The question on everyone’s minds and lips is: are we in an AI bubble?
This is the wrong question. The real question is: Who Are we in an AI bubble, and when will each one burst?
The debate over whether AI represents a transformative technology or an economic time bomb has reached a fever pitch. Even tech leaders like Meta CEO Mark Zuckerberg have acknowledged evidence of an unsustainable financial bubble forming around AI. OpenAI CEO Sam Altman and Microsoft co-founder Bill Gates see the obvious bubble dynamics: overexcited investors, frothy valuations and lots of doomed projects — but they still believe AI will ultimately transform the economy.
but are treating "aye" has been fundamentally misguided as a single monolithic entity destined for uniform collapse. The AI ecosystem is actually three distinct layers, each with different economics, defensibility, and risk profiles. It is important to understand these layers, as not all of them will open at once.
Layer 3: Wrapper Companies (First to Fall)
The Weakest Class isn’t building AI – it’s reengineering it.
These are companies that take OpenAI’s API, add a simple interface and some quick engineering, then charge $49/month for the equivalent of a glorified ChatGPT wrapper. Some have achieved rapid early success, such as Jasper.AI, which reached nearly $42 million in annual recurring revenue (ARR) in its first year by incorporating the GPT model into a user-friendly interface for marketers.
But the cracks have already started appearing. These businesses face threats from every direction:
feature absorption:Microsoft may bundle your $50/month AI writing tool into Office 365 tomorrow. Google could make your AI email assistant a free Gmail feature. Salesforce can build your AI sales tools natively into their CRM. When big platforms decide your product is a feature, not a product, your business model evaporates overnight.
web of commodification: Wrapper companies are essentially just passing input and output, if OpenAI improves prompting, these tools lose value overnight. As foundation models become more similar in capability and pricing continues to decline, margins become zero.
zero switching costs: Most wrapper companies don’t have proprietary data, embedded workflows, or deep integrations. A customer can switch to a competitor or directly to ChatGPT in minutes. There is no moat, no lockout, no security.
The white-label AI market exemplifies this fragility. Companies using white-label platforms face vendor lock-in risks from proprietary systems and API limitations that can hinder integration. These businesses are building on rented land, and the landlord can change the terms at any time, or bulldoze the property.
the exception that proves the rule:Cursor stands out as a rare wrapper-layer company that has built a real security capability. By integrating deeply into developer workflows, building proprietary features beyond simple API calls, and establishing strong network effects through user habits and custom configuration, Cursor has demonstrated how a wrapper can evolve into something more significant. But companies like Cursor are outliers, not the norm – most wrapper companies lack this level of workflow integration and user lock-in.
Time: Significant disruptions are expected in this segment from late 2025 to 2026, as larger platforms absorb functionality and users realize they are paying premium prices for commoditized capabilities.
Layer 2: Foundation Model (Middle Way)
The companies making LLM – OpenAI, Anthropic, Mistral – are in a more defensive but still uncertain position.
Economic researcher Richard Bernstein points to OpenAI as an example of the bubble dynamic, noting that the company has secured nearly $1 trillion in AI deals, including a $500 billion data center buildout project, despite only aiming to generate $13 billion in revenue. Difference between investment and earning potential "Definitely looks bubbly," Bernstein notes.
Still, these companies have real technical advantages: model training expertise, compute access, and performance advantages. The question is whether these advantages are sustainable or whether the models will become commoditized to the point where they will be indistinguishable – turning foundation model providers into low-margin infrastructure utilities.
Engineering will separate the winners from the losers: As foundation models transform into foundational capabilities, competitive edge will increasingly come from predictive optimization and systems engineering. Companies that can increase the memory wall through innovations such as extended KV cache architecture, achieve better token throughput and provide faster time-to-first-token issuance will capture premium pricing and market share. The winners will not only be those who have the greatest training, but also those who can make AI inference economically viable at scale. Technological breakthroughs in memory management, caching strategies, and infrastructure efficiency will determine which leading laboratories will survive consolidation.
Another concern is the cyclical nature of investments. For example, Nvidia is investing $100 billion in OpenAI to bankroll data centers, and OpenAI is then filling those facilities with Nvidia’s chips. Nvidia is essentially subsidizing one of its largest customers, potentially artificially inflating real AI demand.
Yet, these companies have massive capital backing, real technical capabilities, and strategic partnerships with major cloud providers and enterprises. Some will be consolidated, some will be acquired, but the category will survive.
TimeConsolidation: In 2026 to 2028, 2 to 3 major players will emerge while smaller model providers will be acquired or closed down.
Layer 1: Infrastructure (permanently built)
Here’s the paradoxical view: The infrastructure layer — which includes Nvidia, data centers, cloud providers, memory systems, and AI-optimized storage — is the least bubbly part of the AI boom.
Yes, the latest projections show that global AI capital spending and venture capital investment has already exceeded $600 billion in 2025, with Gartner estimating that all AI-related spending worldwide could exceed $1.5 trillion. It feels like a bubble zone.
But infrastructure has one important characteristic: no matter how successful a specific application is, its value remains intact. The fiber optic cables laid during the dot-com bubble weren’t wasted — they enabled YouTube, Netflix, and cloud computing. Twenty-five years ago, the original dot-com bubble burst after debt financing led to the construction of fiber-optic cables for a future that had not yet arrived, but that future finally arrived, and the infrastructure was there waiting.
Despite the stock pressure, Nvidia’s Q3 fiscal year 2025 revenue reached nearly $57 billion, up 22% quarter-on-quarter and 62% year-on-year, with the data center division alone generating nearly $51.2 billion. These are not vanity metrics; They represent real demand from companies making real infrastructure investments.
The chips, data centers, memory systems and storage infrastructure being built today will help any AI application ultimately succeed, whether it’s today’s chatbots, tomorrow’s autonomous agents or applications we haven’t even imagined yet. Unlike commoditized storage alone, modern AI infrastructure includes the entire memory hierarchy – from GPU HBM to DRAM to high-performance storage systems that serve as token warehouses for inference workloads. This integrated approach to memory and storage represents a fundamental architectural innovation, not a commodity game.
Time: Short-term overbuilding and lazy engineering is possible (2026), but long-term value retention is expected as AI workloads expand over the next decade.
Cascade effect: why it matters
The current AI boom will not end with a dramatic crash. Instead, we will see a cascade of failures starting with the weakest companies, and the warning signs are already here.
phase 1: Wrapper and white-label companies face margin compression and facility absorption. Hundreds of low-margin AI startups will shut down or sell for pennies on the dollar. Valuation of over 1,300 AI startups now exceeds $100 million with 498 AI "unicorns" Values of $1 billion or more, many of which would not justify those valuations.
phase 2: Foundation model consolidation occurs when performance is equal and only the players with the best capital survive. Expect 3 to 5 major acquisitions as the tech giant snaps up promising model companies.
step 3: Infrastructure spending has moderated but remains high. Some data centers will remain partially empty for a few years (like fiber optic cables in 2002), but eventually they will be filled as AI workloads increase.
What does this mean for builders
The most significant risk is the absence of a wrapper – its remaining a wrapper. If you own the experience the user engages in, you own the user.
If you’re building in the application layer, you need to jump right to the top:
From Wrapper → Application Layer: Just stop generating output. Own the workflow before and after AI interactions.
From Application → Vertical SaaS: Create execution layers that compel users to stay inside your product. Create proprietary data, deep integration, and workflow ownership that makes switching painless.
Distribution Gap: Your real advantage is not the LLM, but how you acquire users, retain them and expand what they do inside your platform. The winning AI businesses aren’t just software companies – they’re delivery companies.
bottom line
It’s time to stop asking if we’re in "" Hey bubble. We are in multiple bubbles with different characteristics and timelines.
Wrapper companies will likely be the first to emerge within 18 months. Foundation models will strengthen over the next 2 to 4 years. I predict that existing infrastructure investments will ultimately prove justified in the long term, although not without some short-term overbuilding problems.
This is not a reason for pessimism, it is a roadmap. Understanding what layers you are working in and what bubbles you may get trapped in is the difference between becoming the next victim and building something that will survive the shock.
The AI revolution is real. But not every company riding the wave will be able to reach the shore.
Val Bercovici is CAIO at WEKA.
<a href