I've been a vocal skeptic of quantum computing economic opportunities. I poo-pooed a recent poll showing that 25% of business executives felt that quantum computing would have a significant positive impact. But I read today, Satya hyped quantum in a quarterly phone call. Satya's no fool and I don't think he'd be pumping froth at a time when Microsoft's is absolutely roaring in valuation on the belief that it will be among the chief winners of the ML race. (MSFT is up 2% in the last 5 days. Nice for my retirement, not so great for belief in market efficiency.)
In the last few years of my career, I worked hard to understand QP, even though I stuck to a radically simplified version of the math (rotations on a 2D plane) to the extent possible. Which, as it turned out, is sufficient to pass courses, earn certificates, and get through the "Mike & Ike" textbook. I understand the few known algorithms in the relevant BQP complexity class, and can read and write programs with the known quantum computing operators and gates. So I think my opinion is technically informed.
No one knows how ML will transform the economy, but everyone knows how QP will change one very specific thing: the style of encryption on which the entire digital economy relies. Factoring very large numbers will go from "impossibly difficult" to "not only possible, but pretty straightforward, as long as you've got a quantum computer you can run or rent."
There is no doubt that the low-level security infrastructure will have to shift to post-quantum cryptography (PQC). But that's so low-level that it will directly affect only a handful of libraries. For almost every product, the direct use of PQC applies only at a few choke points, all of which are "plumbing," not profit-driving opportunities.
Not only is PQC plumbing, it does not require quantum hardware to run. There are known algorithms for cryptography, and standards for using those algorithms, that are not vulnerable to quantum factoring attacks. Each transaction with will likely be a little more costly than factoring-reliant cryptography. That added friction, multiplied across the countless transactions of digital commerce, becomes a very big cost. But at the individual company level, it'll be somewhere between a scratch and a nick, not an ER visit. Kilobytes not megabytes, milliseconds not seconds.
Beyond encryption, then, what is the value of quantum computing? What makes it something worth drawing attention to when discussing the economics of a company with a $4T valuation?
Quantum simulation. I'm going to be even less specific than saying "factoring large numbers" here, but if you're willing to do the math, you'd agree that quantum hardware legitimately proposes more accurate simulation of atomic/molecular behavior. That means material science, pharma, lots of electrical / energy "stuff," and even lighting (bonus: maybe even CO2 capture) (penalty: yet another finance-bro arms race to make sure no crumbs of Wall Street profit are available to mere mortals).
Downstream of the R&D-heavy "materials science, pharma, and energy" is, basically, most of the world of physical manufacturing. But still the total count of companies doing quantum simulations will be small: low thousands, not millions. Not, going back to that poll, 25% of medium or large businesses. So now we've lemma'd the question to "Why are low thousands, not millions, of companies important to the economics of a $4T company?"
Lock-in. Oh, excuse me, "sticky workloads," as we say in polite society. Not as off-putting. Quantum R&D is going to be hard and expensive. And while the number of companies will be low, those companies are going to be racing, building Jenga towers, not pyramids. And those companies are not going to move those Jenga towers to a vendor who charges 10% less per Hadamard gate.
Microsoft would like their Azure cloud to be quantum sticky. You've heard of Google and IBM, also front-runners. (Would an IBM comeback warm your heart as it would mine?) There are a number of mid-tier companies likely amenable to a Microsoft-OpenAI-style partnership (putting aside the acrimonious lessons that Satya may be taking from that) and smaller companies that are more likely straight acquisitions by a few of the most deeply pocketed. (You can't discount Huang, Musk, or Zuck just buying enough chips to put themselves at the final table.)
And then there are the state-actors: China, the US, and, I dunno', one of the Commonwealth countries benefitting from the US brain drain? The US is publicly "free market innovations!" but the NSA hasn't spent all their money on spy satellites and they were weirdly blasé when encryption at the "why, you'd need a quantum computer to break this!" level became widespread in the late -00s. And China, of course, makes no bones about involving itself with tech transfer and capital support. For the sufficiently paranoid, one might see the lack of any Chinese company among the prominent leaders as a bit of a "dog that didn't bark" absence. China has demonstrated world-beating quantum communication and has a track-record of government-funded High Performance Computing. I doubt BQP is below their notice.
And that's my answer: low thousands of engineering companies building quantum R&D Jenga towers on someone's quantum cloud in order to be competitive in their own very-big sectors. When I was in Azure ML, it took $1M / quarter to get priority access to our limited $40/hour GPUs. And that was in the -10s and early -20s, prior to ChatGPT 3.5 hitting. Quantum workloads will be faster than ML training workloads (setup times will be a bitch) but I think it's a safe assumption that a significant number of qubits are going to cost considerably more per minute than an 8-GPU A100 V4 VM. So... cloud costs of $10M per year per company? That feels laughably low to me, but still multiplies out to $100B/yr in locked-in cloud revenue.
That's enough to pay attention to.