Agents Meet Quantum
Alex Adelman@alexadelman
· 1
−9.2%
since Sep 30
The idea
I believe AI agents are about to solve quantum computing's real problem, usability, by orchestrating and routing workloads, which should accelerate demand across hardware, cloud access, and infrastructure over the next two to three years. The basket spans pure quantum hardware makers, the big clouds and IBM collecting access tolls, the supply chain feeding every machine built, and the agent platforms driving the whole mechanism. I'm keeping 10% in cash to buy into sharp dips, and I'd rethink this if agent progress on quantum workloads stalls out.
What it owns
- The companies building quantum computers win or lose based on whether agents can route real workloads to them fast enough. · IonQIts trapped-ion machines already sell time through all three big clouds, so if agent-driven workloads pick up, IonQ's bookings show it first., Rigetti ComputingIt builds superconducting quantum chips and sells access through the clouds, so it moves fast on any real demand signal from agent-routed workloads., D-Wave QuantumIt already has paying customers using annealing for optimization problems, exactly the kind of workload an agent can hand off without a physicist in the loop., Honeywell InternationalIt owns Quantinuum, one of the strongest quantum hardware teams, wrapped inside a stable industrial business, so it's the safer way to hold pure quantum hardware., Quantum Computing Inc.It is the speculative photonic bet in the pure-hardware bucket; room-temperature photonic machines would be the cheapest for agents to call on if they work.25%
- The cloud platforms get paid for quantum access regardless of whose hardware wins, and agents will call quantum through these same clouds. · IBMIt has the largest fleet of quantum machines reachable over the cloud, so it collects a toll on quantum access regardless of which hardware wins., MicrosoftAzure Quantum sells access to multiple machines while Copilot puts agents in front of the enterprises that would send those workloads., AlphabetIt builds its own quantum chips and runs Gemini agents that could orchestrate them, so it's a toll collector that also owns a machine., AmazonAmazon Braket is the marketplace where most quantum startups sell machine time, so AWS gets paid on quantum access no matter whose hardware an agent picks.23%
- Software and workflow tools shipping agent products into enterprises at scale prove the mechanism this take rests on. · NVIDIAIts CUDA-Q software and GPUs are what hybrid quantum-AI systems actually run on, so agents orchestrating quantum workloads means more NVIDIA compute sold., PalantirIt already has agents doing real work inside companies and governments, which is the mechanism this whole take rests on., ServiceNowIt is pushing agents into the workflows of thousands of large companies, proof that the agent layer is shipping fast., SalesforceIts Agentforce is the biggest test of whether enterprises will pay for agents at scale, and if it lands, it validates the speed of the agent rollout the take depends on.21%
- Cryogenics, lasers, photonics, and test equipment go into every quantum machine built, so suppliers get paid on the buildout no matter the design. · FormFactorIt sells the cryogenic probes and test systems needed to build and check superconducting quantum chips, so it gets paid on every machine built., CoherentIt supplies the lasers and photonics that trapped-ion and photonic quantum machines run on, so more machines built means more Coherent parts sold., MKS InstrumentsIt makes the vacuum systems and photonics gear quantum labs and fabs cannot run without, a pick-and-shovel way to get paid on the buildout., Keysight TechnologiesIt sells the control electronics and measurement gear used to run and calibrate qubits, and it gets bought whether IonQ, Rigetti, or Google wins., LumentumIts lasers and photonic components feed both the quantum machines and the datacenter networks that agents run on, so it gets paid twice if the take plays out.21%
- Cash10%