Why Most AI Startups Will Become Features, Not Products
OpenAI’s recent announcements—ChatGPT app integrations, commerce infrastructure, and an identity layer—reveal a strategic consolidation play. With $375 billion invested in AI this year and over 91,000 companies competing in the AI market, most are building features, not products, and OpenAI is creating the platform that will absorb them.
Listen to “Why Most AI Startups Will Become Features, Not Products” Deep Dive
Last week, OpenAI announced app integrations for ChatGPT, whereby Spotify, Figma, Zillow and other brand partners can now run seamlessly within chat conversations. Users can call upon apps by name or have ChatGPT suggest them contextually.
On the surface, this looks like OpenAI is building its version of the Apple App Store or Salesforce AppExchange. But what appears to be happening is something more strategic: OpenAI is preemptively commoditizing the competition. They’re building a two-sided consolidation strategy. Partnerships with established brands, like Walmart, Spotify and Zillow, create immediate user value and revenue streams that expand ChatGPT’s audience reach and redefine its purpose beyond just answering questions. The flip side? This app ecosystem is creating a landing pad for struggling AI startups. All those voice and image generators, presentation tools, and video editors now have a path to become features rather than competitors. It’s a moat-building strategy against Google, Anthropic and Microsoft that works both ways: attracting proven brands while absorbing would-be rivals.
AI investments are projected to reach $375 billion this year and $500 billion by 2026. That kind of capital creates bubble conditions. And like the dot-com era, it’s masking fundamental problems—e.g., category confusion and failure to demonstrate sustainable revenue. AI companies are creating feature-level solutions while positioning themselves as standalone platforms, often with unclear paths to profitability.
What Makes OpenAI’s Approach Different
OpenAI has transformed ChatGPT from a conversational AI into something closer to an application platform. Instead of treating AI as a feature added to existing apps, OpenAI is bringing entire apps into ChatGPT’s AI-centric context, flipping the script on the usual approach.
Here’s what that looks like in practice: Say I’m planning a trip to Majorca, Spain. ChatGPT can now respond to my prompt with: “I can help you organize your trip. To make things easier, I can connect directly to Expedia to check on hotels, flights and car rentals. Would you like to proceed?” The entire process has just been streamlined, eliminating the need to sift through multiple websites.
“Instead of treating AI as a feature added to existing apps, OpenAI is bringing entire apps into ChatGPT’s AI-centric context, flipping the script on the usual approach.”
This matters because while Google integrates Gemini into its browser and while platforms like Perplexity or Anthropic’s Claude focus on answering questions, OpenAI has expanded its offering to the point where apps aren’t just responding to prompts; they’re interactive, adaptive and personalized.
The Bigger Play: Commerce and Identity
OpenAI’s app integrations are just one piece of a larger infrastructure play. The company recently launched Instant Checkout through its Agentic Commerce Protocol, allowing users to go from chat to purchase in just a few taps. Partners include Walmart, Etsy and more than a million Shopify merchants.
More significantly, OpenAI is developing a ChatGPT login feature that would function like Google or Facebook single sign-on (SSO) logins, letting users access websites with their ChatGPT accounts. This positions ChatGPT as a potential identity layer for the internet, controlling login credentials, data flow and the distribution of new capabilities.
Together, these moves reveal OpenAI’s full ambition: not just to be an app platform, but a foundational infrastructure for commerce and identity across the internet. While competitors focus on answering questions, OpenAI is building its own ecosysterm.
Learning from the Dot-Com Bubble
With over $375 billion being invested in AI in 2025 alone, it’s fair to ask: is this an AI bubble? Aside from the major players (OpenAI, Google, Anthropic, Meta, Microsoft and a few others), what happens to the secondary tier?
As you can see from the chart below CB Insights identified 100 of the most promising AI startups, but as of this year there are over 91,000 companies working in AI. Not all of these businesses are sustainable, which could lead to either a bubble or a market correction that results in industry consolidation, weeding out weaker companies that will either get acquired or liquidated.
The dot-com era offers a useful parallel. The companies that survived weren’t necessarily the most technically advanced; rather, they had proven business models that generated web traffic and revenue. Consider the case of AltaVista, which was one of the first search engines to index significant web content and it quickly became hugely popular—until Google came along. AltaVista’s popularity declined as Google’s superior search algorithms and pay-per-click advertising model took over. Technical capability simply wasn’t enough without also having a sustainable business model.
This market consolidation lesson matters because many AI companies appear to be mis-positioning themselves and failing to understand whether they’re making a real difference in their space or simply riding the AI wave.
In Zero to One, Peter Thiel argued that the business world learned the wrong lessons from the dot-com crash, favoring incremental advances over grand visions, iteration over planning and product over distribution. OpenAI seems to be proving Thiel right. While smaller AI companies stay lean and iterate, OpenAI is pursuing the grand vision of becoming the universal interface for AI. They’re not improving on competitors; they’re absorbing them.
The Three Paths Forward for AI Companies
With over 91,000 AI companies in the market, consolidation isn’t a prediction; it’s already underway. The question isn’t whether smaller AI companies will survive independently, but which path they’ll take. Most are headed down one of three routes:
Path 1: Acquisition
Acquisition is the most common outcome. AI M&A activity has surged 155% in the first half of 2025, with over $100 billion in deals. OpenAI alone has acquired companies for billions, while Google executed a $2.7 billion acquihire of Character.AI. For established players, it’s faster and cheaper to acquire technology than to build it internally, and in many cases the talent may be more valuable than the product itself—AI programmers and data scientists remain in critically short supply.
Path 2: Platform Integration
If OpenAI’s app ecosystem scales beyond brand partnerships, it creates a new category: the AI app marketplace. Smaller AI companies become features within larger platforms, similar to how Salesforce’s AppExchange hosts 6,000+ apps or how WordPress offers 90,000+ plugins. These companies don’t disappear; rather, they just become part of a larger ecosystem, accessed through ChatGPT, Gemini or Claude rather than as standalone products.
Path 3: Shutdown
Without clear differentiation or sustainable revenue, some companies will exhaust their capital and close. This is market reality. The AI boom has funded ideas that work better as features than businesses. These companies will ultimately face a choice: pivot to one of the first two paths or wind down operations.
Very few AI companies will remain independent platforms. Those that do will need more than just impressive technology; they’ll need distribution, differentiation and a business model that can’t be easily replicated by OpenAI, Google or Microsoft.
The Questions that Matter
OpenAI’s app integration strategy signals that consolidation is already underway. As investors, business leaders and observers of this space, we need to start asking harder questions: Is this AI company solving a problem that justifies a standalone business, or is it building a feature that will be absorbed by a platform? Can it demonstrate sustainable revenue, or is it burning capital on impressive demos? Does it have true differentiation, or is it one algorithm update away from irrelevance?
The dot-com era taught us that technological capability alone isn’t enough. The AI era is teaching us the same lesson, just way faster. The companies that survive won’t be the ones with the best technology or the biggest funding rounds; rather, they’ll be the ones who understand what they actually are and can position themselves accordingly before someone else makes that decision for them.
Thanks for reading and/or listening. Questions? Email me here.
Photo by Ashkan Forouzani on Unsplash
