My Recommendations to the US Department of Energy

The US Department of Energy recently posed a question that may sound technical but is really about power: What should the next generation of scientific computing look like? 

Its Quantum Genesis initiative aims to develop fault-tolerant, scientifically relevant quantum computing capability by 2028. It also describes a National Quantum Supercomputing User Facility that would work with high-performance computing, artificial intelligence, and scientific networks. I happen to be one of the people encouraged to provide input on this process.

Most countries, including Bangladesh, will not design the architecture of that future. They will buy access to it, depend on it, negotiate with it, and build local ambitions on top of it. 

If we are going to be clients of these computing empires, we should at least become informed clients.

My first recommendation to DOE was simple: Do not build a quantum facility with the quantum computer at the center. A useful quantum computer will not sit alone like a miracle machine. It will be an accelerator attached to a larger system for classical computing, artificial intelligence, data, and networking. 

Before deciding what to build on the quantum side, one must understand the classical side: AI workloads, memory, storage, networking, scheduling, and user workflows.

The clearest scientific target is materials science simulation. Better materials simulation could affect batteries, semiconductors, catalysis, superconductors, solar cells, and quantum devices.

But the benchmark should not be whether a quantum computer can run an impressive circuit. That is a laboratory milestone.

The benchmark should be whether an integrated quantum-classical facility can produce a better answer to a materials problem than the best available classical AI and high-performance computing workflow.

My second target is computational: average-case hard stochastic gradient descent. In modern AI and scientific computing, we optimize large systems through repeated noisy updates.

Some problems are not impossible for classical computers in an absolute sense. They are hard on average across real workloads.

A useful quantum method might only make the average case slightly easier, reduce failures, save energy, or reach a good answer with fewer evaluations. That could still matter at national or industrial scale.

This is why the field should move away from dramatic language about “classically impossible” problems. Users care less about worst-case theory than about whether a workflow becomes faster, cheaper, more accurate, more reliable, or more energy-efficient.

Quantum computing must compete against the best classical AI systems of the same period, not outdated baselines.

On the hardware side, the key challenge is not only more qubits. It is choosing the right materials for every layer of the quantum computing stack: substrates, wiring, links, cryogenic packaging, control electronics, and interfaces to classical machines. The future machine is not only a physics problem. It is a materials-integration problem.

My strongest software recommendation was an AI-assisted testbed for quantum algorithm discovery. We should not assume that the most important quantum algorithms have already been discovered.

AI systems, human theorists, hardware teams, and HPC researchers should search for useful quantum subroutines, test them against classical baselines, and turn promising ideas into realistic resource estimates.

DOE should also encourage a common language between AI and quantum computing, so qubits and gates can be compared with data, compute, cost, and performance.

This brings me to Bangladesh. I am not saying Bangladesh cannot become a leader in high-performance computing or that it cannot produce original research. It absolutely can.

My point is different: I would not be surprised if that leadership emerges despite ministries, bureaucracy, and politicians, not because of them.

Bangladesh has already produced examples of technical ambition that deserved stronger institutional follow-through.

One example comes from Shahjalal University of Science and Technology, where Dr. Muhammed Zafar Iqbal and his colleagues showed the scientific imagination Bangladesh needs. When he and his young colleagues built a cluster computer from scratch, the government did not adopt it as part of a national computing strategy. When he led work on Bangla computing, the government did not continue building on it. A large number of Bangladeshi educational institutions still use a paper-based admissions application process, while he and his colleagues invented a paperless system decades ago.

That is the pattern: talent appears, prototypes appear, ideas appear, and institutions fail to compound them. A functioning state would connect such moments to schools, laboratories, procurement, industry, and national research priorities. Too often, Bangladesh celebrates individuals but fails to build systems around them.

It is encouraging when public institutions express interest in frontier technology.

But the main challenge of our ICT sector is not a lack of slogans. It is governance. Reports about Aspire to Innovate, or A2I, warrant caution. In 2024, The Daily Star reported that the ICT Division asked UNDP to investigate allegations involving A2I consultants. The Business Standard reported that an Anti-Corruption Commission team inspected the ICT Tower as part of an inquiry into corruption allegations involving A2I.

So my advice to young Bangladeshi professionals is: be interested, but do not be captured. Work with government when useful, but keep at least one foot in another boat: a university lab, an open-source community, a private company, a foreign collaboration, or an independent technical network. Do not build your whole future around ministries, projects, and patronage chains.

The coming computing empires may be American and Chinese. But informed clients are not powerless clients. They negotiate better, choose better, build local capacity, and avoid being dazzled by every new machine.

That is why a DOE questionnaire about quantum computing should matter even in Bangladesh. It is about how future scientific authority will be organized, and whether we will understand it well enough to participate.

Omar Shehab is a theoretical quantum computer scientist. He is currently bootstrapping a stealth startup.