@HPC Podcast Archives - OrionX.net - HPC News Bytes – 20260223

Episode Date: February 23, 2026

- AI power needs - Small Modular Reactors (SMR) - US military, DOE, airlift small reactor - Quantum "teleportation" - How does Quantum Communication work? - Silicon photonics for quantum computing [audio mp3="https://orionx.net/wp-content/uploads/2026/02/HPCNB_20260223.mp3"][/audio] The post HPC News Bytes – 20260223 appeared first on OrionX.net.

Welcome to HPC Newsbytes, a weekly show about important news in the world of supercomputing, AI, and other advanced technologies. Hi, everyone. Welcome to HBC Newsbytes. I'm Doug Black of Inside HBC, and with me is Shaheen Khan of OrionX.net. Previous episodes of this podcast have discussed the problem of demand for electrical power outstripping supply. Well, it's still going on, only more so.

This is leading to new measures to increase power supply, given a recent figure from the Wall Street Journal, that $700 billion will be spent in capital expenditures by tech giants this year, a cash pile that has already begun to generate creativity in the power sector. There was news last week that the U.S. military and the Department of Energy conducted a project to airlift a five-migawatt micro-nuclear reactor without nuclear or fuel, I should quickly add, nearly 700 miles from California to Utah to show how they can rapidly deploy significant sources of energy for defense and civilian needs. The idea was to avoid assembling the reactor on site. That's part of a broader push in the U.S. and globally to commercialize small modular reactors or SMRs, which are compact nuclear units without the huge size and cost of traditional nuclear power plants. So this is an addition to all the work going on in large traditional nuclear reactors as we've seen hyperscalers pursue. According to various reports,

more than 80 SMR designs are being developed in 18 countries, and China and Russia have the only currently operational SMRs so far. The U.S., Canada, Japan, South Korea, and several European in countries have active R&D pipelines backed by government funding. The issues, as you might expect, are economics, safety, and meeting government requirements. So that means compliance and licensing. Forecasts vary, but many analysts see a multi-billion dollar global SMR market by 2035. We first covered this topic three or four years ago. And at that time, the power ratings were in the 20 megawatt to 500 megawatt range, and that was targeting 2030.

Those projections remain the same, so data centers will need several, if not many, since by 2030, a single rack could be 600 kilowatts and possibly approaching one megawatt or more. Some estimates see SMRs providing six gigawatts of capacity in the U.S. by 2035. Globally, over $20 billion in combined public plus private funding, has been invested in SMR development in recent years, and that is expected to grow. The U.S. leads in the number of SMR projects and the research funding that is applied to it.

Some estimates have it at 30% of global funding to give you a ballpark, with federal programs backing multiple designs. China is also a strong contender, deploying early units and funding several SMR types domestically. As you mentioned, Canada, South Korea, and Japan have major government industry efforts, and India recently committed to indigenous SMRs. As we all hear all the time, there is significant pent-up demand for electricity driven by AI data centers, so the market for SMRs could grow rapidly and substantially over the next decade if only the products were available. In case you're curious, the reactor was moved by a Boeing C-17 military transport plane, which is a heavy-lift tactical aircraft, which means it can participate in the reactor.

in combat and take off and land on short runways. It can lift about 170,000 pounds. The much bigger plane you might be thinking of is the Lockheed C-5, which they call a flying warehouse, whose latest model can carry 280,000 pounds. Wow. Okay, fans of Star Trek light up at the mention of the word teleportation. Now several companies have simultaneously announced that they are the first to successfully test quantum teleportation. These include Telecom giant Telos and Quantum Company Photonic, announcing they've demonstrated quantum teleportation over 30 kilometers of active metropolitan fiber, as did Deutsche Telecom and Kinect, a quantum networking company. Kinect also joined with Cisco on this last week to demonstrate high-speed, high-fidelity entanglement swapping on their Gotham Kew network in New York City.

Shaheen, this all sounds significant and naturally I'm tempted to somehow work on the hackneyed phrase, beam me up, but please help us understand what teleportation means in this context. While physicists understand what they mean by the word teleportation, it is pretty misleading to use it in mass communication. The short answer is, quote, teleporting means instant, and in this case, only the correlation between quantum states is instant. So let's break up quantum communication into four parts. First, you have to establish an entangle the state between two particles in location A and location B. Entanglement between far away particles can happen in nature. But if you want to control where it happens, then you start with, say, a photon that you can split into two entangled photons

and then physically send them to location A and B through a fiber optic cable. You are interested in sending the quantum state of the particle. The particle itself is just a vehicle. If the distance is too far, the state can dissipate. So in that case, you need intermediate steps, where the state of the particle is transferred to a new particle before it continues the journey. These intermediate steps are called repeaters in telecommunications,

but in this case, they have to carry the state. Now, quantum states cannot be copied or supplicated. That's part of what makes them secure, but they can be swapped, and that's the state swapping that was mentioned in the news. They need to be removed from one particle and installed on another, so to say. So you need to map out the path from A to B and all the intermediate repeaters. All of this is limited by speed of light. Step two is confirmation that the state successfully arrived in location B. That is usually through a notification process, which is also limited.

by speed of light. Step 3 is measurement. Location A performs a measurement and the state at location B correlates immediately. This is the only part that is instant. But knowing the correlated state is useless by itself, it is correlated, but location B doesn't know how it is correlated, and it needs additional information from location A to interpret it. Step 4 is interpretation. Location A sends its results to location B, again, limited by speed of light. What location A sends is useless to everyone else, but it is the key that location B needs to interpret its measurement, to unscramble the state.

So they use the above for quantum key distribution, KQD. That way, only a shared encryption key is sent through this process, and the actual data is encrypted and decrypted by this key, which is guaranteed to be secure. For quantum internet, the idea is that you transfer quantum states around to quantum computers which operate on them and everything stays in a quantum state until the results are available to its intended user. So that goes beyond confidential computing towards what's called blind computing.

But fully homomorphic encryption, F-H-E, can also accomplish that with classical computers. So quantum internet seems to be only useful when you have quantum advantage or for scale-out types of quantum computing uses. By the way, going from location A to location B is what everyone calls Alice and Bob, but I just stuck to location A and B. Very good. Continuing on the quantum front tower semiconductor and Zanadu have expanded their partnership to develop a manufacturable silicon photonics platform for fault-tolerant. quantum computing. This collaboration utilizes Towers' high-volume foundry infrastructure to industrialize Zanidu's photonic circuit designs. And the companies say they are transitioning hardware from prototype to demonstrate your systems. The joint engineering effort focuses on a

custom production flow for a specialized material stack designed to maintain optical performance and scalability as system complexity increases. Yes, and we should recall that back in 2021, Global Foundries and Photonics quantum computing company, Cy Quantum, formed a partnership to manufacture silicon photonic quantum chips in New York State. Their goal is to reach a full-scale, 1 million-plus-cubit commercial quantum computer

by utilizing existing manufacturing techniques rather than creating a new, separate foundry process. Also notable is that silicon photonics for quantum computing does not need the leading-edge fabrication nodes, which opens the market to other manufacturers and not just the usual TSM-S Samsung Intel Rapidis, who are all pursuing sub-3 nanometer processes. This came up in our conversation with Professor Karen Bergman of Columbia University. I highly encourage you to take a look at episodes 54 and 104 of the FHBC podcast, when she was a special guest of the podcast.

As of early 2025, the companies were reportedly producing, quote, millions of photonic chips, unquote, using Global Foundry's 45 nanometer process. And as of October 2025, a full-fledged prototype was being assembled in California, and the company was expecting to reach a one million cubic system by roughly 2027. So as usual, lots of really good progress and lots to be done yet. All right, that's it for this episode. Thank you all for being with us. HPC Newsbytes is a production of OrionX in association with InsideHPC.

Shaheen Khan and Doug Black host the show. Every episode is featured on Insidehpc.com and posted on OrionX.net. Thank you for listening.