Nvidia has secured Taiwan Semiconductor Manufacturing Company (TSMC) as a production partner to manufacture H200 AI chips, responding to unprecedented demand from Chinese technology companies. The partnership comes as Chinese firms have submitted official orders for 2 million H200 chips designated for 2026 delivery, creating a substantial supply challenge for the semiconductor giant.

Supply and Demand Dynamics

The scale of Chinese demand significantly outpaces Nvidia's current capabilities and inventory levels. Key supply metrics highlight the production challenge:

TSMC is scheduled to begin H200 chip production in the second quarter of 2026, though the exact volume allocation for addressing the supply deficit remains undisclosed.

Manufacturing and Technical Specifications

The H200 chips will be manufactured using TSMC's advanced 4-nanometer process technology. This partnership represents a strategic move to leverage TSMC's established semiconductor fabrication capabilities to meet the substantial order volume from Chinese customers.

Nvidia has set the pricing for H200 chips at $27,000 per unit for Chinese technology companies. The first batch from current inventory is expected to be delivered before the Lunar New Year holiday in mid-February.

Regulatory and Market Considerations

Despite Trump lifting US export restrictions on H200 chips, China has not yet approved the import of these AI processors, creating potential regulatory uncertainty for the arrangement. An Nvidia spokesperson emphasized that licensed sales to authorized Chinese customers would not impact supply capabilities for US customers.

The spokesperson noted that China represents a highly competitive market with rapidly growing local chip suppliers, stating that blocking US exports would undercut national and economic security while benefiting foreign competition.

Strategic Context

Beyond the H200 production partnership, Nvidia continues expanding its chip portfolio through scaling up Blackwell chips and unveiling new Rubin chip lineups. The company maintains existing order obligations while navigating the complex dynamics of serving both domestic and international markets in the competitive AI chip sector.

Nvidia has agreed to license chip technology from AI startup Groq and hire away its CEO in a significant deal that follows the recent pattern of major technology companies acquiring talent and technology from promising startups without formal acquisitions. The agreement was announced by Groq in a blog post on Wednesday.

Deal Structure and Key Personnel

The licensing arrangement includes several key components that demonstrate Nvidia's strategic approach to expanding its AI capabilities:

A person close to Nvidia confirmed the licensing agreement, though financial details were not officially disclosed by either company. CNBC reported that Nvidia had agreed to acquire Groq for $20 billion in cash, but neither company commented on this report.

Strategic Focus on AI Inference Market

Groq specializes in AI inference technology, where artificial intelligence models that have already been trained respond to user requests. While Nvidia dominates the market for training AI models, it faces significantly more competition in the inference segment from traditional rivals such as Advanced Micro Devices and startups including Groq and Cerebras Systems.

The acquisition of Groq's talent and technology represents Nvidia's strategic move to strengthen its position in this competitive landscape. Jonathan Ross, who helped Google start its AI chip program, brings valuable expertise to Nvidia's expanding AI operations.

Industry Pattern and Regulatory Considerations

This deal exemplifies a familiar pattern among the world's biggest technology firms, where companies pay substantial sums to acquire technology and talent while avoiding formal acquisitions. Recent similar transactions include:

• Microsoft's $650 million deal structured as a licensing fee to acquire a startup's top AI executive
• Meta's $15 billion expenditure to hire Scale AI's CEO without acquiring the entire firm
• Amazon's hiring of founders from Adept AI

These arrangements have attracted regulatory scrutiny, though none has been unwound to date. Bernstein analyst Stacy Rasgon noted that "structuring the deal as a non-exclusive license may keep the fiction of competition alive," while acknowledging antitrust concerns as the primary risk.

Groq's Market Position and Technology

Groq has demonstrated strong growth, more than doubling its valuation to $6.9 billion from $2.8 billion following a $750 million funding round in September. The company distinguishes itself by not using external high-bandwidth memory chips, instead employing on-chip memory called SRAM.

Groq's primary competitor using this approach is Cerebras Systems, which Reuters reported plans to go public as soon as next year. Both companies have secured large deals in the Middle East market.

Future Outlook

Nvidia CEO Jensen Huang has emphasized the company's commitment to maintaining its leadership position as AI markets shift from training to inference. The Groq deal represents a significant step in this strategic direction, combining Nvidia's market dominance with Groq's specialized inference technology and experienced leadership team.