The Essence of the 22% Drop in SaaS Stocks—The New Investment Risk of 'Unmodelability' and the Contradiction of Triple-Digit Monthly Token Demand Growth
"For SaaS investors, this is an absolute mess."
A managing partner appearing on Bloomberg Tech asserted this in the face of the sharp decline in software stocks. As of February 2026, the IGB (iShares Expanded Tech-Software Sector ETF) has fallen more than 22% year-to-date. However, this is not merely a cyclical economic adjustment. What investors are facing is a more fundamental problem: "not knowing how to model the traditional SaaS model."
1. AI Bubble or Transformation—The Core of the Contradiction
1-1. The Dilemma That "Both Cannot Be True at the Same Time"
Program host Ed Ludlow pointed out the fundamental contradictionsurrounding AI investment.
"Is AI transforming both the old economy and the new economy, or are we in an AI bubble? For many, both cannot be true at the same time."
In response to this question, the guest urged attention to the phenomenon of "agentic acceleration." This has the power to not just evolve AI technology, but destroy the SaaS consumption model itself.
1-2. What the "Triple-Digit Monthly Growth" of Tokens Indicates
The guest emphasized the astonishing numbers being observed at companies like OpenAI.
"Token growth rates are in the triple digits monthly. This is not quarterly or annually. It is monthly."
Despite cloud giants (Google, Microsoft, Amazon, Meta) planning a combined $650 billion in capital expenditures, the reason they cannot keep up with compute demand is this explosive token consumption.
2. "Unmodelable"—The SaaS Investor's Nightmare
2-1. How to Forecast Free Cash Flow
Traditional SaaS company valuations were based on free cash flow (FCF) multiples. However, the guest explains it this way:
"Moving away from the traditional SaaS consumption model and looking at agentic acceleration, people are, frankly, 'getting out of the way.' And they are moving into six infrastructure-related sectors."
In other words, because investors do not know if traditional SaaS models can transition to a consumption-based model, they are offloading SaaS stocks and fleeing to more certain infrastructure stocks.
2-2. What Amazon's "Mass White-Collar Layoffs" Mean
Ed Ludlow pointed out that capital expenditures are putting pressure on cash flow, and the guest cited Amazon as an example.
"Amazon laid off a 'significant number' of white-collar workers to get closer to 'neutral cash flow' during a phase where cash flow is being pressured (by AI investments, etc.)."
This indicates a structural difference between companies that can supplement FCF with advertising revenue, like Google and Meta, and those that cannot (Amazon and Microsoft).
3. The 'Memory Crisis'—The Bottleneck Within Two Years That No One Is Talking About
3-1. Memory Shortages Created by Tokens and Inference
The guest emphasized what they called 'something no one is talking about': the memory problem.
'When you consider the acceleration of tokens and the expansion of inference, we have a massive memory problem. This is at a crisis level. Over the next two years, there will not be enough memory to support this compute.'
Unlike training, AI inference requires continuous, real-time processing of massive amounts of data. As a result, demand for high-speed memory such as DRAM and HBM is exploding.
3-2. Expectations for Nvidia—What They Must Show in Their Earnings
Regarding the Nvidia earnings report coming up the following week, the guest pointed out two key points.
'First, they must reassure investors that there is no memory problem. And in fact, they don't have one. Jensen is avoiding the memory issue by securing the supply chain.'
'Second, the Blackwell numbers and the timing of the transition to the next-generation Ruben are critical. This is a real shift in infrastructure, and they need to show that it won't cannibalize their GPU and inference business.'
In other words, Nvidia cannot lead the entire market unless it proves the security of its memory supply and a smooth transition to next-generation products.
4. The Temperature Gap Between Public and Private Markets
4-1. What OpenAI's 'Platform Integration' Signifies
Ed Ludlow pointed out the dual structure of selling SaaS stocks in the public market while concentrating capital into AI companies in the private market.
What became a hot topic on social media over the weekend was OpenAI's moves. The company has begun integrating various platforms, which suggests a consolidation of consumer AI platforms.
4-2. 'The Battle Between Old and New'—Which Will Investors Bet On?
The guest summarized it this way.
'This is truly a battle between the old and the new. Those who cannot model the old (SaaS consumption model) and are watching the acceleration of agents are, frankly, 'stepping aside' and moving into infrastructure.'
In short, public market investors are shifting capital toward infrastructure that is guaranteed to grow (semiconductors, cloud, networks) rather than waiting for the uncertain transition of the SaaS model.
Conclusion—What to Believe in the Midst of 'Absolute Chaos'
The 22% drop in the IGB (Software ETF), the Bank of America survey showing a record-high response that 'companies are over-investing,' and the $650 billion capital expenditure plan by the Big Four tech companies—all of these speak to the difficulty of making investment decisions in the AI era.
The phrase 'absolute mess' repeated by the guest is not mere rhetoric. While the traditional SaaS model is collapsing, how the new consumption-based model will function remains unknown. And during this transition period, the physical constraint of memory will become apparent within two years.
What investors are facing now is not a binary choice between 'AI bubble or transformation,' but a decision on 'which layer to bet on'. The SaaS application layer or the infrastructure layer? And that judgment will be tested again by the next touchstone: Nvidia's earnings.
