The Compute Singularity And Quantum Computing's Inevitable Role - Quantum Zeitgeist

The technology world is buzzing with talk of the singularity. Elon Musk recently declared on X that we have entered the Singularity, followed by “2026 is the year of the Singularity.” These statements came in response to engineers marvelling at how AI tools now compress years of work into weeks. It is a dramatic claim, but one that reflects a growing sentiment across Silicon Valley and beyond. The feeling is that we are approaching takeoff, the point where recursive self-improvement kicks in and progress begins to compound at rates that leave human intuition behind (in a business sense, Jeff Bezos, Founder of Amazon, would most liken it to a flywheel). Ray Kurzweil has been the patron saint of this concept since his 2005 book “The Singularity Is Near.” The inventor and futurist is getting on a bit now, but he has spent over six decades working on artificial intelligence. His track record of predictions is almost uncanny. In 1999, he predicted that AI would achieve human-level intelligence by 2029, a timeline that seemed absurd at the time but now appears conservative. Kurzweil is not a talking head wheeled out for soundbites. He is a bona fide technologist who has been at this longer than almost anyone alive. His latest book, “The Singularity Is Nearer,” updates his thesis for the current AI moment, and his core prediction remains unchanged: by 2045, human intelligence will multiply a billionfold through merger with AI. But let us be precise about what we mean by singularity. This is not a physics singularity in the general relativistic sense. This is fundamentally a compute singularity. It represents the point where technological progress becomes so rapid and compounding that prediction becomes impossible, and machines begin improving themselves faster than humans can track. Musk’s framing is instructive. He describes being “on the event horizon of the singularity,” borrowing the language of black holes to suggest we are at a point of no return, where the gravitational pull of accelerating progress becomes irresistible. Sam Altman, CEO of OpenAI, has echoed this sentiment in equally striking terms. “We are past the event horizon; the takeoff has started,” Altman wrote in a recent essay. “Humanity is close to building digital superintelligence, and at least so far it’s much less weird than it seems like it should be.” He predicts that “intelligence too cheap to meter is well within grasp,” a phrase deliberately echoing the early promises of nuclear energy. Altman’s timeline is aggressive: agents capable of real cognitive work in 2025, systems capable of generating novel insights in 2026, and robots performing real-world tasks by 2027. Whether these predictions prove accurate or optimistic, they reflect the prevailing mood at the frontier of AI development. Dario Amodei, CEO of Anthropic, offers a more cautious but no less striking assessment. He expects AI systems with capabilities that match or exceed those of Nobel Prize winners across most disciplines to emerge by late 2026 or early 2027. Anthropic describes this as “a country of geniuses in a datacenter.” Amodei has warned that humanity is entering a period of “technological adolescence” where we will gain unimaginable power, and “it is deeply unclear whether our social, political, and technological systems possess the maturity to wield it.” The risks are real, but so is the trajectory. The trajectory is clear enough. We are on a quest for more and more computation. Moore’s Law has delivered a roughly million-fold increase in computing power over the past 60 years, doubling every 18 to 24 months. But the AI training buildout of the past decade has accelerated this dramatically, with compute for large language models increasing roughly fivefold annually. That translates to 15,000 times more compute in six years, equivalent to 30 years of traditional Moore’s Law progress. The numbers are staggering. Some researchers talk about “slow takeoff” versus “hard takeoff” scenarios, but either way, the curve is bending upward. We may be approaching something like escape velocity, where the gravitational pull of old constraints no longer applies. Share This is where quantum computing comes into play. Classical computing has delivered extraordinary progress, but there are fundamental limits to how far it can go. We are not at those limits yet, and classical hardware will continue to improve. But the state space available to quantum systems is qualitatively and quantitatively different from that of classical systems. The Hilbert space that quantum computers operate in grows exponentially with the number of qubits, enabling computational capabilities that classical systems cannot efficiently explore. This is not speculation. It is mathematics. Richard Feynman understood this half a century ago. In his famous 1981 lecture at MIT, he declared: “Nature isn’t classical, dammit, and if you want to make a simulation of nature, you’d better make it quantum mechanical, and by golly it’s a wonderful problem, because it doesn’t look so easy.” Feynman was responding to a simple observation. If you try to simulate even a modest quantum system on a classical computer, the computational requirements explode exponentially. A system of around 50+ qubits already exceeds the capacity of any classical computer ever built. The only way to efficiently simulate quantum systems is with quantum systems. This is the origin story of quantum computing. But the argument extends beyond simulation. Nature itself operates quantum mechanically. Photosynthesis achieves near-perfect energy transfer efficiency by exploiting quantum coherence. When photons hit chlorophyll molecules, the resulting excitation energy is efficiently transferred to the photosynthetic reaction centre, with an efficiency approaching 100 per cent. Classical random hopping models cannot explain this. The energy appears to explore multiple pathways simultaneously in a quantum superposition, finding the optimal route with remarkable speed. Research from UCL, Berkeley, and elsewhere has demonstrated that quantum mechanical effects play a decisive role in this biological process. Evolution has spent billions of years optimising quantum effects in living systems. There is no reason to think our computing devices should ignore this same physics. Ray Kurzweil’s Exponential Growth of Computing. Could we see Quantum on the right-hand side as we further push out the timelines beyond classical compute, and the only way we remain on the exponential is with quantum compute? [Image from Ray Kurzweil’s Library] The virtuous circle is already forming. Artificial intelligence is accelerating quantum computing, and quantum computing promises to accelerate artificial intelligence. Machine learning is being used to optimise quantum hardware, discover new materials for quantum devices, and improve quantum error correction. Google’s GNoME system identified 2.2 million stable inorganic crystal structures, including 381,000 new materials more stable than any previously known combinations. AI is compressing what would take human researchers decades into months. Meanwhile, researchers at Yale and Emory have demonstrated machine learning techniques that identify complex quantum phases in materials in minutes rather than months. The feedback loops are tightening. AI helps discover better quantum materials. Better quantum materials enable more capable quantum computers. More capable quantum computers can train more powerful AI systems. Each improvement feeds into the next. This is not a linear progression. It is exactly the kind of compounding acceleration that Kurzweil has been describing for decades. The technical term is “recursive self-improvement,” but you could also call it an intelligence explosion in slow motion. Or perhaps not so slow! Jensen Huang, CEO of NVIDIA, has framed this dynamic as a “virtuous cycle” that has now achieved critical mass. “The AIs get better. More people use it. More people use it, it makes more profit, creates more factories, which allows us to create even better AIs, which allows more people to use it,” Huang explained at the APEC CEO Summit. He describes AI as “the most powerful technology force of our time” and predicts it will reshape $100 trillion worth of industries. NVIDIA’s role as the picks-and-shovels supplier to the AI gold rush gives Huang a unique vantage point on the infrastructure buildout, which he calls “the largest infrastructure buildout in human history.” On the quantum side, Google CEO Sundar Pichai has offered a striking prediction. “I would say quantum is where maybe AI was five years ago,” Pichai told the BBC. “So I think in five years from now we’ll be going through a very exciting phase in quantum.” If this comparison holds, quantum computing is approaching the same inflexion point that AI reached around 2020, just before ChatGPT and the current wave of generative AI transformed the landscape. Google’s recent quantum breakthrough, in which its Willow chip solved a problem in 5 minutes that would take classical supercomputers longer than the universe has existed, suggests this is not idle speculation. Consider the current state of the technology stack. Large language models like Anthropic’s Opus, OpenAI’s GPT series, Google’s Gemini, and Mistral are already performing at levels that rival PhD researchers in many domains (there are plenty of benchmarks that verify these). These systems can synthesise vast bodies of literature, generate novel hypotheses, and work through complex technical problems with increasing sophistication. The agentic systems now being developed can chain these capabilities together, creating workflows that autonomously conduct research, write code, and iterate on solutions. When these tools are applied to quantum computing research, the pace of discovery accelerates further. There is a natural objection to be addressed. Quantum computers remain challenging to build and operate. Current systems are noisy, error-prone, and limited in scale. This is true. But the trajectory matters more than the snapshot. The same objections were raised about classical computers in the 1950s, about integrated circuits in the 1960s, and about practical AI as recently as a decade ago. Each time, exponential improvement rendered the objections obsolete. The quantum computing industry is following a similar curve, with qubit counts ( and quantum volume metrics), coherence times, and gate fidelities improving steadily. The argument here is not that quantum computers will replace classical computers wholesale. Classical computing will remain essential for the vast majority of tasks. The argument is that quantum computing represents the next frontier in expanding computational capability. As we push toward the compute singularity, the classical phase space will eventually prove insufficient. Quantum systems offer access to an exponentially larger computational space, one that may be necessary to sustain the compounding growth that defines the singularity. We have perhaps reached a point where quantum computing is no longer a side project or a speculative research direction. It is becoming an integral part of the computational landscape. The technology industry is investing tens of billions of dollars annually. Major cloud providers offer quantum computing services. Governments are establishing national quantum initiatives. The quantum workforce is expanding rapidly. Not the signs of a technology on the margins. The singularity, if (or when) it arrives (or to some may have already arrived), will not be built on classical computation alone. The universe is quantum mechanical, and the most efficient way to compute is likely to follow nature’s lead. Feynman saw this clearly. Kurzweil’s law of accelerating returns suggests that paradigm shifts become increasingly common as technology advances. Quantum computing may well be the paradigm shift that enables continued exponential growth when classical approaches reach their limits. We are on the “ramp”. The on-ramp to accelerated development that Musk and Kurzweil describe is becoming visible. Some call it FOOM, the onomatopoeic term for rapid capability gain or ‘Hard Takeoff’. Others talk about capability overhang, the idea that latent potential has been accumulating and could be released suddenly. Whatever the terminology, the LLMs we use daily are already transforming how research is conducted, how problems are solved, and how knowledge is synthesised. The next generation will be even more capable. And at some point, the classical substrate may no longer suffice. When that moment arrives, quantum computing will not be an exotic option. It will be a necessity. The compute singularity will be at least partially quantum.