Hardware, Cognition, Capital and Valuation in the Age of Expanding Compute
By Christopher Combs, Chief Investment Officer (CIO), Silicon Valley Capital Partners (SVCP)
Executive Summary
The current AI (Artificial Intelligence) cycle should not be viewed as a discrete technology boom or a transient earnings theme. It is better understood as a multi-decade capital expansion cycle, anchored in compute, energy, and systems integration, and reinforced by a gradual progression in machine cognition. While market narratives often oscillate between enthusiasm and skepticism, the underlying drivers of AI (Artificial Intelligence) investment remain structurally intact.
This article imparts three core conclusions. First, artificial intelligence is fundamentally constrained and enabled by hardware, making capital expenditure (CapEx) — not algorithms alone — the binding variable. Second, AI (Artificial Intelligence) cognitive development proceeds in stages, and we remain firmly in a phase where incremental advances justify continued expansion of compute and infrastructure investment (“X”), not retrenchment. Third, valuation expansion tied to AI (Artificial Intelligence) remains defensible as long as CapEx (capital expenditure) efficiency improves alongside cognitive leverage — a condition that appears increasingly likely.
Taken together, these dynamics support a high probability of a sustained AI (Artificial Intelligence) super-cycle, characterized by prolonged capital formation (“X”), rising productivity, and structurally elevated returns for firms that control compute, platforms, and distribution.
I. AI (Artificial Intelligence) Begins With Silicon, Not Software
Artificial intelligence is often discussed as a software breakthrough. In reality, it is an industrial transformation whose pace is set by semiconductor physics, power availability, and balance-sheet capacity. Every meaningful gain in AI (Artificial Intelligence) performance over the past decade has been preceded by a corresponding expansion in compute density, memory bandwidth, and system architecture.
This reality has produced a layered ecosystem of AI (Artificial Intelligence) chips, each optimized for a distinct function within the AI lifecycle: training, inference, edge deployment, autonomy, and efficiency. GPUs (Graphics Processing Units) remain the backbone of large-scale model training. Domain-specific accelerators improve cost efficiency at hyperscale. NPUs (Neural Processing Units) move intelligence to the edge. Autonomous processors enable real-world interaction. Inference-optimized silicon ensures persistence and scale.
The significance of this hardware stack is not technical alone. It defines the capital intensity of the AI (Artificial Intelligence) cycle. Unlike prior software revolutions, AI requires sustained, front-loaded investment in physical assets. This characteristic alone argues against a short or easily exhausted cycle.
II. Capital Expenditure (CapEx) as the Core Variable (“X”)
Consider viewing AI (Artificial Intelligence) primarily through the lens of capital formation. The defining feature of the current environment is not speculative enthusiasm, but continued expansion of CapEx (capital expenditure) (“X”) across the AI value chain.
Compute build-outs, data centers, advanced packaging, networking, and power infrastructure represent long-duration investments with multi-year lead times. Importantly, this CapEx (capital expenditure) is not discretionary in the traditional sense. Firms competing at the frontier of AI (Artificial Intelligence) are compelled to invest simply to maintain relevance. Underinvestment increasingly represents strategic risk.
Historically, technology cycles that demand persistent capital formation — railroads, electrification, telecommunications, the internet — tend to outlast skepticism, not succumb to it. AI (Artificial Intelligence) exhibits similar characteristics, with the added feature that incremental investment continues to produce measurable gains in performance and efficiency.
This creates a reinforcing loop: investment begets capability, capability begets adoption, and adoption justifies further investment.
III. The Three Stages of AI (Artificial Intelligence) Cognitive Ability
While hardware defines the pace of AI (Artificial Intelligence) progress, cognition defines its economic impact. AI development can be framed through three stages of cognitive capability: narrow intelligence, general intelligence, and superintelligence.
Stage One: Narrow Intelligence (ANI – Artificial Narrow Intelligence)
ANI (Artificial Narrow Intelligence) systems perform specific tasks with extraordinary efficiency but lack the ability to reason across domains. This is the phase we occupy today. It is already economically meaningful, driving productivity gains, automation, and revenue expansion across sectors.
Crucially, ANI (Artificial Narrow Intelligence) benefits directly from scale. More data and more compute reliably improve outcomes. This property alone supports continued expansion of CapEx (capital expenditure), as returns on incremental investment remain positive.
Stage Two: General Intelligence (AGI – Artificial General Intelligence)
AGI (Artificial General Intelligence) refers to systems capable of cross-domain reasoning, abstraction, and autonomous planning. While full AGI (Artificial General Intelligence) has not yet arrived, early manifestations — including multi-step reasoning agents and workflow coordination — are emerging.
Importantly, the path toward AGI (Artificial General Intelligence) does not require a sudden breakthrough. It is likely to unfold gradually, with incremental cognitive gains that continue to reward compute investment. Each step toward greater reasoning capacity expands the economic leverage of AI (Artificial Intelligence), increasing addressable markets and reinforcing the case for sustained CapEx (capital expenditure).
Stage Three: Superintelligence (ASI – Artificial Superintelligence)
ASI (Artificial Superintelligence) remains theoretical and should not be treated as a base-case assumption. Its relevance lies primarily in shaping long-term policy, governance, and risk discussions. It is not required to justify today’s investment cycle or current capital formation.
IV. Valuation Regimes Across the Cognitive Timeline
Valuation expansion tied to AI (Artificial Intelligence) must be understood in context. Markets do not price cognition directly; they price cash flows, durability, and optionality.
ANI (Artificial Narrow Intelligence) Valuation Regime
In the ANI (Artificial Narrow Intelligence) phase, valuation expansion reflects confidence in long-duration growth and platform durability. Margins improve gradually, while CapEx (capital expenditure) remains elevated. This supports higher-than-historical multiples, particularly for firms with scale, balance-sheet strength, and pricing power.
Transition Regime: ANI (Artificial Narrow Intelligence) to AGI (Artificial General Intelligence)
As AI systems approach early forms of AGI (Artificial General Intelligence), CapEx intensity may peak, but so does valuation dispersion. Markets increasingly price optionality — the possibility that cognitive leverage accelerates faster than expected. This phase favors platform owners and penalizes under-scaled participants.
AGI (Artificial General Intelligence) Valuation Regime
Should AI achieve durable AGI (Artificial General Intelligence), the economic model shifts. Labor substitution accelerates, decision-making costs compress, and free cash flow inflects materially. At that point, valuation expansion would be driven less by growth narratives and more by structural margin transformation.
Even partial realization of AGI-like capabilities would justify a sustained re-rating.
V. Why the AI (Artificial Intelligence) Super-Cycle Is Likely to Persist
Several factors support a continued AI (Artificial Intelligence) super-cycle.
First, capital commitment is increasingly irreversible. Much of the infrastructure being built has multi-decade utility and becomes a platform for successive innovation waves.
Second, compute efficiency continues to improve. Gains in performance per watt and per dollar increase returns on CapEx (capital expenditure), extending the cycle rather than truncating it.
Third, cognitive progress is incremental but compounding. Each step toward greater reasoning ability expands AI’s economic footprint without requiring a binary breakthrough.
Finally, global competition reinforces investment discipline. AI (Artificial Intelligence) has become a strategic asset. Competitive dynamics alone ensure continued spending independent of near-term macro cycles.
VI. Conclusion
The AI (Artificial Intelligence) cycle should not be evaluated through the lens of prior software booms. It is a compute-driven, capital-intensive super-cycle, reinforced by gradual cognitive progress and sustained by structural demand for productivity.
We do not need ASI (Artificial Superintelligence) to justify continued expansion of CapEx (capital expenditure) (“X”). ANI (Artificial Narrow Intelligence) already delivers returns, while early steps toward AGI (Artificial General Intelligence) increase upside optionality. As long as incremental investment produces incremental capability — a condition that remains firmly in place — the probability of a prolonged AI (Artificial Intelligence) super-cycle remains high.
Footnotes
- Stanford University, AI (Artificial Intelligence) Index Report, 2024–2025 Editions.
- McKinsey Global Institute (MGI – McKinsey Global Institute), The Economic Potential of Generative AI (Artificial Intelligence), 2024.
- NVIDIA, AMD, and hyperscaler CapEx (capital expenditure) disclosures, 2023–2026.
- Bloomberg Intelligence, AI (Artificial Intelligence) Infrastructure and Semiconductor Supply Chain Analysis, 2025.
- OECD (Organisation for Economic Co-operation and Development), Artificial Intelligence, Productivity, and Growth, 2024.