Fierce Debate Between Bulls and Bears: The Current Core Controversies of AI, as Well as the Winners and Losers

The fierce debate on Wall Street reveals that the moat of large language models is being backfired by open-source ecosystems and price wars, with core value shifting sharply to computing power and data ecosystems. Under the brutal reshuffle, Nvidia, Microsoft and Apple remain firmly at the winning end, while SaaS giants relying on past achievements and highly valued Intel are being ruthlessly cleared by the market. A wealth knockout that reshapes the tech landscape has fully kicked off! 

When the trillion-dollar infrastructure frenzy collides with the "price war" surprise attack of low-cost large models, is artificial intelligence (AI) a trillion-dollar wealth feast comparable to the Fourth Industrial Revolution, or a capital bubble about to usher in painful adjustments?

In the latest issue of the Real Eisman Playbook interview, Steve Eisman, the prototype of The Big Short, talks with well-known Wall Street bull Dan Ives and rational valuation analyst Gil Luria, focusing on core controversies such as AI return on investment (ROI), the collapse of large model moats, and the SaaS doomsday theory. In this unprecedented major AI divergence, Nvidia, Microsoft, Apple and the undervalued Micron are firmly occupying the winning position, while traditional SaaS giants that lack innovation and only rely on price hikes, Intel with inflated valuations, and edge software burdened by private equity (PE) are gradually moving to the elimination edge of being cleared by the market.

01 Core Key Takeaways

◦ Capital expenditure and ROI debate: The AI industry has shifted from a light-asset model to a high capital-density model. At present, the AI industrial chain has generated more than 100 billion US dollars in annualized monetized revenue. Although it is still insufficient compared with trillion-dollar capital investment, it has proved the existence of real economic demand.

◦ Collapse of model moats and price wars: Large Language Models (LLMs) are rapidly moving towards homogenization and commoditization. The core value in the future will no longer lie in the model itself, but in data, computing power infrastructure and ecosystem entry points.

◦ Open-source ecosystem and closed-source risks: Closed-source models are facing full erosion from the open-source ecosystem. Excessively binding core enterprise data to a single closed-source model will easily lead to data leakage risks and operational passivity.

◦ Real divergence of the SaaS doomsday theory: AI will not destroy all software, but will trigger drastic differentiation. SaaS manufacturers that lack R&D investment and only rely on price hikes to maintain growth, as well as unlisted small software companies controlled by private equity (PE), will be cleared by enterprise budgets; while enterprises such as Microsoft and Palantir that control core data and ecosystems will emerge as winners.

◦ Review of winning and losing positions: Winners: Nvidia (absolute core "water seller"), Microsoft (dual defense of cloud and ecosystem), Apple (consumer-grade AI "pass"), Micron (severely undervalued storage giant), Palantir (data moat) and the cybersecurity sector (CrowdStrike/Palo Alto). Losers/risk targets: Salesforce (no added value, only price hikes), Intel (severe mismatch between high valuation and weak fundamentals), unlisted edge software held by PE, and traditional software directly impacted by AI substitution (Intuit/Adobe).

02 Capital Frenzy and the Big ROI Question: Two Core Controversies in the AI Industry

A year ago, there were almost no bearish arguments when the market discussed AI, but today a year later, skepticism has begun to dominate the mainstream.

Gil Luria points out that the most core debate in the current tech industry is whether extremely high capital expenditures can deliver sufficient return on investment. Although the industry-wide annualized AI monetization revenue has exploded from zero to hundreds of billions of dollars, proving the real existence of economic demand, this return is still in its early stage compared with the trillions of dollars of capital expenditures laid out.

Steve Eisman also adds that Google, Microsoft and Meta, which never relied on external financing in the past, have now become high capital-density enterprises due to the high investment in AI, which in itself is a hidden worry at the financial level.

However, Dan Ives holds a completely different long-term perspective. He believes that humanity is currently in the third year of an eight-to-ten-year AI infrastructure revolution, which is like building the Las Vegas Strip in 1955, and there will inevitably be several violent fluctuations that test the market's courage. Ives points out that according to the latest survey of his Asian supply chain, the chip supply-demand ratio is still as high as 15:1, indicating that the underlying demand is far from saturated, and the United States has once achieved absolute leadership over global competitors in technology and infrastructure.

03 Collapsing Moats and Open-Source Counterattack: Large Models Are No Longer a "One-Man Show"

In response to Eisman's concern that "large models lack moats and low-cost models trigger price wars", analysts reveal the deep structure of the AI industrial chain.

Gil Luria says that the AI industrial chain consists of equipment vendors, chip vendors, cloud computing power providers, and underlying model vendors. Model homogenization and even price wars mainly impact the most downstream model companies, while the value of upstream equipment, chips and computing power providers will not be weakened. Especially when hardware giants such as Nvidia begin to actively launch free open-source models to the market, the pricing power of closed-source large model vendors is rapidly disappearing.

In addition, the guests also specifically cited the warning from Alex Karp, CEO of Palantir.

In the enterprise market, data control is far more important than the model itself. If enterprises directly expose core data to third-party closed-source large models, they will not only face the risk of commercial secret leakage, but also easily get locked by the underlying model. The inevitable future trend will be "model-agnostic", that is, models will gradually become general basic tools, and the knowledge architecture and application ecosystem built around the enterprise's own data are the real moats.

04 Divergence of the SaaS Doomsday Theory: Who Is Creating Panic, and Who Is Dominating the Market?

Facing the controversy of "whether AI will destroy traditional software", the two analysts give a very detailed differentiated judgment.

Luria points out that the so-called unemployment panic and the remarks spread by large model manufacturers that "AI will replace white-collar workers" are, to some extent, a strategy by the top founders of large models to push the government to introduce strict regulations, thereby "pulling up the ladder behind them" and blocking competition. From the perspective of historical microeconomics, after technological progress improves productivity, it will often drive the simultaneous growth of enterprise benefits and the number of jobs, rather than destroying overall employment.

But the software industry is indeed undergoing drastic reshuffling. The market is not trying to destroy all software, but to clean up low-quality software.

Taking Microsoft as an example, even if AI Agents are fully popularized, users still need underlying infrastructure such as Outlook, Teams and Office. Microsoft will not be subverted at all, but will instead achieve further growth by empowering cloud services and office suites with AI.

In contrast, companies such as Salesforce have failed to provide obvious added value to customers for many years, and only rely on unilateral price increases to maintain performance growth. When enterprises reallocate IT budgets for AI, such software lacking core moats will inevitably be eliminated first.

05 The Final Winning Move: Tech Winners and Losers in the Eyes of Wall Street

Based on the above analysis, the two analysts sort out a clear list of winning and losing positions.

In the winner camp, Nvidia still occupies a unique core "water seller" position.

Ives believes that Nvidia's leading edge in chips is irreplaceable, and every 1 dollar spent on Nvidia's chips can bring a multiplier effect of 8 to 10 dollars to the entire tech ecosystem.

Apple has shown an extremely smart strategic posture, which Ives compares to the "pass" for consumer-grade AI. Apple does not need to bear the risk of hundreds of billions of dollars in computing power infrastructure investment. As long as it firmly controls the entry points of 2.5 billion terminal devices, no matter who wins the model battle in the future, Apple can smoothly integrate the most powerful models and realize ecosystem monetization.

Micron Technology shows extremely significant valuation mismatch. Against the background of the booming memory chip industry, its 6x P/E ratio implies an extremely pessimistic wrong expectation, with a very high risk-reward ratio.

In addition, Palantir and the cybersecurity sector whose demand has doubled due to the expanded attack surface brought by AI (such as CrowdStrike and Palo Alto) are also in the winning position.

In the loser and risk camp, in addition to Salesforce that has no added value and only raises prices, Intel, which has a severe mismatch between inflated valuation and weak fundamentals, is also widely questioned.

What is more hidden is the small and medium-sized unlisted software companies controlled by private equity (PE) funds. Since PE institutions often cut R&D to extract cash flow after acquisition, these innovative-edge software companies are facing mass closure in the face of the reality that enterprise CIOs are drastically reducing the number of software suppliers. At the same time, the inherent business models of traditional tax software Intuit and content creation software Adobe are also facing the deconstruction threat from the underlying large models.

The full text of the interview is as follows:

Steve Eisman:

Hi, I'm Steve Eisman, welcome to a new episode of The Real Eisman Playbook. The most intense debate right now is obviously around artificial intelligence (AI) — how much profit can it actually generate? How sustainable is it? And the focus of this debate changes almost every week, which is really incredible.

So today I've invited two top analysts in the tech field: Dan Ives, who just left Wedbush and founded his own investment bank; and Gil Luria from D.A. Davidson.

The reason I love these two analysts is that their coverage is extremely broad. The vast majority of tech analysts only focus on chips, or hardware devices, or software, but these two cover almost every aspect of the tech industry. Therefore, I believe they can bring profound insights into the breadth and depth of this debate.

At the end of the show, I will come back to share some concluding thoughts. But before we start, if you enjoy our interviews and weekly recaps, the best way to support the show is to subscribe to us for free on YouTube and Substack.

Steve Eisman:

Hi, I'm Steve Eisman, welcome to a new episode of The Real Eisman Playbook. There are so many major events happening in the tech industry every week. As I often say in my weekly notes, I have never seen any industry develop so fast that every piece of news is not just an incremental addition, but almost reshaping the entire investment logic.

Today we have two guests. The first one is our old friend Dan Ives.

Dan Ives:

It's great to be back here again, Steve.

Steve Eisman:

The second is our new guest, Gil Luria.

Gil Luria:

Thank you very much.

Steve Eisman:

Dan is preparing for his new project, and we won't go into details here; Gil is currently at D.A. Davidson.

Gil Luria:

Yes.

Steve Eisman:

One of the main reasons I invited you two is that most people only focus on semiconductors or software, but your coverage is very broad, and the impact of what is happening now is really far-reaching.

Before I hand over the floor to you, I want to make a few points. If a year ago, we stood here discussing AI, I know Dan would be extremely optimistic, he would talk nonstop about how amazing Nvidia's revenue growth is, how fast hyperscalers are developing, and you could hardly find any bearish logic back then.

Today, a year later, it's not that easy to find a bearish argument. So Gil, I'd like to hand the floor to you first. Could you spend a few minutes giving us a high-level summary: what is the core focus of the current debate? What is your position in this debate?

Gil Luria:

No problem. There are two core debates going on in the tech industry right now:

First, can we get returns from all these huge investments? Can all these data centers under construction and the huge funds deployed generate a return on investment (ROI) for the relevant companies that justifies this extreme level of expenditure? We must objectively admit that this expenditure is extremely drastic and unprecedented.

Steve Eisman:

It's truly unprecedented.

Gil Luria:

Yes, unprecedented. This is the first major debate that is being tugged back and forth right now, and we have some nuanced views on it.

Then let's look at the other side of the debate: How will AI affect all other companies, especially software companies? What will the life of software companies be like in the AI era? This is the so-called "SaaS (Software as a Service) Doomsday Theory".

Steve Eisman:

Exactly, SaaS doomsday.

Gil Luria:

The so-called "SaaS Doomsday Theory" is a view that all software companies are doomed — that "there's nothing left to see in this industry, they will all be dead in five years, and software will cease to exist".

But now we have a more nuanced discussion, and we can talk about our position on this later. But overall, these are the two core debates.

We can combine these two points to look at a company like Microsoft. Microsoft is suffering on both ends:

On the one hand, some people say: "Look, you built so many data centers and you can't get any return on investment! Since AI is worthless and AI is a bad thing, that means you're wasting a lot of capital at the same time."

But at the same time, others say: "Look, you're a software company, and AI is so powerful that it will directly destroy your existing business!"

So Microsoft is taking hits from both sides. But we think: wait a minute, I can fully prove to you why I think we are getting good ROI, and ROI will improve further. Therefore, it makes sense for them to keep investing.

Then I can also prove: Five years from now, I will still wake up in the morning, turn on my computer, log in to Outlook, use Teams, use PowerPoint, Excel and Word. By the way, there will be AI Agents helping me with Excel, Word, Outlook and Teams by then, but I will still be there. Guess who is standing in front of all these models when this happens? Microsoft.

So these are the two major debates happening right now, and Microsoft just happens to be the "punching bag" caught in the middle. That's exactly what makes the current market so interesting.

Steve Eisman:

Dan, that's a brilliant summary. It seems you have nothing to add?

Dan Ives:

Haha, here's how I see it: We are currently in the third year of an 8 to 10 year construction period of the AI revolution. In my opinion, it's a bit like building the Las Vegas Strip in 1955.

In this context, there will inevitably be all kinds of questions: When will capital expenditures actually translate into monetization? Will Anthropic (the AI unicorn) take food off everyone else's plate? Are valuations too high? Is this a repeat of the 1999-2000 internet bubble, or is it the real Fourth Industrial Revolution?

I obviously firmly believe it's the latter. Therefore, I think there will be three to four violent moments of volatility every year that I call "gut check" moments.

But if we step back, during our recent trip to Asia, the chip supply-demand ratio was as high as 15 to 1. I'm usually not trapped by the narratives in the market. If you got caught up in those narratives a year ago — like when taxi drivers in New York were all bearish on Alphabet (Google's parent company), shouting "AI will destroy search engines, the DOJ will break up Google" — you'll find that we are now in a period of narrative reshaping.

For example, when memory chip prices skyrocket, or pull back sharply after the SK Hynix deal, it does trigger these white-knuckle moments. But in my opinion, this technology will change society in positive ways. I believe AI will create more jobs than it destroys.

Many times in my life, I just flew back to New York airport from a faraway place, saw all kinds of penny-pinching scenes, and then had to fly back to Taiwan, China to visit those fabs where workers work 18 hours a day.

As for the gap you see between Asia and here, I think that gap has narrowed dramatically, and now it's America's turn to take the driver's seat.

Steve Eisman:

Then let me press you two on this. I'll raise three counterarguments, and you tell me what you think.

First, the problem is not just that they spent too much money, but that these companies involved have barely raised external capital since their inception. Google went public in the early