PsiQuantum’s $6B valuation lures Nvidia into quantum hardware

First came “Quantum Day,” Nvidia’s March 20 showcase at GTC put CEO Jensen Huang side-by-side with quantum players he had dismissed on January 7 at CES. Now, according to The Information and Reuters, Huang is in talks to invest in photonic-qubit upstart PsiQuantum, underscoring potentially changed thinking after quipping at CES that “very useful” quantum machines are likely somewhere between 15 and 30 years out.

Talks are reportedly in advanced stages. If the deal goes through, Nvidia would join BlackRock in PsiQuantum’s current BlackRock-led $750 million raise, potentially valuing the Palo Alto firm near $6 billion, as Reuters also confirmed. This new funding builds on PsiQuantum’s previous successes, including a 2021 Series D of $450 million, which valued the company at over $3 billion. The rumored stake would also mark Nvidia’s first direct bet on hardware qubits. That would extend the mea culpa that began with Quantum Day into a full-blown strategic hedge against the limits of classical compute scaling.

Huang’s potential hedge with PsiQuantum builds on Nvidia’s existing quantum initiatives like cuQuantum (its SDK for GPU-accelerated quantum circuit simulation) and QODA (Quantum Optimized Device Architecture for hybrid systems), plus its Boston-based quantum research lab opened in April. Nvidia also recently took a minority stake in the quantum-inspired startup SandboxAQ in a deal Reuters reported on last month.

Investing in PsiQuantum would also tether Nvidia to the reportedly only quantum shop fabricating qubits on a 300 millimeter GlobalFoundries CMOS line and targeting a million-qubit system by 2028. Early runs deliver 99.98% state-prep fidelity. Significant public investment supports the ambition. For example, Illinois has earmarked over $500 million for a South-Side quantum campus where PsiQuantum will anchor, per WSJ.

As PsiQuantum co-founder Mark Thompson noted, its chips have benefits in terms of manufacturability. They are made in a fab that “normally produces chips for cell phones and the automotive industry and now yields the world’s highest-performance photonic qubits.”

The firm has also made strides in overcoming a traditional challenge in photonic qubits: minimizing loss in waveguides and coupling. PsiQuantum’s latest Nature paper (2025) details ultra-low-loss silicon nitride (SiN) photonics in its Omega chipset, reportedly achieving waveguide propagation losses below 0.1 dB/cm. That is, the company has developed and benchmarked a set of monolithically integrated silicon photonics modules that support the essential operations for photonic quantum computing: generating, manipulating, networking, and detecting photonic qubits. As they note in the Nature paper, they also demonstrate high-fidelity performance for several key operations.

PsiQuantum aims to hit 1 million physical qubits by 2028, north of rivals’ hardware road maps. A million physical qubits, however, does not equal a million useful qubits. For instance, Google’s prototype logical qubit used 49 physical qubits to encode one logical qubit. In general, independent physicists have reported that aggressive error correction would be required to turn physical qubits into the number of logical qubits needed for, say, chemistry or materials workloads.