When Uber launched, it didn’t just provide rides, it subsidized them. By pouring venture capital into cheap fares, it hooked both drivers and riders, building a massive user base before competitors could catch up. The economics were never sustainable at face value, but the strategy worked: the network scaled, user habits hardened, and switching costs rose.

Today, AI is running a strikingly similar play. The true cost of training, hosting, and running large language models is enormous. Yet to end users, access feels close to free: ChatGPT at $20 a month, free copilots bundled into Microsoft products, and generative AI features embedded into search engines. Big tech companies are subsidizing usage heavily, prioritizing scale and adoption over profitability.

Why Subsidize AI?

1.Data Flywheel: More users mean more feedback, which refines models faster.

2.Lock-In: Early habit formation ensures users stay within a company’s ecosystem.

3.Market Power: By undercutting rivals, leaders can entrench their position before AI commoditizes.

4.Platform Play: Subsidies attract developers and startups to build on top of the ecosystem, creating dependence.

The Implications

1.Winner-Takes-Most Dynamics: Just like Uber dominated ride-hailing, AI could consolidate into a few mega-platforms. Smaller players without deep subsidies may never achieve critical mass.

2.Artificial Prices: Current costs don’t reflect reality. When subsidies ease, users may face sticker shock as true compute costs are revealed.

3.Policy and Regulation: Governments may intervene if subsidies distort competition or reinforce monopolies.

4.Innovation Risk: If usage becomes tied to a handful of subsidized platforms, diversity in AI architectures and business models may shrink.

The Long Game

Uber’s subsidies eventually burned off, and prices rose to sustainable levels. AI will follow the same arc. The subsidy era isn’t about profits, it’s about capturing territory in a once-in-a-century platform shift. The question isn’t whether subsidies will end, but who will own the market when they do.

When Uber launched, it didn’t just provide rides, it subsidized them. By pouring venture capital into cheap fares, it hooked both drivers and riders, building a massive user base before competitors could catch up. The economics were never sustainable at face value, but the strategy worked: the network scaled, user habits hardened, and switching costs rose.

Today, AI is running a strikingly similar play. The true cost of training, hosting, and running large language models is enormous. Yet to end users, access feels close to free: ChatGPT at $20 a month, free copilots bundled into Microsoft products, and generative AI features embedded into search engines. Big tech companies are subsidizing usage heavily, prioritizing scale and adoption over profitability.

Why Subsidize AI?

1.Data Flywheel: More users mean more feedback, which refines models faster.

2.Lock-In: Early habit formation ensures users stay within a company’s ecosystem.

3.Market Power: By undercutting rivals, leaders can entrench their position before AI commoditizes.

4.Platform Play: Subsidies attract developers and startups to build on top of the ecosystem, creating dependence.

The Implications

1.Winner-Takes-Most Dynamics: Just like Uber dominated ride-hailing, AI could consolidate into a few mega-platforms. Smaller players without deep subsidies may never achieve critical mass.

2.Artificial Prices: Current costs don’t reflect reality. When subsidies ease, users may face sticker shock as true compute costs are revealed.

3.Policy and Regulation: Governments may intervene if subsidies distort competition or reinforce monopolies.

4.Innovation Risk: If usage becomes tied to a handful of subsidized platforms, diversity in AI architectures and business models may shrink.

The Long Game

Uber’s subsidies eventually burned off, and prices rose to sustainable levels. AI will follow the same arc. The subsidy era isn’t about profits, it’s about capturing territory in a once-in-a-century platform shift. The question isn’t whether subsidies will end, but who will own the market when they do.