Diraq and NVIDIA: Accelerating the Path to Utility-Scale Quantum Computing
- 2 hours ago
- 3 min read
At Qbeat Ventures, we invest at the frontier, and quantum computing doesn't get more frontier than this.
We're proud backers of Diraq, the Australian quantum computing company building silicon-based processors for utility-scale quantum computing.
This week, Diraq announced a deepened collaboration with NVIDIA, and it's a milestone worth unpacking.
The Bet: Build Tomorrow's Quantum Computers on Today's Chip Lines
Diraq's strategy has always been elegant in its simplicity: use the same CMOS manufacturing processes that already power the world's semiconductors to build silicon-based quantum processors. No exotic infrastructure, no reinventing the fab. The goal is a cost target of under $1 per physical qubit, with full compatibility for data center deployment - quantum processors that can sit alongside GPUs as accelerators, rather than requiring bespoke facilities.
That architectural choice is what made this week's announcement possible.
Why Classical Computing Is the Hidden Bottleneck
Diraq's silicon spin qubits operate on nanosecond timescales, making them among the fastest qubits available today. But speed creates its own challenge: quantum processors never operate in isolation. They need classical compute to orchestrate control sequences, analyze measurement data in real time, and run the adaptive feedback loops that make error correction and calibration possible at scale.
One clear example is qubit calibration, identifying transitions in charge stability maps, a multi-dimensional task that has traditionally required expert physicists to spend hours manually labeling data and sweeping parameters by hand. Multiply that across thousands or millions of qubits, and the old way of doing things simply doesn't scale.
The Architecture: GH200, NVQLink, and NVIDIA Ising
To solve this, Diraq integrated its silicon quantum processors with the NVIDIA GH200 Grace Hopper Superchip, connected via NVIDIA NVQLink, achieving an ultra-low-latency round-trip time of approximately 3.3 microseconds between QPU and GPU. The system runs on NVIDIA CUDA-Q, orchestrating quantum and classical workflows in a single pipeline.
On top of that infrastructure, Diraq is now using the NVIDIA Ising family of open models. Ising Calibration's GPU-accelerated, vision-capable AI analyzes Diraq's charge stability maps and identifies functional qubits automatically, turning raw measurement data into calibration signals without manual labeling.
The GH200 was first integrated into Diraq's system in May 2025. Within a single week, the team had already demonstrated three real-time applications addressing core scaling challenges.
The Results
The before-and-after is stark:
Calibration, compressed. What used to consume a full year of an expert physicist's time can now be trained in days.
Real-time, not offline. Diraq can now close the loop between measurement and analysis live, during the experiment itself, rather than waiting days to redesign the next run.
Physicists freed up. With manual tuning automated, Diraq's team can spend its time on what actually moves the needle toward utility scale: scalable architectures and new error-correction pathways.
As Diraq CEO and founder Andrew Dzurak put it: NVIDIA is changing Diraq's route to utility-scale quantum computing by enabling real-time feedback and high-throughput automation across the stack, and the two companies are, in his words, the quantum and classical sides of the same coin in terms of cost efficiency and deployability.
Why We're Excited
This collaboration is a strong proof point for the thesis we've held since backing Diraq:
✦ Nature-published results confirming manufacturability at scale
✦ Selection for Stage B of DARPA's Quantum Benchmarking Initiative
✦ $38M in U.S. CHIPS Act funding - one of only 9 companies selected globally, with the U.S. government taking an equity stake
✦ Australia's National Reconstruction Fund as a strategic investor
✦ A Series B round on the horizon
And now, a real-world demonstration that Diraq's CMOS-first architecture isn't just easier to manufacture, it's easier to integrate into the AI and HPC ecosystems that will define utility-scale quantum computing. By building on infrastructure that's already ubiquitous in data centers, Diraq gains access to mature software stacks, deep documentation, and a talent pool that already speaks the language of GPU computing.
The quantum computing industry has moved past the question of whether quantum computing is possible. The defining challenge now is utility-scale, delivering real-world value at a cost and footprint the world can actually deploy. Diraq's approach, leaning on dense on-chip integration and existing chip ecosystems, is exactly the kind of pragmatic engineering that gets quantum computing out of the lab and into production.
This is what deep tech progress looks like: rigorous science, patient capital, and a technology architecture built for the real world.
Diraq is a Qbeat Ventures portfolio company. Read the full announcement on Diraq's newsdesk.