The AI wars are ending. Not because someone built AGI. But because frontier models are becoming commodities.
For the past two years, the industry has obsessed over one question: Who has the best model? OpenAI. Anthropic. Google. Meta. xAI.
Every benchmark. Every leaderboard. Every product launch has reinforced the same belief: The smartest model wins. History says otherwise. Every technology revolution follows the same pattern. Technology becomes abundant → Value moves somewhere else
Today, our friends in China are accelerating that shift.
Good Enough Changes Everything
For years, frontier AI companies competed on intelligence. Today, they're increasingly competing on economics. Chinese AI labs, including DeepSeek, Moonshot AI (Kimi), MiniMax, Z.ai, and Qwen, have released increasingly capable frontier models that perform well across many enterprise tasks while often costing substantially less than leading U.S. alternatives. That changing cost-performance equation is prompting businesses to rethink where they run AI workloads.
Recently, one of our interns spent several days evaluating the top Chinese models. The conclusion wasn't that they outperform OpenAI. It was something much more interesting. For a surprising number of enterprise workflows… They were good enough.
And "good enough" has a habit of reshaping industries. When products become good enough, buyers stop paying enormous premiums simply because something is slightly better.
They optimize for value.
History Keeps Repeating Itself
This isn't the first time we've seen this happen.
Amazon
Servers became a commodity → The value shifted to software
Information became a commodity → The value shifted to discovery
Spotify
Music distribution became a commodity → The value shifted to recommendation
Apple
Hardware became a commodity → The value shifted to the ecosystem
Salesforce
CRM infrastructure became a commodity → The value shifted to applications
Notice the pattern? Technology doesn't eliminate value. It relocates it.
Models Are Becoming Infrastructure
This is exactly what we're beginning to see in AI. Today's debate revolves around frontier models. Tomorrow's debate won't. Because once multiple companies can deliver capable intelligence at dramatically different price points, intelligence itself becomes harder to differentiate.
That doesn't mean frontier models stop improving. It means they start behaving like infrastructure. Infrastructure matters. But infrastructure rarely becomes an enduring competitive advantage.
Alex Karp Thinks Everyone Is Looking in the Wrong Place
Palantir CEO Alex Karp has been making an argument that deserves more attention. He believes organizations are becoming too focused on the models themselves.
The long-term advantage, he argues, comes from operationalizing AI inside organizations, not simply owning access to the newest frontier model. We think recent developments reinforce that point.
If organizations can increasingly choose between multiple capable models, then the strategic question changes.
It stops being: Which model should we buy?
And becomes: What capabilities are we building on top of those models?
The Profit Pool Is Moving
This is the mistake every technology revolution invites. People assume the money stays with the underlying technology. History says otherwise.
The real winners usually build one layer above it. We believe AI is following the same trajectory.
Today companies compete on:
Model intelligence
Benchmarks
Context windows
Token pricing
Tomorrow they'll compete on:
Institutional expertise
Governance
Proprietary workflows
Organizational knowledge
Execution
Those things don't disappear when another model launches.
The New AI Stack
The AI stack is already evolving.
Commodity
Models, Inference, Compute (eventually)
↓
Value (scarcity)
Institutional expertise, Governance, Execution
The scarcity isn't intelligence anymore. It's knowing how to apply it.
The Companies That Win Won't Own Better Models
Every company will soon have access to extraordinary AI. Open-source models will improve. Commercial models will improve. Chinese models will improve. American models will improve. Prices will continue to fall. That isn't the threat. It's the opportunity.
Because once intelligence becomes abundant, organizations are forced to compete somewhere else. On judgment. On expertise. On governance. On execution.
Those are much harder to copy than a model endpoint.
The Next AI Race Has Already Started
The AI race isn't ending. It's changing. China isn't forcing companies to rethink AI because its models are magically superior. It's forcing them to rethink what actually creates value. The organizations that continue chasing whichever model tops the leaderboard will always be playing catch-up.
The organizations that capture their expertise, govern it effectively, and operationalize it across the business will build advantages that survive every model release.
Because history has never rewarded the companies that simply adopted the technology. It has always rewarded the companies that understood where the value moved next.