Kalshi’s Harrison Shows Where the Next AI Trade Is HeadingAI agents need far more compute than chatbots — and that changes the winners
By Luke Lango
Agentic AI is shifting AI from one-off chatbot responses to always-on workflows that plan, act, check and revise.
Enterprise AI agents can consume far more inference compute than single-shot prompts, driving demand for chips, memory, networking, power and storage.
The biggest beneficiaries may be the infrastructure suppliers powering agentic AI: accelerators, custom silicon, HBM, optical connectivity, servers, and cooling.
Gartner estimates that agentic workflows consume 5-30x more tokens per task than single-shot generative queries.
Goldman Sachs sees the monthly token count for agentic AI applications reaching roughly 120 quadrillion by 2030.
Accelerators: Nvidia and AMD Power the Reasoning Loop
Networking and Custom Silicon: Lowering the Cost per Token; AVGO
Memory: The Bottleneck Behind Long-Context Agents; MU
Servers, Racks, and Power: The Always-On Agent Layer; Dell
Storage: Fast Retrieval for Enterprise AI Agents; Seagate
Optical Connectivity: The Overlooked Agentic AI Bottleneck; Corning
Source: Investor Place
https://investorplace.com/hypergrowthin ... s-heading/
It's all about "how much you made when you were right" & "how little you lost when you were wrong"