story. Company: new entity, not a classic down round.
Why Disruptive / NVIDIA — portfolio fit
Disruptive is both lead investor and, via Davis, operator of the remaining company. That is concentrated control, not a diversified syndicate.
NVIDIA as planned participant is supply + certification + ecosystem. Same pattern as other NCPs: the GPU vendor finances demand for its own racks.
Likely founder/board rationale: stop competing with NVIDIA at the chip layer; become a distribution and ops company NVIDIA can live with.
| Investor type | What they bring |
|---|---|
| Disruptive | Capital + executive control |
| NVIDIA (planned) | GPUs, NCP standards, customer signal |
Competitive map
| Player | Lane |
|---|---|
| CoreWeave / Lambda / Nebius | GPU neoclouds |
| Firmus | NVIDIA AI factories / campus build |
| Fireworks | Inference software/serving, not MW of sites |
| Volta Infra | AI cloud buildout (earlier August) |
| Hyperscalers | Default alternative for many enterprises |
Market signal
$1B in two months for a reset company says LPs will fund megawatts, not nostalgia for LPUs. It also says equity value and capex need can move in opposite directions.
The useful distinction aggregators skip: this is a Series A label at $3.5B after a prior higher mark, plus a planned NVIDIA transaction that is not cash already closed.
When not to use this as a template
- Wrong if you describe Groq as primarily an LPU chip startup as of this round.
- Wrong if you call $3.5B a simple down round without the NVIDIA transaction.
- Wrong if you treat “planned participation” as cash already closed.
Practical takeaway
- Founders (infra): If NVIDIA is the customer and the competitor, the viable path may be NCP, not a second CUDA.
- Investors: Underwrite MW, offtake, and depreciation — not token demos.
- Operators: Compare Groq to other neoclouds on cluster availability and price, not brand history.
Sources
- Groq (Aug 17, 2026): https://groq.com/newsroom/groq-closes-usd350-million-series-a-building-the-world-s-leading-ai-inference-cloud
- TechCrunch: https://techcrunch.com/2026/08/17/groq-raises-350m-to-fuel-its-pivot-from-ai-chips-to-neocloud/
- Related: /startup/groq · /2026-august-14-18-investment-news-ai-infra-physical
