MC.76: Benedict Evans Presentation

Hey everyone,

Benedict Evans' latest presentation "AI Eats the World" landed at the perfect moment, just as we're witnessing the AI industry's pricing chaos unfold in real-time.

The Processor Wars All Over Again

The slide that hit hardest was his comparison between today's frontier model race and the Intel vs. AMD processor wars of the early 2000s. Remember when choosing between a 2.4GHz or 2.8GHz processor felt like a life-or-death decision? When megahertz ruled everything, and we obsessed over benchmark scores?

Today, we're living through the exact same dynamic with AI models—except instead of gigahertz, it's billions of parameters. Instead of clock speeds, it's benchmark leaderboards. The pattern is identical: everyone's racing toward specs that will eventually become commoditized.

Just as most people today couldn't tell you what processor powers their laptop (and don't care), we're heading toward a world where the underlying AI model becomes invisible infrastructure. The question isn't which model has the highest MMLU score—it's which application solves your actual problems.

The Real Question: What Is AI Actually For?

Evans cuts through the hype with the fundamental question that matters: What is AI for?

Not what can it do in demos. Not what benchmarks it can ace. What problems does it solve that people will pay for, repeatedly, at scale?

Right now, the clearest answer remains code generation and productivity enhancement. We see this pattern clearly at Alter—developers and knowledge workers finding genuine value in AI-assisted workflows. But the bigger question looms: can we move beyond content production into truly transformative use cases?

The Adoption Reality Check

Here's the sobering parallel Evans draws: despite decades of cloud evangelism, most enterprise infrastructure still isn't cloud-native. The transition is massive, but it's also incomplete and slower than the hype suggested.

AI adoption will follow the same pattern. We're still in the early bundling/unbundling phase—taking specific features from established applications and rebuilding them AI-first. Sometimes it works spectacularly (think Canva starting with simple social media graphics and now challenging Adobe's entire creative suite). Sometimes it doesn't.

The winners won't be determined by model performance alone, but by who best understands organizational decision-making, enterprise adoption cycles, and the messy reality of changing established workflows.

The Long Game

Evans reminds us that technology transitions take decades, not quarters. The cloud revolution is still happening. The mobile revolution is still unfolding in many industries. AI will be no different.

The companies that win won't be those with the highest benchmark scores or the most aggressive pricing. They'll be those that best navigate the messy, slow, human process of changing how work actually gets done.

As the frontier model race intensifies and pricing wars escalate, the real opportunity lies in taking a step back from the bubble and focusing on the hard numbers that matter: enterprise adoption, user retention, and genuine productivity gains.

The processor wars eventually ended not with a winner, but with commoditization. The AI model wars will likely follow the same path. The question is: what are you building on top of that infrastructure?

Cheers,
Olivier

Modern Chaos explores the intersection of technology, business, and society in an age of rapid transformation. Subscribe for weekly insights on how AI and other emerging technologies are reshaping our world.

Like this newsletter? Forward it to a friend and have them sign up here.

Until next Thursday 🎉