⚛️ Majorana 2, 💡 Valleytronics and 🔧 ChipStack: the virtuous circle of hardware

Microsoft used AI to design materials for its Majorana 2 quantum chip — and the result is qubits 1,000 times more stable. A team at Monash University created the first chip that processes information with light instead of electricity. And Nvidia and Cadence built an AI agent that designs chips by itself, with 10x productivity improvements. They're not three separate stories: it's the same virtuous circle. AI designs chips that run AI that design better chips.

⚛️ Microsoft Majorana 2: the quantum chip that AI helped design

On June 2, Microsoft unveiled Majorana 2, its second-generation quantum chip. The big news isn't just that its qubits are 1,000 times more stable than Majorana 1 (20 seconds vs milliseconds) — it's that the chip's materials were optimized by AI.

Microsoft used AI models to discover the exact combination of lead-based materials that allow topological qubits to function at scale. The result: a quantum chip capable of maintaining quantum coherence 1,000 times longer, paving the way for a viable commercial quantum computer — which Microsoft says will arrive in 2029.

The scientific community is cautious. Physicists point out that 12 qubits are still far from the million needed for real commercial problems. But the jump in qubit quality is undeniable, and using AI to discover new materials is a milestone in itself.

COMPUTEX 2026 showed that the chip race is accelerating. Microsoft just upped the ante on the quantum front, and did it with a tool that until recently seemed like science fiction: AI designing hardware.

From silicon to light: while Microsoft advances on quantum, Monash shows there's another way to compute.

💡 Monash creates the first valleytronics chip: light instead of electrons

On the same June 2, a team from Monash University (Australia) published in Nature Photonics the creation of the first integrated chip capable of generating, directing, and reading information using light instead of electricity.

The technology is called valleytronics and is one of the great promises of future computing. Instead of using electrons (like traditional chips) or qubits (like quantum), it uses a quantum property of light called a "valley" to encode information. The chip operates at room temperature, without needing extreme cooling.

Dr. Chi Li, lead author of the study, explains that until now you could generate or detect these signals, but not do both on a single integrated device. His team solved that problem by combining ultrathin materials (just a few atoms thick) with meta-surfaces designed to control light at minuscule scales.

We talked at the start of this blog about how traditional computing is approaching its physical limits. Valleytronics is exactly the kind of paradigm shift needed to surpass them — and now we have the first functional chip.

And while new computing methods are invented, AI is already designing today's chips 10 times faster.

🔧 Nvidia + Cadence: the first AI agent that designs chips autonomously

On June 1, at COMPUTEX, Nvidia and Cadence announced the ChipStack AI Super Agent: an autonomous Level 5 AI system capable of designing and verifying chips without human intervention.

The agent receives the design team's requirements, generates RTL code (the abstraction describing how a chip works), creates verification plans, runs simulations, detects bugs, and corrects them automatically. All in a closed loop that previously required weeks of specialized engineers' work.

Jensen Huang, Nvidia's CEO, presented it at the COMPUTEX keynote with visible pride. And with good reason: this agent isn't a lab experiment. Nvidia has tested it with its own design team, and the result is 10x productivity improvements in design and verification code generation.

The agent runs inside OpenShell, a secure sandbox developed by Nvidia that allows AI agents to execute code and access tools safely. We already covered OpenShell in the COMPUTEX analysis — infrastructure for autonomous agents advances at the same pace as the chips that make it possible.

Microsoft, Monash, Nvidia and Cadence. Quantum, light, silicon. Three stories that seem from different labs but tell the same story: we're entering a feedback cycle where AI and hardware accelerate each other. Every advance feeds the next. The interesting question isn't when it will arrive, but who will be ready when it does.

— Max