Last Updated: May 2026 — Enhanced with AI business impact analysis
While the tech world obsesses over AI capabilities, two companies are making fundamentally opposite bets on how AI businesses will actually make money. Cognition’s human-centric approach and Groq’s infrastructure — as explored in the economics of AI compute infrastructure — play reveal a critical fork in the road for AI business models.
The Human-AI Partnership Model vs. The Speed Infrastructure Play
Cognition, the startup behind AI coding agent Devin, is doubling down on what founder Scott Wu calls “AI agents shouldn’t replace humans.” This isn’t just philosophical posturing—it’s a deliberate business model choice. Cognition makes money by selling AI tools that amplify human developers, not replace them. Their revenue model depends on sustained human involvement in the coding process.
Groq, meanwhile, is pursuing the exact opposite strategy with their reported $650M funding round. — as explored in the economics of AI-era business models — They’re betting that AI will eventually automate so much work that the real money lies in owning the infrastructure that makes AI blazingly fast. Groq’s Language Processing Unit (LPU) chips promise 10x faster inference than traditional GPUs—a pure infrastructure play.
Why Business Models Matter More Than Technology
These aren’t just different products—they represent fundamentally different theories about where AI value accrues. Cognition’s model requires ongoing human subscription revenue, creating recurring relationships with development teams. They make money when humans stay in the loop, so they design AI that enhances rather than eliminates human work.
Groq’s infrastructure model follows the classic “picks and shovels” playbook. As Nvidia’s recent $20B near-acquisition spree shows, whoever controls the computational infrastructure captures massive value regardless of which specific AI applications succeed. Groq doesn’t need to guess whether AI will replace humans—they just need AI usage to explode.
The Dependency Trap vs. The Commodity Risk
Each model faces distinct existential risks. Cognition’s human-centric approach could become obsolete overnight if AI truly achieves human-level coding ability. Their entire revenue model collapses if customers no longer need human developers in the loop. It’s the classic innovator’s dilemma—optimizing for today’s market while potentially missing tomorrow’s disruption.
Groq faces the opposite problem: commoditization. Infrastructure businesses live or die by maintaining technological advantages. If competitors match their speed improvements or if software optimizations make raw computational power less critical, their premium pricing disappears. They’re racing to establish market position before their technical moat erodes.
The Real Competition: Business Model Validation
Neither company is really competing on technology alone—they’re competing to validate their vision of AI’s economic future. Cognition needs to prove that human-AI collaboration creates more value than pure automation. They must show that their “enhancement” model generates better outcomes than “replacement” alternatives.
Groq needs to prove that computational speed becomes the primary bottleneck in AI adoption. They’re betting that as AI gets smarter, the constraint shifts from capability to performance. Their success depends on AI applications becoming so compute-intensive that speed premiums justify their infrastructure costs.
The Winner Takes All—Or Nothing
Here’s the bold prediction: both models can’t simultaneously dominate. If Cognition’s human-centric approach proves sustainable, it suggests AI won’t fully automate knowledge work—limiting the computational demands that drive Groq’s premium infrastructure model. If Groq’s speed-focused infrastructure becomes essential, it implies AI is moving toward full automation—exactly what would kill Cognition’s human-partnership revenue model.
The market will ultimately decide which vision of AI’s economic future proves correct. But right now, these two companies are making opposite bets with hundreds of millions in funding—and only one theory about AI business models will survive.
Want more business model insights delivered weekly? Subscribe to the FourWeekMBA newsletter for deep-dive analysis on how the world’s most innovative companies actually make money.
FourWeekMBA AI Business Intelligence — strategic analysis of the moves that matter.