Overview
Nvidia just bought out its most promising challenger. On Christmas Eve 2025, the chip giant paid $20 billion—three times Groq's September valuation—to acquire the AI inference startup founded by the creator of Google's revolutionary TPU chip. The deal eliminates Nvidia's fastest competitor in inference, the next battleground where AI chips actually make money.
It's the latest—and largest—sign that Big Tech's AI land grab has entered a new phase. Microsoft, Amazon, and Google spent 2024 scooping up AI talent through quasi-mergers designed to dodge antitrust scrutiny. Now Nvidia is doing the same thing, calling its acquisition a 'non-exclusive licensing agreement' while hiring Groq's entire leadership team and absorbing all its assets except the cloud business. AMD and Intel, who spent all year positioning inference chips as their path to challenge Nvidia, just watched their best case study disappear.
Key Indicators
People Involved
Organizations Involved
The $3+ trillion chip giant controls 90%+ of AI training chips and is now consolidating inference.
Built the Language Processing Unit to challenge Nvidia with 10x faster, more efficient AI inference.
Nvidia's closest rival, targeting inference market with MI350 series chips.
Legacy chip giant betting on Gaudi 3 for AI inference comeback.
Timeline
-
Nvidia Acquires Groq for $20B
AcquisitionNvidia's largest-ever deal eliminates leading inference competitor. Called 'non-exclusive licensing agreement' to ease antitrust review.
-
Amazon Doubles Anthropic Bet to $8B
InvestmentAmazon invested another $4B in Anthropic, total $8B commitment in AI partnership race.
-
Groq Raises $750M Series E
FundingValuation jumped to $6.9B, more than doubling in 13 months as inference market exploded.
-
AMD Launches MI350 Series
Product LaunchAMD released inference-optimized chips claiming 35x generational performance leap, targeting Nvidia's dominance.
-
Groq Raises $640M Series D
FundingBlackRock led round valuing Groq at $2.8B as inference demand surged.
-
Microsoft's $650M Inflection Acquihire
Talent AcquisitionMicrosoft hired most of Inflection's 70 employees, including founder Mustafa Suleyman, via licensing deal to dodge antitrust.
-
Amazon Invests $2.75B More in Anthropic
InvestmentAmazon's total Anthropic investment reached $4B, part of Big Tech's AI talent war.
-
Nvidia's $40B ARM Deal Collapses
RegulatoryRegulators worldwide blocked Nvidia's ARM acquisition over competition concerns, biggest deal failure in semiconductor history.
-
Mellanox Acquisition Closes
RegulatoryDeal cleared by EU, US, and Chinese antitrust authorities after 13-month review.
-
Nvidia Acquires Mellanox for $6.9B
AcquisitionNvidia's then-largest deal, 17x bigger than any previous acquisition, added networking to AI infrastructure.
-
Groq Founded by Google TPU Creator
Company FormationJonathan Ross and Google X veterans founded Groq to build inference-optimized AI chips.
Scenarios
Regulators Block Deal, Groq Remains Independent
Discussed by: Bernstein analyst Stacy Rasgon, antitrust observers citing Microsoft-Inflection precedent
Chinese or European authorities could block the deal, forcing Nvidia to unwind the transaction. Regulators already designated Microsoft's similar Inflection acquihire a 'merger' and blocked Nvidia's ARM purchase. The 'licensing agreement' framing may not survive scrutiny given Nvidia is acquiring all assets, hiring all leadership, and paying triple the valuation. If blocked, Groq would remain independent with its $750M war chest, continuing to challenge Nvidia in inference while validating regulatory concerns about AI consolidation.
Deal Clears, Nvidia Dominates Inference Market
Discussed by: Financial analysts noting inference market shift, semiconductor consolidation trend
The deal closes after 6-12 months of regulatory review, potentially with minor concessions. Nvidia integrates Groq's LPU technology into its AI factory architecture, extending dominance from training into inference. AMD and Intel lose their best competitive case study for specialized inference chips. With 90%+ training market share and Groq's low-latency inference tech, Nvidia becomes the end-to-end AI infrastructure provider. Remaining inference startups like Cerebras face pressure to sell or partner, accelerating consolidation.
Conditional Approval Creates New Competitors
Discussed by: Antitrust scholars, observers of tech merger remedies
Regulators approve the deal with behavioral remedies: Nvidia must license Groq technology to competitors, maintain Chinese walls between teams, or divest certain assets. The 'non-exclusive' licensing language becomes legally binding, forcing Nvidia to share LPU innovations with AMD, Intel, or startups. This could accelerate inference chip development across the industry while still giving Nvidia first-mover advantage with Groq's team. Enforcement would be challenging, and Nvidia's history of leveraging acquisitions (Mellanox's $10B networking business) suggests it would find ways to capture most value.
Historical Context
Microsoft's Inflection AI Acquihire (2024)
March 2024What Happened
Microsoft paid $650M to license Inflection's AI models and hire most of its 70 employees, including co-founder Mustafa Suleyman, without formally acquiring the company. The structure was designed to avoid antitrust scrutiny since traditional acquisitions trigger mandatory regulatory review.
Outcome
Short term: UK's Competition and Markets Authority designated it a 'relevant merger situation' despite the licensing structure, though ultimately cleared the deal.
Long term: Established the 'quasi-merger' playbook Big Tech now uses to acquire AI talent and IP while minimizing regulatory friction.
Why It's Relevant
Nvidia is copying Microsoft's template by framing Groq as a licensing deal while acquiring all assets and hiring all leadership—the exact structure regulators are scrutinizing.
Nvidia's Failed ARM Acquisition (2021-2022)
September 2021 - February 2022What Happened
Nvidia announced a $40B deal to buy ARM Holdings from SoftBank, which would have given it control over chip architecture used across the industry. Regulators in the US, EU, UK, and China raised competition concerns about Nvidia controlling technology its competitors rely on.
Outcome
Short term: The deal collapsed in February 2022 after 17 months, the largest failed semiconductor acquisition ever.
Long term: Nvidia became cautious about large acquisitions, leading to the licensing-deal framing for Groq. Regulators signaled they'll block deals that concentrate AI infrastructure control.
Why It's Relevant
The ARM failure explains why Nvidia calls Groq a 'licensing agreement'—it's learned to structure deals to avoid regulatory triggers, though the substance may be identical to acquisition.
Nvidia's Mellanox Acquisition (2019-2020)
March 2019 - April 2020What Happened
Nvidia paid $6.9B for Israeli networking company Mellanox, 17x larger than any previous Nvidia acquisition. The deal added InfiniBand and Ethernet networking technology to Nvidia's GPU portfolio, enabling complete data center solutions.
Outcome
Short term: After 13-month regulatory review, the deal closed in April 2020 with approval from Chinese, EU, and US authorities.
Long term: Mellanox became a $10B+ revenue business for Nvidia, integral to its AI data center dominance. The acquisition's success emboldened Nvidia to pursue larger deals.
Why It's Relevant
Mellanox proved Nvidia can successfully integrate acquisitions and build massive businesses from them. The $20B Groq deal follows the same playbook: buy complementary technology, integrate into AI infrastructure, dominate new market segment.