The technology sector is currently experiencing a historic AI stock bifurcation. While semiconductor giants are reaching all-time highs, many established software companies are struggling to maintain their valuations.
This “inner fight” in tech has created a massive performance gap. For example, the iShares Semiconductor ETF (SOXX) has skyrocketed by over 150% in the last year, while the iShares Expanded Tech-Software Sector ETF (IGV) has decline 8% over the same time period.
As an investor, understanding why this split is happening is crucial for protecting your portfolio. We are witnessing a transfer of value from the application layer to the infrastructure layer.
The Rise of Infrastructure: Why Hardware is King
Currently, the market is in the “Build” phase of the AI revolution. Companies are racing to build the data centers required to train Large Language Models (LLMs).
This has created a massive tailwind for companies that provide the “physicality” of AI. These are the “picks and shovels” of the modern gold rush.
- Nvidia (NVDA): The undisputed leader in AI training chips.
- Broadcom (AVGO): A dominant force in networking and custom AI silicon.
- Vertiv Holdings (VRT): A crucial provider of cooling and power systems for data centers.
The Software Struggle: The Fear of “Seat Compression”
The AI stock bifurcation is most painful for traditional Software-as-a-Service (SaaS) providers. The primary reason is a phenomenon known as “seat compression.”
Historically, software companies sold licenses per human user. However, as AI agents become more capable, companies may need fewer human employees to perform the same tasks. If an AI agent can do the work of five people, the software provider loses four paid “seats.”
The Potential Losers
Companies that rely on simple workflow automation or administrative tasks are at the highest risk. If a generic AI can perform the task without a specialized interface, the software becomes redundant.
- Traditional HR/Payroll: Basic administrative tools are easily disrupted.
- Generic Productivity Apps: Tools that just “organize” rather than “create” are seeing users migrate to integrated AI assistants.
The Winners of the Software Pivot
Not all software is doomed. The winners will be those who control the “System of Record” or successfully pivot to outcome-based pricing.
- Salesforce (CRM): By launching Agentforce, they are charging for “successful outcomes” rather than just logins.
- ServiceNow (NOW): Deeply integrated into corporate workflows, making them hard to displace.
- Cybersecurity (Palo Alto, CrowdStrike): AI makes attacks more frequent, making security software a non-discretionary utility.
The Next Layer: Where to Focus Now
If you feel you missed the initial semiconductor surge, look at the “Secondary Infrastructure” layer. This includes power generation and inference-specific technology.
As the world moves from training AI to running AI (inference), the demand for energy and edge computing will explode. Watch companies involved in nuclear power, electrical grid modernization, and specialized inference chips.
Companies to consider:
GE Vernova (GEV): A powerhouse in electrical grid equipment and natural gas turbines. As data centers face a “power bottleneck,” GE Vernova is seeing record orders for the infrastructure needed to connect these facilities to the grid.
Constellation Energy (CEG): The largest operator of nuclear plants in the U.S. They recently made headlines by agreeing to restart the Three Mile Island nuclear unit specifically to provide carbon-free, “always-on” power for Microsoft’s AI data centers.
Vistra Corp (VST): Another major utility player that has pivoted heavily toward nuclear and natural gas. Vistra is positioned to monetize the sheer volume of electricity consumption required as AI shifts from “training” to constant “inference.”
Eaton (ETN): An intelligent power management company. They are investing tens of millions into new U.S. facilities to produce switchgear—the essential equipment used to protect and control the massive electrical loads required by over 3,000 planned AI data centers.
Frequently Asked Questions
1. Why is software falling while chips are rising? Investors are prioritizing hardware because it is a tangible requirement for AI. Software faces uncertainty due to “seat compression” where AI replaces human users who previously paid for licenses.
2. Is it too late to buy semiconductor stocks? While valuations are high, the transition to the “Inference Era” suggests long-term demand remains strong. However, focus on companies providing cooling and power (the infrastructure) rather than just the chips.
3. Which software stocks are safe? Software companies with a deep “data moat” or those that own the “System of Record” are safer. Look for companies moving toward usage-based or outcome-based pricing models.
Previous stock market posts
Leave a Reply