OpenAI & Google's 2026 Declaration: The AI Price War Has Begun, and Now Is the Best Time to Jump In

Sam Altman said it at yesterday’s town hall: “By the end of 2027, you’ll be able to use today’s top-performing models at one-hundredth of the current price.” OpenAI and Google have both launched budget-friendly AI plans almost simultaneously, reversing course from the premium-tier push both companies were running just months ago.

The shift signals that subscriber growth at the top has started to plateau. That is good news for individual users, though it is worth noting that the race to the bottom on pricing cuts both ways: some of these budget products are not yet as capable as their premium counterparts, and the value gap is real.

• Sam Altman’s Town Hall Livestream
• Town Hall Summary (if you don’t have time for the full video)

The Price Disruption That Started in China

Backed by substantial government subsidies and low electricity costs, Chinese AI companies have been pushing price floors down faster than most Western observers expected. Z.ai’s GLM-4.5 raised its performance bar while offering a $3/month lite plan that delivers a Claude Code-level experience. Minimax briefly offered a $2/month plan right before its IPO, and its flagship M2.1 model performed surprisingly well against more expensive alternatives. Moonshot AI’s Kimi 2.5, unveiled yesterday, enhanced parallel agent capabilities for running multiple tasks simultaneously at an even lower price point.

I subscribed to Kimi 2.5 immediately after reading the release. The pattern I keep seeing: American teams set the AI standards and frameworks, and Chinese companies rapidly train follow-on models that perform exceptionally well within those frameworks. Whether that dynamic holds as frontier capability continues to advance is genuinely unclear, but for individual users right now, the value is hard to ignore.

• Kimi 2.5 API Pricing

Anthropic’s Partial Opening

Claude still maintains its Max subscriber-focused strategy, but as of yesterday, free users can now generate Excel spreadsheets, documents, presentations, and PDFs. This looks like a dual-pronged approach: addressing the user-count gap while keeping the core paying audience of business professionals willing to spend $20 or more as the priority. A budget-tier plan from Anthropic has yet to materialize.

Why the Entry Point Matters Now

Andrej Karpathy shared something striking this week. In November, his workflow was 80% manual coding and 20% AI agents. By December, it had completely flipped: 80% agents, 20% manual edits. His entire working method shifted in a single month. Projects like Clawdbot (now Moltbot), which went viral as non-developers set them up on their own, show that this shift is already happening for people well outside the engineering world.

The pivot from premium expansion to budget options means these companies want more users, fast. Start with a $2-3/month Chinese model as a secondary tool to build your intuition, then scale up to a $20 plan when you need it. I have been running Claude as my primary and Kimi as my secondary, and the combination has been more effective than either alone.

If you are still using AI exclusively on free tiers, the gap between you and people who pay will continue to widen. The pricing window is open now. It will not stay this wide.

• Andrej Karpathy’s Jan 27 Tweet