Arm CEO Says AI Is 'Not a Bubble'

Key Concepts:

• Revenue growth (royalties and licensing)
• Cautious outlook due to licensing deal timing
• Royalty model shift to compute subsystems
• AI data center growth and potential
• Impact of tariffs and export regulations
• Customer concentration and diversification
• Demand for compute driven by AI

Financial Performance and Outlook

• Record Quarter: The company achieved a record fiscal fourth quarter with over $1 billion in revenue.
• Record Royalties and Licensing: Royalties and licensing both exceeded $600 million, contributing to a record $4 billion year.
• Growth Projections: Royalty growth is projected at 25-30% on the licensing side.
• Cautious Outlook: The company provided a cautious outlook due to the unpredictable timing of licensing deals, specifically whether deals close in June or July.
• Visibility Challenges: The company chose not to provide early guidance beyond the first quarter due to limited visibility from partners regarding end equipment shipments in the next 2-3 quarters.

Business Model and Tariffs

• Design Licensing: The company designs and licenses IP for chips used in various end products (smartphones, cars, data centers, etc.).
• Indirect Impact of Tariffs: The company is not directly impacted by tariffs, but the end product manufacturers (e.g., smartphone companies) are.
• Lack of Clarity: There is a lack of clarity from partners regarding the impact of tariffs, rather than a clear downward trend.

Royalty Model and Pricing Power

• Shift to Compute Subsystems: The company is transitioning to licensing comprehensive "compute subsystems" instead of individual IP blocks.
• Higher Royalties: Compute subsystems command higher royalties, generally double that of individual blocks.
• Customer Benefits: This approach accelerates time to market and improves design predictability for customers.

AI Data Center Growth

• Strong Momentum: The AI data center segment is experiencing strong growth.
• Resilient Spending: No signs of uncertainty in infrastructure spending from hyperscalers.
• Stargate Program: Mention of the Stargate program between SoftBank and OpenAI, focused on capital and power for data centers.
• Untapped Potential: Enterprise usage of AI is still modest, with significant potential for growth.
• Data Harnessing: Companies like J.P. Morgan have vast amounts of data (150 petabytes) that can be harnessed for valuable information and productivity gains.
• Example: OpenAI's Chatty beta used a petabyte of data to create the model.

Export Regulations and Certainty

• Air Diffusion Rule: Uncertainty surrounding the air diffusion rule and its potential impact on chip sales.
• Case-by-Case Negotiations: The possibility of countries negotiating on a case-by-case basis introduces uncertainty.
• No Bubble: The company does not believe AI demand is a bubble.
• Ubiquitous AI: AI models will run on various devices (earbuds, cars, mobile phones, cloud).

Customer Concentration and Diversification

• Concentrated Client Base: Top five customers account for 60% of revenue, with 20% of that from China.
• Industry Consolidation: The industry has seen consolidation, reducing the number of companies making chipsets.
• Hyperscaler Growth: Large hyperscalers (Microsoft, Meta, Google, Amazon) have significant capital.
• Focus on Quantity: The company is more concerned with the overall quantity of chips demanded than the number of buyers.
• AI-Driven Demand: Insatiable demand for compute is driven by AI.
• AI as an Overlay: AI compute is an addition to traditional compute (Windows, iOS, etc.), not a replacement.

Conclusion:

The company reported strong financial results, driven by growth in royalties and licensing, particularly with the shift towards compute subsystems. While a cautious outlook was provided due to the timing of licensing deals and limited visibility from partners, the overall health of the business remains strong. The AI data center segment shows significant growth potential, and the company believes AI demand is sustainable and will drive further compute needs. Customer concentration remains a factor, but the company is focused on the overall demand for chips, which is increasing due to AI.