It’s been a big month in AI and Silicon Valley, with the headlines clustering around several themes: a backlash to tokenmaxxing, growing regulatory pressure, and major IPOs, including SpaceX and Cerebras. The moment reflects a shift from idealizing companies like Anthropic to a healthier debate about where and when value will accrue.
On tokenmaxxing, attitudes have shifted sharply over the past month. The race to see which employees can spend the most tokens has reached a crossroads, as CFOs see high costs but little measurable impact to justify the expense. Uber is the canonical example, reportedly burning through its 2026 budget in just four months.
Sentiment has also shifted rapidly around regulation. It was only in February of this year that Anthropic refused to grant the U.S. Department of Defense unrestricted access to its model. As of our publication date, the U.S. has barred foreign nationals from using Anthropic’s frontier models. Access is becoming a geopolitical question; in Europe and India, there is renewed support for sovereign models like Mistral and Sarvam.
On the IPO front, SpaceX’s IPO is a signal that a new, exceptionally well-funded player has joined the field.
Instead of a single leader in AGI, as seemed possible a year ago, we now have four powerful, well-resourced players, each with different assets to bring to the table: OpenAI, with the largest AI consumer footprint through ChatGPT; Anthropic, with best-in-class models and harnesses as of today; Google, with the most money, IP, and talent; and xAI, still a distant fourth but flush with capital and compute access, and partnered with Cursor at the application layer.
These leaders are trailed by several others that are not far behind: Microsoft, Apple, and Meta. Then there are the Chinese players, which have the advantage of being open-weight and dramatically cheaper. DeepSeek’s latest model scores close to Opus on SWE-bench at a fraction of the price, and open-source serving can be cheaper still. When a credible alternative costs a fraction as much and functions well enough, why not make the switch?
This is all adding up to an increasingly competitive model ecosystem, with many well-positioned players.
The model is not the moat
The prevailing assumption has been that there will be a winner-take-all scenario, with the maker of the best model emerging as the clear winner. But that’s not what’s happening.
Competition continues to be fierce among the labs, making it difficult for any one of them to emerge as the clear leader.
Two main forces are driving this.
First, competitive dynamics are driving down the price of intelligence. Models improve quickly—a model at the frontier today will look ordinary within a year—and some capabilities can be reproduced through distillation and other techniques. Any lead that a lab gains appears hard to maintain.
Regulatory forces will also continue to play a bigger part. As AI becomes a national-security asset and risk, governments are becoming more involved in which models are released, where infrastructure is built, and who gets access to it. That will slow and complicate the way labs reach customers.
This means the model is not the moat. It never was. Intelligence is not a winner-take-all market. Its frontier is also jagged. There is no single frontier but many, and a different lab will lead on each.
Instead, the real value for most companies will come from the application layer, through their proprietary data and workflows—the specific knowledge that sits within companies and industries, along with the decision traces and context graphs that aren’t easily replicated.
This is where the product comes in: the harness plus the model, wired into a specific customer’s environment to drive a business outcome. A product can turn a model into something durable: a workflow, a habit, a distribution channel, a customer relationship, or a store of usage data. The model remains central, but it becomes the replaceable engine inside something that is much harder to replace.
The labs are steadily moving into products, not just selling models and harnesses. In the B2B realm, Anthropic has been the most aggressive, with Claude Code, Cowork, and most recently Claude Tag. OpenAI is turning ChatGPT into a broader platform spanning coding tools and agents, while moving into devices and robotics. Google is leveraging its broad application footprint across Android and Google Workspace to embed Gemini. xAI’s partnership with Cursor gives it a path into coding products
Looking ahead, Anthropic and OpenAI have begun hiring subject-matter experts across law, finance, and healthcare; we should expect to see more domain-specific products from them soon.
The labs are also expanding their deployment options. Microsoft and Google already have an advantage through their large cloud-services teams. OpenAI and Anthropic have partnered with private equity firms to deploy models within their portfolio companies. OpenAI has launched its Frontier Alliance program to work with global consulting firms and use their reach to accelerate deployment.
Infrastructure platforms are already seeing the writing on the wall and updating their offerings to be model-agnostic. Databricks’ new Agent Bricks platform supports model choice across ecosystems, from OpenAI to Qwen. Snowflake’s agent harness is likewise designed to make the underlying LLM interchangeable without rebuilding the guardrails around it.
We’re moving out of the winner-take-all framing and into a moment when access to a growing number of high-quality models is expanding quickly, giving founders more tools to build with than ever before, at lower prices than ever.
Get insights directly to your inbox.
How startups win in this era of AI
More competition in the model ecosystem is good for startups building at every level of the stack. There are certainly opportunities for new model companies, and there will also be an increasing number of opportunities for harness providers, but the largest opportunity continues to be building AI-native apps. At the same time, model providers will continue to expand their product footprints and compete with their customers.
For founders today, the question is which parts of your product will prove defensible and enduring.
As we have written before, product companies have the opportunity to build a context graph for their customers and domains, which, if done well, can become an enduring advantage. Neither the model companies nor incumbent systems of record typically capture the decision traces required to build the context graph. The orchestration layer sees the full picture, and that’s where the real value is created. This is one more reason why value accrues in the product layer, not the model.
In addition to having a point of view on what makes your product sticky, founders need to think carefully about customer and market segmentation. A recent conversation with a friend of mine, the chief AI officer of a large telecom company, illuminated this for me.
There are two questions founders need to ask themselves.
First, what do your customers care about most when it comes to AI?
First movers and early adopters are willing to buy a product that’s still rough around the edges and invest in co-development. Companies that value control want local models that can be run within their VPCs and meet stringent security requirements. Cost-conscious buyers tend to move more cautiously and may run long trials with multiple vendors before making a commitment.
Second, what’s the nature of the problem your product solves?
There are several categories your product might fall into. One is horizontal productivity tools, like Claude, deployed company-wide. There is still room for startups here—including Glean, which brings enterprise knowledge into AI workflows, and our portfolio company Viven, which creates AI digital twins of employees—but you are truly competing with the giants in this category.
Next, there are products that help make departmental decisions, mostly focused on corporate functions like finance, HR, and GTM operations. These customers are often looking for reliable off-the-shelf solutions that offer clear ROI. They want customization but rarely need bespoke builds. Eightfold, in HR and recruiting, and Maximor, in finance operations, are good examples.
Finally, there are solutions that address strategic, often customer-facing, board-level priorities. These are heavily customized and bespoke solutions, often for teams that want to be the first to market. These customers are more likely to work with companies like Turing or the AI services ventures that the labs have launched.
Building exceptional products is still hard, even without the rapid shifts of the AI era. Many AI apps that look indefensible may not be suffering from some new law of the model era. They may simply not be very good products.
For founders building at the earliest stages today, the advice remains the same as ever: choose the right market and customer segment, make an informed bet on the technology curve, build something people actually want, and then make it easy for them to buy it from you.
We believe that most of the value in this phase of the AI era will be built at the app layer. The model was never the product. It was never going to be your moat either.