Chad Rigetti returns with Singletree, a quantum-accelerated AI server company targeting data center deployment

Chad Rigetti's new bet: quantum co-processors for AI data centers

Chad Rigetti spent a decade building Rigetti Computing, one of the first quantum computing companies to go public, before stepping back and asking a different question: what if the right architecture for quantum wasn't a single qubit technology scaled up, but a system designed backwards from what AI workloads actually need?

That question is Singletree. The Ann Arbor and San Francisco company is building what Rigetti describes as quantum-accelerated AI servers, designed to sit alongside GPU and XPU pods in the data center and act as a co-processor. The pitch is not quantum replacing classical compute. It's quantum augmenting it, with simulations already indicating potential speed-ups of several orders of magnitude on key training tasks.

The multimodal qubit bet

Every major quantum hardware company, IBM, Google, IonQ, Quantinuum, has built its identity around a specific qubit technology: superconducting qubits, trapped ions, photonics, neutral atoms. Rigetti argues that's the wrong frame. Customers are buying a computer, not a qubit type. Modern classical computers aren't built from one transistor type, and he believes quantum systems for AI won't be either.

Singletree's approach is to architect across modalities, pulling from whichever hardware substrates best meet the requirements of AI workloads as those technologies mature. Rigetti calls this multimodal quantum hardware, and argues it has to be baked into a company from the start. You can't bolt it onto an organization that has years of institutional commitment to one qubit choice.

“At Singletree, we're building quantum accelerated AI servers for the data center to bring quantum technologies directly into the data center. It acts as a co-processor for the GPU or XPU pods that have become the unit of compute in AI infrastructure today. Our simulations indicate we expect several orders of magnitude potential speed up for key training tasks — not a factor of two or five, but several orders of magnitude when all the pieces come together. I think quantum can happen in the next five to seven years in the data center running production workloads.”

Where quantum fits in training

The insertion point Singletree is targeting is specific. Rigetti describes looking for places in the existing training and inference workflow where a quantum algorithm can take a step that would take a day or two classically and compress it to hours or minutes. The fundamental constraint is the classical-quantum interface: data goes in classically and comes out classically, so the less translation required, the better the performance. That boundary shapes the architecture.

Longer term, Rigetti is pointing toward quantum-native models, designed from the ground up to use quantum compute tightly integrated with classical infrastructure. That's not a near-term product; it's the asymptotic direction. The near-term product is the co-processor server unit, targeting a one-to-one attach rate with GPU and XPU pods in future data center builds.

Timeline and commercial conversations

Singletree is already in conversations with potential customers, which Rigetti frames as essential rather than premature. Understanding what actually moves the needle for an AI infrastructure buyer shapes hardware requirements early. He puts the broader inflection point for quantum in the data center at five to seven years out, which is when he expects quantum computers running production AI workloads without users needing to know or care that quantum is involved.

The sentiment problem in quantum is real, he acknowledges. The companies making genuine progress can't be evaluated on growth metrics. The honest framework is whether a company is systematically buying down technical risk, which is a harder read than a revenue chart.

Why not one of the existing quantum companies

The application of quantum to frontier AI is not yet consensus within the quantum industry, Rigetti says. Ask a cross-section of quantum hardware leaders about use cases and the median answer is quantum chemistry and optimization, not AI training. That's partly because the algorithms needed for AI-specific quantum acceleration are still being developed. Singletree is functioning simultaneously as a quantum-native AI research lab and a hardware company, which is a combination that didn't exist in the previous generation of quantum startups.

Rigetti took Rigetti Computing through Y Combinator in 2014, part of Sam Altman's first YC batch to include hard tech companies alongside Ginkgo Bioworks and Aqua. The company went public via SPAC in early 2022. Singletree is the reset: same founder, different architecture, different market target.