Key Takeaways
- Quantum computers operate at clock speeds measured in hundreds or thousands of hertz, whereas classical silicon chips run at multiple gigahertz.
- The core theoretical speedup relies almost exclusively on Shor's algorithm for integer factorization and the elliptic curve discrete logarithm problem, which offer zero utility for enterprise workloads.
- Without quantum RAM or persistent storage, quantum hardware cannot compete with Nvidia GPUs on everyday matrix multiplication.
- Martin Shkreli tested current machines directly: “Chelsea and I have tried every single quantum computing quantum computer in the world. We cannot multiply 300 by 200 on any of these.”
The Academic Magic Trick
Martin Shkreli is shorting quantum computing stocks because he thinks the industry sells an academic parlor trick as enterprise infrastructure. Wall Street treats quantum processors like the natural successors to classical microchips. Shkreli sees them as an evolutionary dead end, closer to vacuum tubes than next-generation GPUs.
“The modern quantum computers run at a clock speed of hundreds of hertz or thousands of hertz, not at the gigahertz that we're used to,” Shkreli said. “They're single core and they're single threaded when it comes to doing a job like a merge sort or something like that.”
To run classical workloads, speed depends on clock rate, cache coherence, and memory bandwidth. Quantum processors lack all three. When investors hear that quantum systems solve problems in polynomial time that take classical machines exponential time, they assume that speed applies to everything from database indexing to machine learning.
It does not. That mathematical shortcut only applies to a tiny class of problems, specifically integer factorization and the elliptic curve discrete logarithm problem through Shor's algorithm. “The only thing it's really good at which indeed is a magic trick is going from exponential time algorithm which is integer factorization or elliptic curve discrete log problem,” Shkreli noted. “The quantum computers can do those in polynomial time, which is remarkable.” Outside of breaking specific encryption keys, those equations do not help you run a web service or train a neural network.
The Hardware Bottleneck
Beyond specialized math, practical computing requires memory. Modern AI workloads scale because graphics chips move petabytes of data across high-bandwidth memory at blistering speeds. Quantum architecture has no equivalent medium.
“So the other algorithms that quantum could do are not much faster,” Shkreli explained. “So they'll never beat an Nvidia machine. They'll never beat a million Nvidia machines. And quantum doesn't have storage. There's no such thing as quantum RAM.”
Without quantum RAM, a system cannot hold state between complex operations. Every input must be prepared, executed before decoherence sets in, and read out probabilistically. If you cannot store intermediary states, you cannot run standard software pipelines. Shkreli compares the current wave of venture-backed quantum startups to historical dead ends: “Do you buy, would you go long punch cards and vacuum tubes? Mercury delay lines, core memory, these are all things that disappeared and went bankrupt.”
Investors keep writing checks because they confuse theoretical physics papers with functional commercial engineering. When you strip away the laboratory press releases and test the machines on basic arithmetic, the commercial narrative collapses.
What to Do With This
Audit your technical roadmap for vendor claims that promise exponential efficiency gains through speculative hardware. If a vendor pitches you on quantum computing or unproven specialized chips, ask them to execute a standard three-digit multiplication or a basic database query on live hardware during your pitch meeting. If their architecture cannot beat an off-the-shelf GPU on basic arithmetic today, remove them from your enterprise budget tomorrow.