1997-Style AI Phase: The AI boom appears analogous to 1997, where transformative technology was real but capital intensity and infrastructure overbuild created hidden investment risks.
AI as a Utility: AI is becoming a universal productivity tool, with the biggest long-term winners likely to be traditional firms using AI to structurally improve margins rather than pure AI sellers.
Disciplined Playbook: The recommended strategy is participation with rigor—maintain target allocations, rebalance gains, and emphasize quality beneficiaries over speculative “arms dealers.”
Seven-Month Win Streak Snapped: The S&P 500 snapped its multi-month winning streak, but the pause came alongside a healthy rotation in market leadership rather than a deterioration in underlying fundamentals.
Yield Curve Normalization: December saw wider yield spreads as the Fed cut rates by 25 bps, pushing short-term yields lower while intermediate and long-term rates moved higher—continuing the long-discussed normalization of the yield curve into year-end.
Rates Enter 2026 on Firm Ground: Despite a divided Fed and lingering policy uncertainty, long-term yields rose within forecasted ranges, leaving U.S. bond markets on solid footing and a normalized curve that offers greater flexibility heading into 2026.
Executive Summary
As we transition into 2026, the parallels between the current Artificial Intelligence (AI) boom and the technology cycle of the late 1990s have moved beyond mere observation to a structural framework for portfolio management. While the market faces stretched valuations and macro “fog,” the core question for investors is no longer if AI is transformational, but where we sit in the cycle. By examining the “Efficiency Trap” of 1997 and the tactical rebalancing needs of 1999, we provide a roadmap for navigating a phase that demands both participation and extreme discipline.
The “1997 Phase”: Innovation vs. Investment Reality
Our central thesis is that we are likely in a “1997-style” phase of the cycle. This is a period where the fundamental promise of a technology (the Internet then, AI now) is being realized, but the infrastructure “arms race” creates hidden risks for investors.
A critical historical parallel is the collapse of Global Crossing. They were correct that internet traffic would explode, but they failed to account for a supply-side breakthrough: Dense Wavelength Division Multiplexing (DWDM). This invention expanded fiber capacity by nearly 10,000% per cable, commoditizing bandwidth overnight. The investment case vanished even as the technology won.
In 2026, we see a similar risk. The “Magnificent Seven” are projected to invest $450 billion in AI infrastructure this year. However, a “DWDM-equivalent” breakthrough—such as a massive leap in algorithmic efficiency or specialized “edge” silicon—could suddenly expand compute capacity to a point where the unit price of AI processing collapses, challenging the returns on these massive capital expenditures.
The Universal Beneficiary: Every Company is an “AI Company”
Just as the late-90s boom eventually led to the realization that “every company is a tech company,” we are now entering an era where AI is not a standalone sector, but a universal utility.
The Productivity Multiplier: Our research has already identified a notable rise in productivity as firms leverage technology to optimize SG&A and improve margins. This is the “hidden” bull case: the greatest long-term winners may not be the companies selling the chips, but the “traditional” companies—in healthcare, manufacturing, and services—that use AI to structurally lower their cost of doing business.
The “Netscape” Risk: We expect a shakeout similar to the early 2000s. While we will all be beneficiaries of the trillions spent on AI infrastructure, many “AI-first” startups (the Netscapes and Pets.com of this era) will likely disappear as their specialized functions are absorbed into broader platforms like Alphabet’s Gemini or internal corporate tools.
The 1997 Echo
The 1997 Echo
Performance Review & Outlook
Highlights
Executive Summary
As we transition into 2026, the parallels between the current Artificial Intelligence (AI) boom and the technology cycle of the late 1990s have moved beyond mere observation to a structural framework for portfolio management. While the market faces stretched valuations and macro “fog,” the core question for investors is no longer if AI is transformational, but where we sit in the cycle. By examining the “Efficiency Trap” of 1997 and the tactical rebalancing needs of 1999, we provide a roadmap for navigating a phase that demands both participation and extreme discipline.
The “1997 Phase”: Innovation vs. Investment Reality
Our central thesis is that we are likely in a “1997-style” phase of the cycle. This is a period where the fundamental promise of a technology (the Internet then, AI now) is being realized, but the infrastructure “arms race” creates hidden risks for investors.
A critical historical parallel is the collapse of Global Crossing. They were correct that internet traffic would explode, but they failed to account for a supply-side breakthrough: Dense Wavelength Division Multiplexing (DWDM). This invention expanded fiber capacity by nearly 10,000% per cable, commoditizing bandwidth overnight. The investment case vanished even as the technology won.
In 2026, we see a similar risk. The “Magnificent Seven” are projected to invest $450 billion in AI infrastructure this year. However, a “DWDM-equivalent” breakthrough—such as a massive leap in algorithmic efficiency or specialized “edge” silicon—could suddenly expand compute capacity to a point where the unit price of AI processing collapses, challenging the returns on these massive capital expenditures.
The Universal Beneficiary: Every Company is an “AI Company”
Just as the late-90s boom eventually led to the realization that “every company is a tech company,” we are now entering an era where AI is not a standalone sector, but a universal utility.
Dreams are important.
Aspirations are what help make goals reality.