Expert: Varun Rajwanshi, Senior Vice President, Lazard Asset Management Moderator: Chris Rushworth, Associate Director, Alpha Financial Management
Headlines:
- When it comes to the AI hype-cycle, investors tend to overestimate near-term and underestimate mid-term impact following a technology trigger
- The AI-arms race has triggered a front-loading of investments by hyper-scalers and internet service providers to support AI training/inference tasks
- Sustainability of this demand will be determined by the evolution and adoption of enterprise and consumer use-cases
Context:
The discussion reflected the complex interplay of technological advancements, market dynamics, adoption challenges, and the evolving landscape of AI's impact across industries and sectors, including:
AI's cost and capex problem:
The discussion highlighted concerns about the sustainability of AI's cost, especially with the significant increase in spending on AI servers and networking. AI-related capex is on the rise, potentially crowding out investments in other tech infrastructure. This poses questions about the sustainability of this spending trajectory and how it might affect non-AI capex recovery, particularly for hyper-scalers.
Deflationary and inflationary drivers:
Factors such as Moore's law driving semiconductor advancements and model optimization lead to deflationary pressures. However, the increase in model size and datasets acts as an inflationary driver, impacting the pricing considerations and ROI/payback periods for users.
Enterprise IT budgets and adoption challenges:
While AI and digital transformation are expected to impact IT budgets positively, the adoption in the enterprise sector faces barriers like ROI assessment, skill gaps, data protection, and regulatory considerations. Unlike consumer cycles, enterprise adoption tends to be slower due to more integrated workflows and longer replacement cycles for existing software.
Value-capture framework:
The value from AI's evolution is anticipated to move up the stack, from early beneficiaries like semiconductors towards software and services as use-cases and adoption expand. A similar pattern was observed in the mobile internet cycle (Semis -> Devices -> Software + Services).
Semiconductors in the AI/IoT era:
There's a focus on the dominance of semiconductor players in the AI/IoT era, particularly in edge deployments for AI inferencing. Innovation in semiconductor design and manufacturing, aimed at higher performance and energy efficiency, is expected to shape the market.
Impact on IT services:
Gen AI deployments are expected to have a net positive impact on IT services, leveraging AI advancements and increasing demands for services associated with AI integration and deployment.
Key takeaways:
- Investors often overestimate short-term impacts and underestimate mid-term effects following an AI technology trigger
- The cycle of hype and expectation doesn't always align with the actual timeline of AI's influence on markets and industries
- The AI arms race has led hyper-scalers and internet service providers to heavily invest in infrastructure to support AI tasks. This front-loading of investments may influence market dynamics, but the sustainability of this demand depends on the evolution and adoption of enterprise and consumer use-cases
- Key questions to ask are:
- Do they have a product?
- What is the market opportunity?
- What is the pricing power?
- What is the pace of adoption?
