FEBRUARY 14, 2026 · RESEARCH

Pushing Limits: AI Valuations & the Efficiency Frontier

Pushing Limits: AI Valuations & the Efficiency Frontier

Theme Label: Analytical Frame

Key Takeaways

Bubble framing as reductive: We view the pervasive “AI bubble” narrative as an oversimplification. While valuations are elevated—S&P 500 tech traded at ~27x forward earnings in late 2025—current multiples remain significantly below the >45x extremes of 1999.
Overcoming “GFC PTSD”: Institutional memory of 2008 continues to bias analysts toward premature terminal calls. This focus on imminent collapse ignores the role of momentum; as the adage goes, the market can stay irrational longer than many can stay solvent.
Connectivity as a valuation floor: The US ecosystem’s integration with established enterprise software (Salesforce, Notion) creates a “connectivity floor” that justifies a premium over cheaper, fragmented offshore alternatives.
Sustainability of self-funded CAPEX: Current capital intensity is fundamentally healthier than the dot-com era. Unlike the debt-fueled speculative excesses of 1999, hyperscalers are funding the build-out almost entirely through record earnings and cash flow.

Investment Environment

We view the current phase of high capital expenditure (CAPEX) as a necessary infrastructure cycle, rather than a speculative misalignment. The history of technological adoption suggests that infrastructure must precede the delivery of end-market utility. In 2024, primary hyperscalers deployed $241 billion in CAPEX; we expect this figure to reach approximately $500 billion by 2026.

Crucially, this investment is concentrated in tangible assets—data centers and semiconductor procurement—that form the physical backbone of the AI economy. Unlike the dot-com era, current spending is self-funded by established firms with high earnings quality. The market is currently pricing in the cost of the foundation, while the full efficiency gains from autonomous agents and large-scale workflow automation remain in the pipeline.

Theme 1: The Timidity Gap

Management friction, not technological ceiling, is the primary drag on AI-driven ROI. While critics point to the “monetization lag” as evidence of a bubble, we believe this misdiagnoses a “Timidity Gap” within the enterprise layer. Businesses in regulated industries have progressed beyond experimentation, yet full-scale deployment remains constrained by organizational risk-aversion and regulatory frameworks that remain “left behind” by the rapid velocity of model improvement.

We view this as a temporary bottleneck in the adoption cycle. Once internal “AI agents” move from sandbox to production, we expect a structural break in productivity.

Investment implications • Selective Exposure: We prefer enterprise software names (SaaS) that are actively bridging the “timidity gap” through pre-integrated AI workflows. • Risk considerations: Regulatory uncertainty remains a load-bearing risk; any sudden “hard” regulation could extend the monetization latency.

Theme 2: Connectivity as a Valuation Floor

Ecosystem integration provides a “Stickiness Premium” that raw model efficiency cannot bridge. The emergence of low-cost, high-reasoning models (e.g., DeepSeek R1) has led some to conclude that the U.S. valuation model is fragile. We view this as a misreading of enterprise utility. The winner is not necessarily the provider of the cheapest tokens, but the one with the deepest “pipes” into existing organizational workflows.

Organizations do not consume AI in a vacuum; they require seamless connectivity with day-to-day tools like Notion, Salesforce, and Slack. U.S. leaders have established a multi-year lead in ecosystem integration that creates a structural floor for their valuations. Until offshore competitors can replicate this API maturity and enterprise-grade security, the “Connectivity Moat” remains the dominant narrative.

Theme 3: Capital Discipline

Self-funded infrastructure differentiates the current cycle from 1999 imbalances. Critics highlight “circular financing” as a systemic risk. While we monitor this as a priority, we view it as a manageable idiosyncratic risk rather than a harbinger of collapse. During the dot-com peak, the Russell 3000 capex-to-FCF ratio reached an extreme of nearly 4.0x. As of early 2026, this ratio remains below 1.0x, reflecting a “pay-as-you-go” model for AI infrastructure. The lack of systemic financial strain suggests that the build-out is anchored in balance sheet strength, not speculative leverage.

Granular Views

Asset Class / Sector	View	Commentary
US Hyperscalers	Overweight	Strongest “Connectivity Moat.” Self-funded CAPEX and enterprise “pipes” justify the premium.
Secondary Chips / Power	Neutral	Secular demand exists, but supply chain friction and energy bottlenecks are near-term drags.
Enterprise SaaS	Selective	Preference for names crossing the “Timidity Gap” with integrated, shipped AI agents.

Big Calls

Tactical: Momentum is your friend. We believe analysts attempting to “time the top” are misreading the current cycle’s momentum. The market is rarely wrong in its direction, even if it is often early in its exuberance. We stay long the primary beneficiaries with proven ecosystem stickiness. The risk of missing the momentum-driven ride remains higher than the risk of a systemic collapse found in debt-fueled bubbles.

Strategic: Watch the Efficiency Frontier. Long-term, we monitor the “Efficiency Threat” from offshore clusters. While they currently lack the connectivity floor of the U.S., their capital efficiency represents a strategic discount threat. However, the “Connectivity Moat” remains the dominant structural filter for institutional allocation in the near-to-mid term.