Kai ChenIBM wants Starling to deliver 20,000 times the compute of today’s quantum computers by 2029. Around that goal, IBM plans to acquire HRL’s electron-spin work, has run real-time control on AMD hardware and plans a Cisco network proof for 2030. One performance target now touches a pending acquisition, an off-the-shelf controller and a long-distance network.
Key takeaways
- Quantum progress is shifting from physical-qubit counts toward dependable logical operations produced by error correction, low-latency classical control, software and application workflows working together.
- IBM’s proposed HRL acquisition would add electron-spin circuits alongside its superconducting program, giving IBM a second hardware architecture to evaluate against common reliability milestones.
- Starling’s target of 20,000 times today’s quantum compute by 2029 is a system-level integration test, not a processor-only benchmark; IBM has not yet deployed the fault-tolerant product it describes.
- IBM’s AMD control work and planned Cisco networking proof put classical controllers and interconnects on the roadmap, while acquisitions by IonQ and D-Wave show competitors also bringing manufacturing and error-correction capabilities in-house.
- Prospective users should benchmark complete hybrid workflows—including runtime, end-to-end errors, control latency and support—rather than judging a quantum pilot by processor specifications alone.
Dependable operations now set the milestones
In 2015, IBM learned to detect two types of critical quantum errors. The machine could then identify what had gone wrong, an early step toward preserving a computation long enough to use it.
In 2017, IBM planned a 20-qubit cloud service and a 50-qubit prototype. In 2019, it made a 53-qubit universal machine available externally. Developers could run programs on larger machines and expose failures under use.
IBM also appeared more often in enterprise-oriented coverage and less often in research-oriented coverage across the compared 2024 and 2026 periods:
| Frame | 2024 compared period | 2026 compared period |
|---|---|---|
| Enterprise | 32.3% | 42.9% |
| Research | 22.6% | 5.7% |
IBM now defines Starling as a fault-tolerant performance target for 2029. Microsoft and Quantinuum, meanwhile, reported more than 14,000 experiments without errors. Both milestones focus on computation that survives error correction.
IBM must make error correction, decoding, control electronics and software work together to produce dependable logical operations. Starling’s 20,000× figure measures that output.
HRL would give IBM a second circuit path
IBM’s agreement to acquire HRL Laboratories matters because IBM already operates a substantial quantum-hardware program. IBM plans to integrate HRL’s electron-spin quantum circuits into its quantum computers while continuing its superconducting roadmap.
IBM’s main program uses superconducting hardware, while HRL builds electron-spin circuits. Microsoft’s Majorana 1 uses Majorana particles as a possible route to one million qubits. IBM, HRL and Microsoft are still pursuing distinct physical routes to practical machines.
If the deal closes, IBM can evaluate electron-spin circuits against the same reliability milestones as its superconducting hardware. The company gains an internal architecture comparison while the underlying physics remains unsettled.
IBM engineers still have to integrate HRL’s circuits with the control and software layers already on the roadmap. Starling remains a 2029 target, and IBM has not yet deployed the fault-tolerant product it describes.
IBM has put control and networking on the clock
IBM said it ran a key quantum algorithm in real time on commonly available AMD field-programmable gate arrays. IBM thereby put classical hardware on the algorithm’s critical path, where latency and reliability affect the result.
IBM and Cisco plan a proof of concept for linking quantum computers over long distances by 2030. IBM has assigned dated milestones to two interfaces around the processor: real-time classical control now and long-distance quantum interconnect by 2030.
Three buyers are moving capabilities in-house
IonQ acquired SkyWater for about $1.8 billion, giving the quantum company ownership of a chipmaker. D-Wave agreed to pay $550 million for Quantum Circuits, a developer of error-corrected gate-model technology. IBM’s planned HRL purchase would add electron-spin circuit research to its existing hardware program.
IonQ can now coordinate device development with SkyWater’s manufacturing operations. D-Wave’s pending deal would bring gate-model error-correction capability inside the company. IBM would gain another physical architecture to test.
Quantinuum raised $1.68 billion in an upsized IPO at a $15.6 billion valuation. Its filing reported a $136.6 million net loss on $5.2 million of revenue for the three months ended March 31.
Investors should test Quantinuum’s valuation against repeatable logical work, manufacturing progress and workflow gains. The company’s three-month revenue and loss figures leave that valuation dependent on years of technical delivery.
Chemistry teams need a workflow benchmark
IBM announced ten quantum projects focused mainly on quantum-physics simulation and chemistry and materials-science problems. Microsoft and Quantinuum paired their error-correction work with a chemistry problem that used high-performance computing, AI and quantum computing together.
A chemistry team evaluating a quantum pilot should define a molecular or materials task, run its current high-performance computing baseline, then measure total runtime, end-to-end error rates, control latency and support across the hybrid run. The team can then decide whether the quantum component improved the result enough to justify integration.
IBM’s 20,000× target compresses a 2029 machine into one figure. Starling will earn it by preserving logical operations through error correction, real-time control and customer workflows. The number belongs to the system that preserves the work.
IBM coverage shifted from research toward enterprise
| Coverage frame | 2024 compared period | 2026 compared period | Change |
|---|---|---|---|
| Enterprise | 32.3% | 42.9% | +10.6 percentage points |
| Research | 22.6% | 5.7% | −16.9 percentage points |
Frequently asked questions
What is IBM’s Starling quantum computer target?
IBM says Starling should deliver 20,000 times the compute of today’s quantum computers by 2029. Achieving that target requires dependable logical operations across error correction, decoding, control electronics and software.
Why does IBM want to acquire HRL Laboratories?
HRL would add electron-spin quantum circuits to IBM’s existing superconducting program. If the deal closes, IBM could compare two physical architectures internally while integrating both with its control and software stack.
What role does AMD hardware play in IBM’s quantum roadmap?
IBM said it ran a key quantum algorithm in real time on commonly available AMD field-programmable gate arrays. That makes classical-control latency and reliability part of the quantum system’s performance.
How should companies evaluate a quantum chemistry pilot?
They should define a specific molecular or materials task, establish a high-performance-computing baseline, and compare total runtime, end-to-end error rates, control latency and hybrid-run support. The relevant question is whether the quantum component improves the complete workflow enough to justify integration.
What does the IBM-Cisco project aim to demonstrate?
IBM and Cisco plan a proof of concept for linking quantum computers over long distances by 2030. It tests an interconnect layer needed to move beyond isolated processors toward networked quantum systems.