Groq grabs $350M at $3.5B valuation for AI cloud

Groq grabs $350M at $3.5B valuation for AI cloud

August 17, 2026

SAN FRANCISCO, August 17, 2026, 11:18 PDT — Groq said it raised $350 million at a $3.5 billion valuation to boost its AI inference cloud.

  • The valuation comes in roughly 49% under Groq’s $6.9 billion figure from September 2025.
  • Groq is looking to boost power capacity from 54 MW to over 200 MW by 2027.
  • NVIDIA said it will take part, and Groq is now an NVIDIA Cloud Partner.

Groq’s new financing shifts it from an upstart chipmaker to a cloud operator with large capital needs. The company now has to go beyond showing LPU speed and start putting dependable compute into data centers.

The shift is key for AI customers. Groq now offers medium and large NVIDIA (NASDAQ:NVDA) accelerated clusters for both training and inference, alongside its own LPU tech. NVIDIA licensed that technology and picked up Groq founder Jonathan Ross along with other senior leaders back in December.

MeasureBefore the NVIDIA agreementCurrent Groq model
Strategic roleRan LPU chip and offered an alternative to cloud providersActs as NVIDIA Cloud Partner and manages inference cloud
Compute offeredProvided Groq LPU cloud plus local cluster setupsRuns Groq gear plus NVIDIA-powered cluster options
Primary workloadsFocused on fast-response inferenceHandles inference, now includes larger clusters for training
LPU ownership pathBuilt tech internally at GroqHolds non-exclusive rights for NVIDIA, Groq stays independent
Commercial testRelied on chip price/performance to sign up usersNow aims to keep data center use high worldwide

Disruptive is leading the new Series A. NVIDIA’s planned involvement is still pending standard closing checks. Groq said this raise and its $650 million round in June together bring in $1 billion in new capital.

DateCapital raisedDisclosed valuationCompany position
August 2024$640 million$2.8 billionBoosting Groq LPU output
September 2025$750 million$6.9 billionPushing as a standalone inference-chip rival
June 2026$650 millionNot disclosedShifting focus to inference neocloud
August 17, 2026$350 million$3.5 billionNVIDIA partner rolling out bigger worldwide clusters

The $3.5 billion figure is 49.3% lower than the September 2025 valuation. Groq told TechCrunch it considers this number a valuation for the post-licensing business, not a standard down round. The company called the distinction important since it handed core tech access over and lost its founder to NVIDIA.

Groq says it has 13 data centers in operation and over six million developers now. That’s up from the June update, when the company said more than five million developers were using trillions of tokens every week.

Operating metricEarlier disclosed levelLatest disclosed levelChange
Developer reachOver 2 million, September 2025Over 6 million, August 2026At least tripled
Data-center footprint13 sites, June 202613 sites, August 2026No change
Power capacity54 MW, August 2026More than 200 MW planned in 2027More than 3.7x
Recent funding$650 million, June 2026$1 billion total in June and AugustUp by $350 million

The target is high. Going from 54 MW up to over 200 MW means adding at least 146 MW. That works out to about a 270% jump, not considering any old equipment getting replaced or retired.

Groq Executive Chairman Alex Davis said the company is building “the world’s leading AI inference cloud,” calling inference the biggest and most important AI infrastructure layer. Davis made the comments in Groq’s financing release. Source

CoreWeave (NASDAQ:CRWV) posted 2025 revenue of $5.13 billion, but losses keep piling up. Net loss came in at $1.17 billion and net interest expense reached $1.23 billion. The numbers point to how much neoclouds can scale—and the heavy debt they take on.

CompanyOwnershipLatest disclosed scaleFinancial visibilityKey exposure
GroqPrivateRuns 13 data centers, 54 MW capacity, and says it has over 6 million developersNo numbers out for revenue or lossesPace of trying to hit 200 MW-plus expansion
CoreWeavePublic, NASDAQ:CRWVProjects $5.13 billion in revenue for 2025Public numbers; targeted 2025 loss is $1.17 billionRising debt and GPU value dropping

Risks: Groq hasn’t shared revenue, utilization, debt figures or key customer data. The 200 MW goal is still just a plan, not actual capacity. NVIDIA is investor, supplier, and tech licensee. That might help drive demand, but it also means Groq is heavily reliant on NVIDIA.

The new financing gives customers more potential capacity and a bigger range of hardware. The real test for Groq is keeping prices in line as it tries to fund a four-times jump in power. Whether the valuation reset sticks or turns into a red flag depends on that.

BEZ KABLI • EXTENDED COVERAGE

Further analysis

What did Groq announce on August 17, 2026?
Groq announced a $350 million Series A led by Disruptive, with planned participation from NVIDIA. The financing values Groq at $3.5 billion and remains subject to customary closing conditions.
Why is the $3.5 billion valuation important?
It is about 49% below Groq’s $6.9 billion valuation from September 2025. Groq says the new figure applies to its post-licensing business, after NVIDIA licensed Groq technology and hired founder Jonathan Ross and other senior staff.
Is Groq still an AI chip company?
Groq still operates LPU-based inference infrastructure, but its commercial model has widened. It is now an NVIDIA Cloud Partner offering NVIDIA accelerated clusters for training and inference while operating its global cloud platform.
How much capacity does Groq plan to add?
Groq plans to expand from 54 MW to more than 200 MW in 2027. That requires at least 146 MW of additional capacity, or growth of roughly 270% from the current level.
What is the main uncertainty for customers and investors?
Groq does not disclose revenue, utilization, debt or customer concentration. The new funding supports a large build-out, but the company must convert that capacity into sustained demand without losing cost competitiveness.

Marcin Frąckiewicz

Marcin Frąckiewicz is the CEO of TS2 Space and a longtime technology entrepreneur focused on telecommunications, satellite communications and digital innovation. A graduate of the Warsaw School of Economics (SGH), he writes about space technology, artificial intelligence and publicly traded technology companies. His analysis covers major market trends, emerging technologies and the businesses shaping the future of the global economy.