MarketsJul 21, 2026
The Falling Cost of AI Could Supercharge Demand for Chips and Memory
Hardware Bottlenecks
Rhea Lobo
Discover
The AI trade is entering its bargain-bin era. Moonshot AI's surprise release of a powerful, low-cost AI model rattled markets and sparked a selloff in chip stocks. The initial reaction focused on the risks, but it may have overlooked the opportunities.
Cheaper intelligence, bigger demand: Moonshot AI released Kimi K3 last week, a powerful AI model that matched Anthropic's Fable 5 and OpenAI's GPT-5.6 Sol on several benchmarks while costing roughly half to one-third as much. The launch was so overwhelming that Moonshot AI had to pause new sign-ups over the weekend due to compute constraints. If Kimi is any indication, lower AI costs may accelerate demand for the infrastructure behind it as adoption spreads.
Kimi K3's size underscores the growing importance of high-bandwidth memory. Even though the model actively uses only a small fraction of its 2.8T parameters at a time, the entire model still needs to be stored in memory. That makes K3 difficult for most enterprises to run on their own servers and reinforces demand for cloud-scale HBM infrastructure. Wedbush analyst Matt Bryson noted larger AI models may require more AI chips working together, which would also boost networking providers alongside memory vendors.
The policy wildcard: OpenAI and Anthropic are urging Washington to tighten restrictions on open-weight Chinese AI models, citing national security risks. Critics argue the effort is also aimed at limiting competition ahead of future public listings. The Trump administration has yet to act. For investors, Kimi's viral adoption suggests the next AI race may be won as much by hardware as software.
