AI Build-Out: Underinvestment and Thematic Precision for Investors

The AI build-out is not a sprint, but a decades-long marathon, and current market sentiment may be dangerously underestimating its true scale and demand. While many investors flock to traditional tech sector ETFs, believing they are gaining exposure to artificial intelligence, they might be missing the more precise, targeted opportunities that thematic investing offers. This conversation with Jay Jacobs of BlackRock reveals that the current surge in AI capital expenditure, though significant, is merely the prologue to a much larger story. The hidden consequence? A potential missed opportunity for those who fail to grasp the long-term, structural shift AI represents, mistaking early stages for peak maturity. Investors, particularly those managing portfolios or seeking to refine their market approach, will find an advantage in understanding the multi-layered demand and infrastructure requirements that traditional sector allocations may not fully capture.

The AI Infrastructure Boom: An Underinvestment in Disguise?

The prevailing narrative around AI investment often centers on fears of overspending. Yet, the data presented by Jay Jacobs suggests a starkly different reality: a potential underinvestment in AI infrastructure. While capital expenditures are accelerating, the demand for AI services, particularly token consumption, is growing at an even more astonishing rate--seventeen times last year. This disparity between supply and demand is the first layer of consequence. It implies that the current build-out, which might seem substantial, is struggling to keep pace. The market is shifting from worrying about companies overspending to questioning if they are spending enough.

"Essentially, as much money as the major large language model providers are plowing into capital expenditures, they can't keep up with AI demand. So even just in the last several months, I think the narrative has shifted in the market from that of, 'We worried companies are overinvesting in capex,' to, 'What if companies are actually underinvesting in capex?'"

This dynamic challenges conventional wisdom. The typical investor might see massive capex figures and assume saturation is near. However, Jacobs highlights that AI capex as a percentage of GDP is still relatively low compared to historical infrastructure booms like railroads or electricity. This suggests we are in the early innings of a multi-year, potentially multi-decade, build-out. The immediate consequence of this underinvestment perception is a potential bottleneck: powerful AI models might have to be throttled due to insufficient compute power. The long-term advantage lies with those who recognize this early stage and position for sustained growth, rather than assuming the current investment is the peak.

Thematic Precision: Navigating Beyond Sector Allocations

A critical insight emerges from the distinction between sector investing and thematic investing. Many investors believe allocating to the technology sector provides adequate AI exposure. However, Jacobs points out that the tech sector is a mixed bag; it contains AI builders but also software companies potentially threatened by AI's disruptive business models. This is where the second layer of consequence unfolds: imprecise exposure can lead to unintended risks.

Thematic ETFs, on the other hand, offer a more targeted approach. They allow investors to fine-tune their exposure to specific, disruptive forces like AI. The data shows a significant gap between BlackRock's internal model allocations to thematic ETFs and the average advisor portfolio. This gap represents a missed opportunity for advisors and retail investors to gain more precise exposure to structural growth trends. The delayed payoff here is significant: by using thematic ETFs, investors can better capture the specific growth drivers of AI without being diluted by unrelated or even negatively impacted companies within a broader sector. Conventional wisdom suggests broad sector exposure is sufficient, but this overlooks the nuanced impact of disruptive technologies.

The Exponential Surge of Agentic AI and the Tech Stack

The conversation then dives into the potential impact of agentic AI--AI capable of completing multi-step tasks autonomously. The report highlights that this can increase token intensity by a thousand times. This is a profound consequence, as it dramatically amplifies the demand for AI infrastructure across the entire tech stack. The beneficiaries are not limited to one segment but span from the foundational layers of power and hardware to the semiconductors, data, large language models, and finally, the applications and products built upon them.

This exponential demand surge creates a cascading effect. Companies providing the underlying infrastructure--data centers, semiconductors (memory, GPUs, CPUs), and power--will see increased demand. The data layer, including proprietary data training models, becomes more valuable. The LLMs themselves will be in higher demand, and the software layer supporting them will evolve rapidly. Finally, applications leveraging AI agents for tasks like financial analysis or report consolidation will proliferate. The advantage for investors lies in understanding this entire value chain. Those who can identify the key enablers at each layer, from raw compute to end-user applications, are better positioned to benefit from the sustained growth driven by agentic AI.

Infrastructure's Quiet Strength and the Rise of Tokenization

Beyond AI, the discussion touches upon global infrastructure investment, projected to exceed $100 trillion by 2040. Despite this massive capital rollout, infrastructure remains a surprisingly small allocation in many portfolios, with the S&P 500 averaging only about 3%. This presents another instance where conventional thinking may be misaligned with long-term trends. The drivers for infrastructure are numerous: growing populations, aging infrastructure in developed markets, and the increasing need for digital infrastructure. The consequence of this underallocation is missing out on a fundamental, long-term growth driver.

The conversation also explores tokenization and digital assets, exemplified by the rapid growth of the iShares Bitcoin Trust ETF. This trend signifies a growing demand for assets that behave differently from traditional stocks and bonds, acting as a hedge against geopolitical uncertainty, institutional distrust, and currency debasement. The promise of tokenization--24/7 trading, instantaneous settlement, and easier access to DeFi tools like lending--could reshape access, liquidity, and transparency. However, the development of a robust ecosystem, including market-making capabilities and sensible regulation, is crucial. The delayed payoff here is the potential for tokenization to unlock new forms of liquidity and access across a broader range of real-world assets, creating opportunities for investors who understand its evolving landscape.

Actionable Insights for the Forward-Thinking Investor

• Re-evaluate AI Exposure: Move beyond broad technology sector allocations. Investigate thematic ETFs that specifically target AI infrastructure, semiconductors, and AI-driven applications. Immediate Action.
• Understand Token Consumption: Recognize that token consumption growth (17x) significantly outpaces capex investment, signaling potential supply constraints and sustained demand. Immediate Action.
• Consider Infrastructure as a Core Holding: Given the projected $100 trillion investment by 2040, increase allocations to infrastructure, recognizing its fundamental role in global development and its current underrepresentation in portfolios. Immediate Action.
• Explore Agentic AI Beneficiaries: Map the AI tech stack and identify companies poised to benefit from the exponential increase in demand driven by agentic workloads, from hardware providers to software developers and application creators. Immediate Action.
• Monitor Tokenization Ecosystem Development: Stay informed about the regulatory and infrastructure developments in tokenized real-world assets, as this trend has the potential to unlock new liquidity and investment opportunities. Ongoing Monitoring.
• Embrace Long-Term Thematic Investing: Adopt a multi-year, even multi-decade, perspective on themes like AI, robotics, and healthcare innovation, understanding that significant payoffs often require patience and a commitment beyond short-term market fluctuations. Long-Term Investment Strategy (12-18 months+).
• Prepare for Infrastructure Bottlenecks: Anticipate potential bottlenecks in AI compute and power grids due to demand outpacing supply. This requires strategic investment in companies that can scale to meet this demand, creating an advantage for those who invest ahead of the curve. This pays off in 12-18 months as demand continues to surge.