Palo Alto – February 18, 2025 – Now that we have Generative AI, surely, we don’t need Quantum computing!

Last Thursday I was rigorously persuaded that this statement is far from true! The two technologies are complementary not competitive. AI is being trained today on the output of classical computers and cannot help solve, problems that can only be calculated using quantum techniques. When quantum computers start producing useful output, Generative AI could be trained on that output but until then there is simply no training data available.

What will Quantum Computing be useful for?  Here are 3 big ideas:

1. Drug design: today even with the massive amounts of classical computing power now available, drugs are discovered not designed.
2. Catalysts for chemical reactions: Imagine new catalysts that would reduce the temperature and pressure required to create industrial chemicals. Imagine a new catalyst that would massively reduce the cost of extracting CO2 from the air.
3. Superconductors: Since the 1950s’ engineers have been dreaming of discovering a room-temperature, ambient pressure, and super-conducting material. A quantum computer could design this magic material.

And there we have the classic quantum chicken & egg problem; a quantum computer could design a much better quantum computer.

Last night’s fireside chat at the Magic Playground featured:

• Peter Barrett, Founder & General Partner at Playground
• Ashley Montanaro, Co-Founder and CEO of Phasecraft and Professor of Quantum Computation at the University of Bristol, UK
• Jennifer Dionne, Co-Founder of Pumpkinseed and Professor of Materials Science and Radiology at Stanford University
• Peter Shadbolt, Chief Science Officer and Co-founder at PsiQuantum

Here are my 3 key takeaways:

1. Quantum sensors are already very useful!

Whilst we wait for Quantum computing to become useful. Some of the money invested in research has already yielded incredibly useful quantum sensing systems.

Ideon, a Canadian Playground portfolio company, detects cosmic muons created by supernova explosions in deep space. The detection can take place deep underground and be used to identify valuable mineral deposits deep in rock that classical sensors cannot see.

NVISION, a German Playground portfolio company, has invented a new kind of easy-to-use MRI scanner. The scanner uses quantum sensing to determine whether a cancer treatment is working just days after the onset of treatment. With a traditional MRI scanner, you would have to wait for months for that same insight.

Pumpkinseed is still in stealth mode but is developing the means to detect diseases in crops and livestock in real time, thereby reducing the need for unnecessary antibiotics, vaccinations and harmful agrochemicals.

2. Better Quantum Algorithms do not need much capital!

So, if you’re the kind of climate tech investor that only does software, you may not have the stomach to start or invest in a quantum computer developer. Do not panic there is big opportunity in the capital light realm of quantum algorithms. To create a useful quantum computer, we need more Qubits, better error correction, and we need better algorithms so that a useful problem can be solved in a reasonable time frame on a computer with not too many more qubits. In recent years algorithms have been improving 1000 times faster than Quantum hardware. Phasecraft is a British Playground portfolio company leading that algorithmic charge.

3. Quantum compute is using the semiconductor supply chain to scale!

The ENIAC was the world’s first electronic digital computer. It was built from 1943 to 1945 for the US Army. It weighed 30 tons and used over 18,000 vacuum tubes, errors, caused by the failure of these tubes, were frequent and hard to correct.

Not for another twenty years would Moore’s law, the observation that the number of transistors on a microchip doubles roughly every 2 years, be first described, by a co-founder of Intel.
 
Google’s Willow quantum processor found fame at the end of 2024, it was particularly good at error correction, and had 105 Qubits. This is (almost) double the number of Qubits that Google’s Sycamore quantum processor had, when first used 6 years ago in 2019.
 
It is commonly believed that a really-useful Quantum computer needs to have 1 million Qubits. So, if we assume we speed up and start doubling the number of Qubits every 2 years, as per Moore’s Law, then a 1 million Qubit computer would still be more than 25 years away.
 
But there is one Quantum computer company, PsiQuantum, that has a strong belief that Moore’s need not be a predictor of progress in Quantum computing. PsiQuantum’s approach seems to be more inspired by the pre-Moore’s law computers of 80 years ago. PsiQuantum is aready building a 1million Qubit computer in Brisbane, Australia and has aggressive plans for it to be operational by the end of 2027.
 
Instead of spending time trying to fit more and more Qubits into a small-space, PsiQuantum will use the scales of the global high-volume semiconductor industry to make millions of photonics devices on standard silicon wafers.  Photonics devices will then be integrated with electronics and temperature sinks into a package that can be made by conventional contract manufacturers. This turns the challenge of building a million Qubit computer into something closer to the challenge of building a million iPhones from a working prototype.
 
PsiQuantum was founded a years ago and unlike most Silicon Valley start-ups that have been around that long, their story has never wavered, they have never pivoted and have never shown any interest in building a Minimum Viable Product.

In Conclusion: Quantum Magic is Real!

So after another evening at the Magical Playground I can conclude that Quantum computing has already produced useful commercial products. More bizarrely I have learnt that Quantum Magic is a real thing. For any given quantum system, magic is a measure that tells us how hard those calculations are on a classical computer. The higher the magic, the more we need quantum computers!