Zero Friction Agent Economy
Shraced@shraced
−3.3%
since Oct 15
The idea
I believe AI agents will take over knowledge work within 6-7 years, running company back offices end to end while handling customer interactions directly. When thinking and admin become nearly free, value shifts to what agents can't skip: chips and power, the platforms they run on, identity and security checks, payment rails, and proprietary data. This basket owns those six layers with large steady leaders as ballast, avoids anything living off friction or per-seat pricing, and keeps a modest cash reserve while holding for years.
What it owns
- Every agent action runs on chips, so more agents doing more work means more chips sold. · NVIDIA CorporationIt sells most of the chips and software that agents are trained and run on, so demand for AI computing directly drives its growth., Taiwan Semiconductor ManufacturingAlmost every advanced AI chip, whoever designs it, gets made in its factories, so it gets paid regardless of which chip designer wins., Broadcom Inc.Big platforms building their own AI chips and wiring data centers together rely on it for both, so it gains as agent workloads scale.20%
- When agents talk to agents, every one must prove who it is and be kept from being hijacked. · Cloudflare Inc.A big slice of internet traffic passes through it, which positions it to decide which agents get in and block bad bots for a fee., CrowdStrike HoldingsWhen agents can act on a company's systems, each one is a new way in for attackers, and it protects the machines and workloads those agents touch., Zscaler Inc.It sits between users, apps and machines and checks every connection, which matters more when most of the traffic is agents talking to other agents., Okta Inc.Agents talking to business agents need to prove who they are and what they are allowed to do, and it sells exactly that kind of identity checking.17%
- Agents can copy any process but not proprietary data, and they pay for it per use rather than per person. · S&P Global Inc.Agents doing financial work need credit ratings, benchmarks and market data they cannot make up, and it owns a lot of that proprietary information., Verisk AnalyticsInsurance agents pricing risk need years of claims and property data, and it owns the industry's shared pool of that information., Snowflake Inc.It charges for computing resources consumed rather than per human login, so revenue scales with agent-driven query growth as long as usage patterns shift toward AI workloads., Datadog Inc.Back offices run by agents still need someone watching whether everything works, and it bills based on how much it monitors, so more agent software means more to watch.16%
- The companies that build the models and rent out the cloud are where agents actually live and work. · Alphabet Inc.It owns frontier models, AI chips and a giant cloud, so both customer agents and business agents can live and work there., Amazon.com Inc.Its cloud hosts a huge share of business software where business agents will run, and its warehouses handle the physical work agents cannot do.14%
- Agents buying from agents still need a trusted network to move money, and these networks are built for it. · Visa Inc.It operates a trusted network that handles agent purchases, fraud detection and dispute settlement when agents buy things for people., Mastercard Inc.It operates a card network that agents will run payments through and is building tools for agents to pay with checked credentials.14%
- Data centers running billions of agents need huge amounts of steady electricity and grid equipment. · Vistra Corp.It owns power plants that can sell to data centers, and agents can skip paperwork but they cannot skip the electricity those plants provide., GE VernovaMore data centers mean more gas turbines and grid equipment, and it builds both, so it gains from the power buildout regardless of which utility wins., Constellation EnergyIt runs the largest US nuclear fleet, which gives data centers steady, around the clock, low carbon power that is hard to replicate.12%
- Cash7%