@reddingecon.bsky.social

⭐ TL;DR:
Some superstar firms don’t just sell technology — they engineer the skills that make it indispensable. 
Market power can hinder allocation but also fuel learning.
Read the paper 👉 eumena.github.io/files/em_jmp...

Why do we care?
⚖️Regulators often see big tech’s “free” tools as pro-competitive.


🤯 But these giveaways can reshape labor markets and the direction of technological specialization.

🧠Competition policy needs to account for the skill-formation channel of market power. 7/7

🧩How does it all fit together? The logic:
1️⃣ Google makes it easier (cheaper) to learn TensorFlow.
2️⃣ More workers acquire the skill.
3️⃣ More firms adopt compatible cloud tools.
4️⃣ Google’s ecosystem (and rents ) grow.
Training is a strategic lever of market power. 6/n

Empirically, I study Google’s open-sourcing of TensorFlow in 2015, a turning point in AI skill diffusion.

After the release:

📈 Firms adopting Google Cloud Platform surged.

👩‍💻 Demand for ML skills rose.

💰 Wage premia for those skills declined as the trained workforce expanded. 5/n

This creates a policy-relevant trade-off:
- Market power distorts resource allocation.
- But it can also stimulate investment in high-quality training that diffuses advanced technology.
Some monopolists, paradoxically, teach the world to use their tools. 4/n

I build a general-equilibrium model where technology providers decide both:
1. how much to subsidize training in their ecosystem’s skills, and
2. how much to charge firms for using their capital.
Their market power lets them internalize how training affects adoption and wages. 3/n

When production depends on technology-specific skills, dominant providers of that technology (e.g. cloud platforms or equipment makers) have a stake in making those skills abundant.
Even if the trained workers never work for them directly, more skilled users = higher demand for their technology.2/n