I was thinking this morning... Why are big-tech companies pushing hard on building bigger and bigger clusters for AI models? I mean, I know it's a race, and even half a year could make a difference (and the way Grok 3 jumped at the top of the models after being trained on a bigger cluster than the competition says they are right in doing so), but... at what stage is quantum computing now?... was my inner thinking. Is it worth it to build huge datacenters now using potentially soon-to-be obsolete technology, or wait for stable quantum processors?
So... I started doing a bit of research on the subject.
First of all, as I suspected, quantum computing would really speed up the training phase of AI models, by parallelizing the process. It wouldn't have a big impact on inference, except on very specific tasks, particularly related to cryptography, but not only.
When Useful Quantum Computer?
I remembered Google is focusing intensely on its quantum processor, and trying to reduce errors associated with this new technology, and the fact that a qubit's value is not discrete (either 0 or 1), but rather anything in the range between 0 and 1.
So, I wanted to see if other top AI companies focus on quantum computing too. And they do. We also have Microsoft and Nvidia working on that. Meta, surprisingly, decided it's too far away, and ignores it. Maybe they got burned with the metaverse and don't want to be too early in something else.
They are not the only ones interested in quantum computing in the US. We also have IBM, Amazon, and Intel at this party.
Not the best image Ideogram generated for me.
Both Meta and Nvidia say it's decades away until quantum computing could have useful applications. Jensen Huang, the CEO of Nvidia, said at the beginning of the year at the Consumer Electronics Show in Las Vegas:
If you said 15 years for very useful quantum computers, that would probably be on the early side, If you said 30, it’s probably on the late side. But if you picked 20, I think a whole bunch of us would believe it.
Only a month and a half later, from an article in GeekWire we find that the CEO of Microsoft, Satya Nadella, comes to contradict everyone who thinks it would take decades for quantum computing to be useful. In his words:
We believe this breakthrough will allow us to create a truly meaningful quantum computer not in decades, as some have predicted, but in years.
What Microsoft did compared to competition was to focus on developing new quantum technologies instead of trying to stabilize (reduce errors) for existing quantum chips.
This is interesting to say the least... Years is not decades. Of course, even if a useful QC is reached within years, it won't scale rapidly.
Quantum computing is as bigger of a race as AI, but probably not as much hyped up, since regular people can't use it. But let's take a quick look...
China versus Europe versus the United States on Quantum Computing
Surprisingly... Europe is still in this race, unlike in the AI one, where we just woke up.
Europe is competitive on research in this field but is lacking in deployment of applications. China dominates quantum communications, including two quantum satellites. Quantum communications allow unbreakable encrypted messages even using quantum computers, without alerting the original sender and receiver.
The US has a lead on quantum computers and recently started to take interest into quantum communications.
There is also quantum sensing, for better measuring devices, where Germany seems to have a lead, with the US and China at about the same level.
For a lot more information on how things stand and how they are likely to evolve, you could dive into this report.