A Sydney-based startup with fewer than a hundred employees has just signed a letter of intent that could quietly redraw the semiconductor power map: $38 million in potential funding from the U.S. CHIPS and Science Act’s R&D Office. Diraq may be obscure to mainstream tech audiences, but in quantum computing circles, it’s long been tagged as “the most manufacturable.” This isn’t an investment or a subsidy—it’s a high-stakes bet: that silicon spin qubits can be scaled using existing CMOS fabrication lines.
Don’t let the word “quantum” intimidate you. This isn’t another academic debate over superconducting loops, trapped ions, or topological states. Diraq’s real ambition? To run fault-tolerant quantum processors through TSMC’s 12-inch fabs—the very same facilities churning out NVIDIA’s H100 GPUs. If successful, it would shatter the industry dogma that quantum computing must build entirely new infrastructure from scratch. And TSMC, famously risk-averse, has already quietly reserved pilot production slots for Diraq. In my view, they’re not chasing short-term wafer revenue—they’re positioning for dominance in the post-Moore era.
NVIDIA’s role is even more intriguing. On paper, it’s absent from this deal. But recall Jensen Huang’s cryptic remark at last year’s GTC: “AI needs new physics.” As GPU scaling hits diminishing returns, quantum co-processors emerge as the only plausible next leap. Diraq’s silicon-based approach is inherently CMOS-compatible—meaning future quantum accelerators could be monolithically integrated into Grace Hopper-class superchips. No fiber-optic interconnects. No stacks of dilution refrigerators. Just a dedicated “quantum island” inside the SoC package. That’s the holy grail of heterogeneous computing, and I suspect NVIDIA’s architects are already stress-testing Diraq’s roadmap internally—even if they haven’t gone public yet.
History offers parallels. ARM leveraged the iPhone in 2007 to shake the x86 empire; Cambricon ignited China’s AI chip frenzy in 2016 with TPU-inspired designs. But quantum is different—it lacks clear commercial pathways and standardized benchmarks. $38 million sounds substantial until you realize it barely covers three months of electricity for a leading-edge fab. Why is the U.S. Department of Commerce betting on it? The answer lies in the deeper logic of the CHIPS Act: rather than exhausting resources chasing 2nm and 1.4nm transistor shrinks, why not open a second front? If silicon qubits succeed, America secures both its classical computing moat and a head start in the next computational paradigm.
Yet the risks are existential. Coherence time remains the Achilles’ heel of silicon spin qubits. Even if Diraq’s claims of millisecond-scale operation at 1 Kelvin hold up—and that’s a big if—it’s still orders of magnitude short of practical error correction. More critically, will TSMC truly retool its line for a product with zero customers? A single validation tape-out consumes precious EUV machine hours that could otherwise serve Apple or AMD. And if the effort fails, this $38 million becomes just another sunk cost. Ask yourself: if in five years we discover that “CMOS-compatible quantum chips” were merely an engineer’s mirage, was today’s funding visionary—or just political placebo for techno-nationalism?
The real battle won’t be won in labs, but in ecosystems. If NVIDIA integrates Diraq’s tech, it instantly gains market access. If TSMC releases a quantum-aware PDK, silicon qubits enter the semiconductor mainstream. Without that, even the most elegant physics remains confined to academic journals. I’ve always believed breakthroughs never happen in isolation—they demand corporate endorsement, patient capital, and a dash of historical luck. Diraq now has all three. But time is running out. Moore’s graveyard is already littered with chips that “could have changed the world.” This time, will silicon-based quantum be the gravedigger—or the next tombstone?