IBM has signed a definitive agreement to acquire HRL Laboratories, a U.S. research organization jointly owned by Boeing and General Motors. The deal amount has not been disclosed, and the transaction is expected to close by the end of Q3 2026, pending regulatory approval and other conditions. The value IBM cites for the acquisition spans a wide range—from silicon spin qubit design to manufacturing technology. HRL has developed, as an integrated whole, the semiconductor processes for making quantum dots, cryogenic control circuitry, and signal wiring.
This acquisition is not an announcement that IBM is switching its superconducting quantum computer "Starling," slated for delivery in 2029, to a spin-based approach. Rather, it is a move to add in-house capability for bringing a different qubit modality to manufacturing, in order to go beyond the existing roadmap. Tracing the prototype HRL unveiled in 2026 reveals what IBM is trying to buy beyond mere qubit count.
An Agreement to Acquire a Jointly Owned Boeing/GM Research Lab
HRL was founded in 1948 as Hughes Research Laboratories and transitioned to a limited liability company in 1997. It is currently jointly owned by Boeing and GM, conducting R&D for both companies as well as the U.S. government and private-sector clients. Even after the deal with IBM closes, Boeing and GM will continue their partnership with IBM in developing quantum applications and advanced technologies.
The acquisition agreement covers HRL as a company; it is not described as a carve-out deal limited to the quantum computing division alone. Besides quantum sensing and quantum networking, HRL works on advanced materials, high-speed/high-power communications, electronics, and more. In Southern California, it operates a 250,000-square-foot research facility including a 10,000-square-foot Class 10 cleanroom, and it holds Trusted Foundry certification for U.S. Department of Defense work. IBM also cites HRL's capability to move research for government and commercial clients into prototyping as complementary value gained through the acquisition.
Meanwhile, the acquisition price, the number of employees transferring, and the treatment of government contracts and facilities remain undisclosed. HRL's own announcement states that the agreement marks the start of regulatory review. What is confirmed at this point is the acquisition agreement itself; the transition under IBM has not yet been completed.
The Exchange-Only Approach: Making One Qubit from Three Electrons
The superconducting qubits IBM currently uses encode information in the quantum states of circuits that include Josephson junctions. HRL's exchange-only qubits work differently. Three quantum dots are formed in a silicon/silicon-germanium (Si/SiGe) quantum well, and the combined spin state of three electrons confined in each dot encodes a single qubit. Here, "quantum dots" refers not to the light-emitting nanocrystals used in display devices, but to fine structures that confine single electrons using voltages applied to metal gates.
Control is achieved via voltage pulses that bring neighboring electrons closer together or move them apart. Rather than individually rotating electron spins, the exchange interaction between electrons partially swaps spin states. This configuration uses more physical resources—three electrons per qubit—but avoids correlated errors associated with microwave driving and allows gates to be operated using electrical signals alone.
In 2023, HRL formed two encoded qubits from six quantum dots and demonstrated, in Nature, a universal gate set sufficient for arbitrary quantum computation. The encoded CNOT fidelity was 96.3%±0.7%, and the encoded SWAP fidelity was 99.3%±0.5%. Here, "universal" means the necessary logical operations can be combined—it does not mean large-scale computation or fault tolerance has been achieved.
An 18-Qubit System Supported by a 4K Controller and Superconducting Wiring
A preprint the HRL team released in April 2026 expanded the scope of research from two-qubit gates to system integration. Using a proprietary 200mm wafer process, 54 quantum dots are arranged in three rows, enabling a configuration of up to 18 exchange-only qubits. The chip can be used as a 3×6 lattice, but "up to 18" represents the configurable scale; the multi-qubit experiments in the paper use only a portion of this.
What stands out in the prototype QPU is that a 130nm RF CMOS controller carrying about 70 million transistors was operated at 4K. The controller generates 150 time-varying signals and 35 static bias lines, sent to the qubit chip—with an average electron temperature of 150mK—via low-thermal-conductivity superconducting ribbon cables. This design reduces the need to route large amounts of precise analog wiring from room temperature down to the cryogenic stage, shifting qubits, control, and wiring toward components that can be fabricated using semiconductor wafer processes.
Gate performance has also advanced. The preprint reports average errors of 2×10^-4 for single-qubit gates and 3×10^-3 for CNOT. In a distance-5 repetition code using 7 qubits, the logical error rate was approximately 5×10^-3, a 4.7x improvement over the average of 2.3×10^-2 for a distance-3 subset drawn from the same data.
However, this is a pre-peer-review report. In a [[4,2,2]] error detection experiment using 6 physical qubits, about 77% of 10,000 execution results were discarded based on detection outcomes, and the retained data yielded a logical fidelity of 0.95. Using all the data, the average was 0.59. Moving from a stage of detecting errors and selecting only good results, to a fault-tolerant system that corrects errors without halting computation, will require further refinement of device uniformity, controller power consumption, and wiring manufacturability.
Overlapping Semiconductor Manufacturing Between Superconducting and Spin Approaches
The superconducting and spin approaches differ in the physical systems that hold quantum information. Even so, IBM cites, as points of overlap justifying the acquisition, the fact that both approaches use silicon fabrication, cryogenic environments, and gate control via external signals. In IBM's explanation, spin qubits may potentially operate at temperatures as high as 1K in the future—higher than the 0.015K used for the superconducting approach. However, HRL's qubit experiments this time were conducted at an average of 150mK. The general potential and the prototype's actual achievements should be read as separate matters.
As temperature rises, the amount of heat a cooling system can remove increases, making it easier to place control electronics closer to the qubits. The reason HRL's prototype separated the 4K controller from the 150mK qubit chip and connected them via low-thermal-conductivity ribbon wiring is precisely to handle this thermal design and wiring density simultaneously. What IBM gains is not just a new qubit modality, but also a design approach for breaking down cryogenic systems into manufacturable components.
HRL also conducts research in quantum sensing, superconducting nanowire single-photon detectors, and silicon carbide photonics. In its acquisition announcement, IBM expressed intent to eventually connect computing, sensing, and networking. However, it has not indicated any integrated product or delivery timeline for these. There remains a product-development process between broadening the scope of quantum research and increasing the number of systems customers can actually use.
Starling in 2029 Remains Superconducting
Under IBM's published roadmap, the 2029 Starling is a fault-tolerant superconducting quantum computer executing 100 million quantum gates across 200 logical qubits. The subsequent Blue Jay targets 2,000 logical qubits and 1 billion gates from 2033 onward. Both remain future targets, and this acquisition announcement does not change the modality or timeline. HRL's spin qubits are described as technology for IBM to explore expansion methods for "the next decade."
A potential manufacturing connection point is "Anderon," the quantum wafer foundry concept IBM and the U.S. Department of Commerce announced plans for in May. Under the plan, the U.S. government would invest $1 billion in CHIPS support and IBM would invest $1 billion in cash to build a 300mm wafer fab in Albany, New York. It would initially handle superconducting qubits and associated electronics, later expanding to other modalities. However, this concept itself remains at the memorandum-of-understanding stage, contingent on finalizing definitive agreements.
IBM has mentioned that the HRL acquisition could deepen its collaboration with Anderon and potentially lead to developing spin qubit manufacturing. If HRL's proprietary 200mm process is to be connected to Anderon's planned 300mm process, points to confirm will include how process design kits, in-line inspection and characterization, and baseline processes are adapted for the spin approach. IBM has not indicated a timeline or volume for any such process transfer.
In June, IBM announced plans to invest over $10 billion over five years in the quantum field, covering R&D, manufacturing, and M&A. The HRL acquisition gives concrete form to this investment strategy, but the $10 billion figure is not the acquisition price. Once the deal closes, when will IBM present manufacturing targets and a product roadmap for spin qubits? Only then will it become possible to gauge how the lab acquisition connects to IBM's next-generation quantum computers.