Why the World’s Most Valuable Company Is Buying Into Quantum Computing

In the last two weeks, NVIDIA, the enabler and chief beneficiary of the AI craze, has bought into quantum computing. “I’m a little surprised they haven’t done it before,” says Richard Shannon, an investment analyst at Craig Hallum. The investments from NVIDIA’s venture capital arm and other investors—which collectively value quantum computing startups Quantinuum, QuEra and PsiQuantum at over $17 billion—represent a shift in tone for NVIDIA’s CEO, Jensen Huang. In January, Huang said that useful quantum computers were 15 to 20 years away, sending publicly-listed quantum computing firms tumbling. He walked this back in March, saying in June that quantum computing had reached an “inflection point” and could “solve some interesting problems in the coming years”. A NVIDIA spokesperson declined to comment. By positioning itself at the center of the hardware that AI companies need to run their models, NVIDIA has become the world’s most valuable company, valued at $4 trillion. NVIDIA designs GPUs (the chips specialized for running AI algorithms), develops CUDA (which allows the chips to talk to each other), and packages it all into supercomputers the size of fridges—which AI companies are falling over each other to put into their data centers. Quantum computing is unlikely to help NVIDIA’s AI customers. “Quantum computing and AI are sort of diametrically opposed,” said Pete Shadbolt, chief science officer at PsiQuantum, one of the startups that NVIDIA invested in. AI systems are powerful because they learn patterns from vast quantities of data. Quantum computers, by contrast, “hate data, and they love precision,” said Shadbolt. Proponents believe that quantum computing could usher in a new computing paradigm. Rather than performing numerous simple calculations in parallel, as NVIDIA’s GPUs do, quantum computers could solve a few, extremely valuable equations. Quantinuum does this using ions (charged atoms), PsiQuantum uses photons (light particles), and QuEra uses neutral atoms. These tiny particles obey the strange laws of quantum mechanics: they can be in multiple states at the same time, allowing quantum computers to explore multiple paths of a complex calculation simultaneously, arriving at solutions that current “classical” computers cannot reach in any reasonable length of time. This includes shortening the millions of years that it would take classical computers to break the encryption schemes that underlie much of the digital economy to hours, which has sent banks scrambling for “quantum-resistant” cryptography. There are beneficial applications, too. Quantum computers can simulate quantum mechanical systems, which can’t be done using classical computers, said Hsin-Yuan Huang, a senior research scientist at Google Quantum AI. Quantum mechanics is our most fundamental description of the physical world, and the ability to simulate it could help design new drugs, materials, and chemical processes. “That’s not going to be solvable with just many GPUs. It’s just inherently too hard,” said Hsin-Yuan Huang. For example, quantum computing may unlock greener ways of producing ammonia, which currently accounts for 2 percent of global energy consumption. PsiQuantum has partnered with Mercedes-Benz to understand how quantum computers could simulate lithium-ion battery electrolytes—which might accelerate electric vehicle battery design—and is working with Boehringer Ingelheim, a pharmaceutical company, to understand an enzyme that is involved in the human body’s metabolism of drugs. But without large enough quantum computers to put these ideas into practice, the concrete utility of quantum computers has yet to be demonstrated, said Jan Ole Ernst, a Ph.D. researcher in quantum information and computation at the University of Oxford. It’s possible that this will change when larger quantum computers are available for application development. “But I think that’s also kind of far-fetched, because there’s so much research going into applications, and we haven’t really found anything really clear-cut, apart from factorizing large numbers.” It’s unclear how long it will take for quantum computers to work on the scale where these questions can be answered definitively. Timelines in quantum computing, as in many technically challenging fields, have a habit of stretching. PsiQuantum says that it has started work on sites in Australia and Illinois, and is testing the equipment that will cool their chips to operating temperatures. The company says it will be running a quantum computer large enough to be useful by 2027, two years later than it thought in 2021. If and when quantum computers do start running, NVIDIA is likely to be in the center of the action. “You’ll never be able to run a quantum computer without a ton of classical processing,” said Ernst. Classical computers are needed to control quantum computers, perform error correction and analyze their readouts. PsiQuantum uses NVIDIA’s hardware to prepare their quantum computation and process its outputs. In 2022, the company launched CUDA-Q, which would allow quantum computers to talk to classical computers. This year, it announced the NVIDIA Accelerated Quantum Computing Research Center in Boston, “dedicated to shortening the timeline to useful quantum computing.”  “I think it’s fair to say that they’re doing everything but the quantum computer,” said Shadbolt.  That could change. NVIDIA’s recent investments will give the company “advance notice about which platforms are scaling better,” said Shannon. “I believe, given enough time, it’s a virtual certainty that NVIDIA buys one or more quantum companies.”