{"id":1240,"date":"2026-11-16T08:27:58","date_gmt":"2026-11-16T07:27:58","guid":{"rendered":"https:\/\/hadamard.com\/c\/ibm-nighthawk-r3-520-qubits-explained\/"},"modified":"2026-11-16T08:27:58","modified_gmt":"2026-11-16T07:27:58","slug":"ibm-nighthawk-r3-520-qubits-explained","status":"publish","type":"post","link":"https:\/\/hadamard.com\/c\/ibm-nighthawk-r3-520-qubits-explained\/","title":{"rendered":"IBM Nighthawk r3 Explained: 520 Qubits, 1,040-Qubit System in 2027"},"content":{"rendered":"<p>IBM Nighthawk r3 is a planned IBM quantum processor with 520 programmable qubits, designed to run 10,000 quantum operations. IBM also plans to connect two of them into a single 1,040-qubit system in 2027. You can&#8217;t use r3 yet. The chip clients run today is Nighthawk r2, a 120-qubit processor that IBM says reaches up to 100,000 samples per second.<\/p>\n<p>This article covers what IBM announced on November 11, 2026, how Nighthawk went from r1 to r3 in about a year, what the qubit counts mean, and where Nighthawk stops and IBM&#8217;s fault-tolerant roadmap toward Starling in 2029 takes over. For how IBM physically links chips together, see our separate explainer on <a href=\"https:\/\/hadamard.com\/c\/ibm-l-coupler-explained-quantum-chips\/\">IBM&#8217;s l-coupler and modular chip linking<\/a>.<\/p>\n<h2>What is IBM Nighthawk r3? The November 11, 2026 announcement<\/h2>\n<p>IBM&#8217;s <a href=\"https:\/\/newsroom.ibm.com\/2026-11-11-ibm-quantum-nighthawk-demonstrates-new-levels-of-speed-and-quality,-with-plans-to-scale-beyond-1,000-qubits-in-2027\" target=\"_blank\" rel=\"noopener\">November 11, 2026 press release<\/a> carries the dateline Yorktown Heights, NY. It describes Nighthawk r3 as a processor that &#8220;will be built to accurately execute 10,000 quantum operations across 520 qubits.&#8221; In more detail, it is &#8220;designed to support 10,000 quantum operations on 520 programmable qubits \u2026 with 2,034 total physical qubits.&#8221;<\/p>\n<p>The release gives three more points about r3:<\/p>\n<ul>\n<li>It keeps the square-lattice connectivity used by earlier Nighthawk chips.<\/li>\n<li>It adds &#8220;10 chip-to-chip interconnects, called l-couplers.&#8221;<\/li>\n<li>&#8220;In 2027, IBM plans to connect two Nighthawk r3 processors into a single 1,040-qubit system.&#8221;<\/li>\n<\/ul>\n<p>The release presents r3 as a future processor and gives no firm availability date for it. A photo caption says r3 is designed for 10,000 operations &#8220;in 2027.&#8221; Jay Gambetta, Director of IBM Research and an IBM Fellow, summed up the focus: &#8220;Useful quantum computing hinges on whether results can be delivered with accuracy and speed, both are fundamental.&#8221;<\/p>\n<h2>Nighthawk r1, r2, r3: how the chip changed in a year<\/h2>\n<p>IBM <a href=\"https:\/\/newsroom.ibm.com\/2025-11-12-ibm-delivers-new-quantum-processors,-software,-and-algorithm-breakthroughs-on-path-to-advantage-and-fault-tolerance\" target=\"_blank\" rel=\"noopener\">announced the original Nighthawk on November 12, 2025<\/a>. It had 120 qubits &#8220;linked together with 218 next-generation tunable couplers&#8221; in a square lattice, with each qubit connected to its four nearest neighbours. IBM said this gave &#8220;30 percent more complexity&#8221; than its previous processor and supported up to 5,000 two-qubit gates. At the time, IBM projected 7,500 gates by the end of 2026 and 10,000 in 2027.<\/p>\n<p>Nighthawk r2, described in an <a href=\"https:\/\/www.ibm.com\/quantum\/blog\/nighthawk-r2\" target=\"_blank\" rel=\"noopener\">IBM Quantum blog post on August 31, 2026<\/a>, stayed at r1&#8217;s scale. It reached accurate observable estimation on circuits with more than 7,500 gates using Probabilistic Error Amplification (PEA). r3 is the first version to grow the chip itself.<\/p>\n<table>\n<thead>\n<tr>\n<th>Version<\/th>\n<th>Status<\/th>\n<th>Programmable qubits<\/th>\n<th>Operations \/ gates<\/th>\n<th>Notable change<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>Nighthawk r1<\/td>\n<td>Announced Nov 12, 2025<\/td>\n<td>120<\/td>\n<td>Up to 5,000 two-qubit gates<\/td>\n<td>Square lattice, 218 tunable couplers<\/td>\n<\/tr>\n<tr>\n<td>Nighthawk r2<\/td>\n<td>Available on IBM Quantum Platform<\/td>\n<td>120<\/td>\n<td>More than 7,500<\/td>\n<td>120 reset elements, 100,000 circuits per second<\/td>\n<\/tr>\n<tr>\n<td>Nighthawk r3<\/td>\n<td>Announced Nov 11, 2026; not yet available<\/td>\n<td>520<\/td>\n<td>Designed for 10,000<\/td>\n<td>10 l-couplers; two chips planned as one 1,040-qubit system in 2027<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>Nighthawk r2 is open to users through IBM Quantum Platform, free accounts included. The blog gives its throughput as &#8220;100,000 circuits per second,&#8221; about 25x Heron&#8217;s roughly 4,000. Idle time between runs can drop to about one microsecond, down from hundreds of microseconds on earlier processors.<\/p>\n<h2>520 programmable qubits vs 2,034 physical qubits: reading the numbers<\/h2>\n<p>The two r3 numbers count different things. The 520 programmable qubits are the ones a user&#8217;s circuit actually computes on. The 2,034 &#8220;total physical qubits&#8221; is IBM&#8217;s own term for a larger count of hardware elements on the chip.<\/p>\n<p>IBM&#8217;s r2 blog shows how it counts these. Nighthawk r2 has 120 programmable qubits, 218 dedicated couplers and 120 independent reset elements, one per qubit, for 458 &#8220;physical quantum elements&#8221; in total. So for r2, the larger figure includes couplers and reset hardware, not only qubits that run computations.<\/p>\n<p>The r3 release doesn&#8217;t break down the 2,034 figure, so don&#8217;t assume r3 has the same mix as r2. The safe reading: when comparing processors, the 520 programmable qubits is the number that tells you how large a circuit r3 can run.<\/p>\n<p>There&#8217;s a second trap. Nighthawk qubits are physical qubits, and IBM handles their errors with error mitigation, not error correction. Starling&#8217;s planned 200 qubits are logical, meaning error-corrected, qubits. A 520 next to a 200 doesn&#8217;t mean Nighthawk is the more capable machine. The two numbers measure different things.<\/p>\n<h2>From one chip to 1,040 qubits: l-couplers and 99.3% transfer fidelity<\/h2>\n<p>The 1,040-qubit system depends on linking chips, not on building one bigger die. According to the <a href=\"https:\/\/newsroom.ibm.com\/2026-11-11-ibm-quantum-nighthawk-demonstrates-new-levels-of-speed-and-quality,-with-plans-to-scale-beyond-1,000-qubits-in-2027\" target=\"_blank\" rel=\"noopener\">November 2026 release<\/a>, IBM connected Nighthawk r1 processors using four l-couplers, &#8220;achieving 99.3 percent quantum state transfer fidelity and progressing toward its 99.9 percent target.&#8221;<\/p>\n<p>IBM also reported connecting two cryogenic modules into one system for the first time, a result it announced on August 19. The combined system reached operating temperature in under five days and cooled below 15 millikelvin.<\/p>\n<p>IBM&#8217;s <a href=\"https:\/\/www.ibm.com\/roadmaps\/quantum\" target=\"_blank\" rel=\"noopener\">public roadmap<\/a> shows where this is going. For 2026, it says Nighthawk &#8220;will run circuits with 7,500 gates in up to three 120-qubit modules.&#8221; For 2028, it lists Nighthawk at up to 15,000 gates on up to 1,080 qubits. The 2025 announcement described the 2028 goal as up to 15,000 two-qubit gates on 1,000 or more connected qubits using long-range couplers. The 1,040-qubit system planned for 2027 sits between those two steps.<\/p>\n<h2>What clients ran on Nighthawk r2, and what &#8220;speedup&#8221; means<\/h2>\n<p>The November release reports that Nighthawk r2 reaches up to 100,000 samples per second and more than 7,500 quantum operations on 120 qubits. Clients and partners, including E.ON, Algorithmiq and Q-CTRL, have run 34 billion circuits on it. IBM reports speedups of up to 19x over its previous processors.<\/p>\n<h3>The reported results<\/h3>\n<ul>\n<li><strong>E.ON:<\/strong> a 15x speedup in quantum machine learning for energy-grid optimization.<\/li>\n<li><strong>Oak Ridge National Laboratory (ORNL):<\/strong> a magnetic-materials simulation that ran 12x faster.<\/li>\n<li><strong>Algorithmiq:<\/strong> an experiment on a heterogeneous quantum material that ran 8x faster than the original Heron run. The release estimates that leading classical tensor-network methods would need about 240 times more memory for comparable accuracy.<\/li>\n<li><strong>Quantum Elements and ParityQC<\/strong> also report gains on r2.<\/li>\n<\/ul>\n<p>Hybrid quantum-classical workflows sped up too. IBM and RIKEN ran their sample-based quantum diagonalization (SQD) workflow about 5x faster while doing 100x more computations than in the original study. Cleveland Clinic and RIKEN processed 608 million samples, 33x faster than without workflow automation. On November 5, IBM and ORNL launched a quantum error correction co-design platform with AMD. It places Nighthawk next to ORNL&#8217;s Umbra HPC system, which uses AMD Instinct MI355X GPUs and 5th Gen AMD EPYC CPUs.<\/p>\n<h3>How to read these numbers<\/h3>\n<p>The E.ON and ORNL figures compare against earlier work on IBM&#8217;s Heron processor, and the Algorithmiq figure compares against its original Heron run. They show how much faster the same kind of job runs on IBM&#8217;s newer hardware. They don&#8217;t show a quantum computer beating a classical one. The Algorithmiq memory figure is the only classical comparison in this list, and the release presents it as an estimate.<\/p>\n<h3>Has IBM reached quantum advantage?<\/h3>\n<p>In November 2025, IBM said it was aiming for &#8220;quantum advantage by the end of 2026.&#8221; The quantum advantage experiments IBM and its partners published in July were first run on Heron. Their circuits and results are open for independent testing through the community-led Quantum Advantage Tracker. That openness matters because advantage claims depend on whether classical methods can catch up. Treat the claims as submitted for testing, not confirmed.<\/p>\n<h2>Where Nighthawk ends and IBM Starling&#8217;s 2029 fault tolerance begins<\/h2>\n<p>IBM runs two hardware tracks. Nighthawk is the near-term track: physical qubits plus error mitigation, aimed at useful results and quantum advantage. A separate series of processors is building toward error-corrected computing, and that is the path to Starling.<\/p>\n<ol>\n<li><strong>Loon (2025):<\/strong> an experimental processor that, <a href=\"https:\/\/newsroom.ibm.com\/2025-11-12-ibm-delivers-new-quantum-processors,-software,-and-algorithm-breakthroughs-on-path-to-advantage-and-fault-tolerance\" target=\"_blank\" rel=\"noopener\">according to IBM<\/a>, &#8220;demonstrated all the key processor components needed for fault-tolerant quantum computing,&#8221; including multiple routing layers and qubit-reset technology. It tests qLDPC components such as c-couplers. IBM also reported real-time qLDPC decoding in &#8220;less than 480 nanoseconds,&#8221; a year ahead of schedule.<\/li>\n<li><strong>Kookaburra (2026):<\/strong> described as &#8220;IBM&#8217;s first modular processor designed to store and process encoded information.&#8221; The roadmap says it will show a logical processing unit plus quantum memory in one module.<\/li>\n<li><strong>Cockatoo (2027):<\/strong> links two Kookaburra modules with l-couplers.<\/li>\n<li><strong>Starling (2029):<\/strong> &#8220;The first fault-tolerant quantum computer,&#8221; which IBM says &#8220;will be available in 2029.&#8221; It is planned for Poughkeepsie, NY, and is meant to run &#8220;100 million quantum operations using 200 logical qubits,&#8221; as reported by <a href=\"https:\/\/thequantuminsider.com\/2025\/06\/10\/ibm-offers-roadmap-toward-large-scale-fault-tolerant-quantum-computer\/\" target=\"_blank\" rel=\"noopener\">The Quantum Insider<\/a>.<\/li>\n<li><strong>Blue Jay (2033+):<\/strong> planned for &#8220;1 billion quantum operations over 2,000 logical qubits.&#8221;<\/li>\n<\/ol>\n<p>IBM says its qLDPC approach cuts error-correction overhead by &#8220;approximately 90 percent&#8221; compared with other leading codes. Fabrication has also moved to a 300mm wafer facility at NY Creates&#8217; Albany NanoTech Complex. For a sense of how logical-qubit quality gets measured across the industry, see our piece on <a href=\"https:\/\/hadamard.com\/c\/quantinuum-helios-logical-fidelity-record\/\">what Quantinuum&#8217;s logical fidelity record measures<\/a>.<\/p>\n<h2>What Nighthawk r3 does not mean<\/h2>\n<p>A few clear limits help put the announcement in proportion:<\/p>\n<ul>\n<li><strong>r3 is not available.<\/strong> It was announced as a future processor. The 1,040-qubit two-chip system is a plan for 2027.<\/li>\n<li><strong>It is not fault-tolerant.<\/strong> Nighthawk relies on error mitigation. IBM&#8217;s own timeline puts its first fault-tolerant machine, Starling, in 2029.<\/li>\n<li><strong>Quantum advantage is not confirmed.<\/strong> The July experiments are open for independent testing through the Quantum Advantage Tracker.<\/li>\n<li><strong>The speedups are relative.<\/strong> They compare against IBM&#8217;s earlier hardware, mostly Heron.<\/li>\n<\/ul>\n<p>None of this is about breaking today&#8217;s public-key cryptography. That would need large-scale, error-corrected machines. Our breakdown of <a href=\"https:\/\/hadamard.com\/c\/how-many-qubits-break-rsa-2048\/\">how many qubits it would take to break RSA-2048<\/a> covers the published estimates. Still, roadmaps like IBM&#8217;s are why systems that rely on ECDSA or RSA are planning their migration now. Hadamard builds <a href=\"https:\/\/bb84.com\/\">BB84<\/a>, an Ethereum Layer 2 that uses hash-based one-time addresses instead of ECDSA keys, as one approach to that migration.<\/p>\n<h2>FAQ<\/h2>\n<h3>When will IBM&#8217;s Nighthawk r3 be available, and is the 1,040-qubit system available now?<\/h3>\n<p>Neither is available. IBM&#8217;s November 11, 2026 release gives no firm availability date for r3, and a photo caption says it is designed for 10,000 operations &#8220;in 2027.&#8221; IBM plans to connect two r3 processors into a 1,040-qubit system in 2027. Today, users can run Nighthawk r2 on IBM Quantum Platform.<\/p>\n<h3>Why does Nighthawk r3 have 520 qubits but 2,034 physical qubits?<\/h3>\n<p>The 520 are programmable qubits that circuits run on. The 2,034 &#8220;total physical qubits&#8221; is IBM&#8217;s wording for a broader hardware count. For r2, IBM&#8217;s 458-element total includes 120 qubits, 218 couplers and 120 reset elements. IBM hasn&#8217;t published an equivalent breakdown for r3.<\/p>\n<h3>Has IBM achieved quantum advantage in 2026?<\/h3>\n<p>IBM has aimed for quantum advantage by the end of 2026. The advantage experiments it published with partners in July are open for independent testing through the community-led Quantum Advantage Tracker. They should be treated as claims under review, not confirmed results.<\/p>\n<h3>Is Nighthawk r3 a fault-tolerant quantum computer, and could it break RSA?<\/h3>\n<p>No. Nighthawk uses physical qubits with error mitigation, not error-corrected logical qubits. IBM&#8217;s first planned fault-tolerant system is Starling in 2029, with 200 logical qubits. Breaking RSA would need large-scale error-corrected hardware, which r3 is not.<\/p>\n<h2>Sources<\/h2>\n<ul>\n<li><a href=\"https:\/\/newsroom.ibm.com\/2026-11-11-ibm-quantum-nighthawk-demonstrates-new-levels-of-speed-and-quality,-with-plans-to-scale-beyond-1,000-qubits-in-2027\" target=\"_blank\" rel=\"noopener\">IBM Quantum Nighthawk demonstrates new levels of speed and quality, with plans to scale beyond 1,000 qubits in 2027<\/a> (IBM Newsroom, November 11, 2026)<\/li>\n<li><a href=\"https:\/\/www.ibm.com\/quantum\/blog\/nighthawk-r2\" target=\"_blank\" rel=\"noopener\">Nighthawk r2 announcement<\/a> (IBM Quantum blog, August 31, 2026)<\/li>\n<li><a href=\"https:\/\/newsroom.ibm.com\/2025-11-12-ibm-delivers-new-quantum-processors,-software,-and-algorithm-breakthroughs-on-path-to-advantage-and-fault-tolerance\" target=\"_blank\" rel=\"noopener\">IBM delivers new quantum processors, software, and algorithm breakthroughs on path to advantage and fault tolerance<\/a> (IBM Newsroom, November 12, 2025)<\/li>\n<li><a href=\"https:\/\/www.ibm.com\/roadmaps\/quantum\" target=\"_blank\" rel=\"noopener\">IBM Quantum Roadmap<\/a> (IBM, accessed November 16, 2026)<\/li>\n<li><a href=\"https:\/\/thequantuminsider.com\/2025\/06\/10\/ibm-offers-roadmap-toward-large-scale-fault-tolerant-quantum-computer\/\" target=\"_blank\" rel=\"noopener\">IBM offers roadmap toward large-scale fault-tolerant quantum computer<\/a> (The Quantum Insider, June 10, 2025)<\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>IBM Nighthawk r3 is a planned 520-qubit processor built for 10,000 quantum operations, and IBM wants to join two of them into a 1,040-qubit system in 2027. Here is what has shipped, what is still a plan, and how far all of it is from fault tolerance.<\/p>\n","protected":false},"author":1,"featured_media":1241,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"jetpack_post_was_ever_published":false,"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[791],"tags":[829,874,3,846,875],"class_list":["post-1240","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-quantum-computing","tag-fault-tolerance","tag-ibm-quantum","tag-quantum-computing","tag-quantum-hardware","tag-quantum-roadmap","entry"],"jetpack_featured_media_url":"https:\/\/hadamard.com\/c\/wp-content\/uploads\/2026\/11\/2026-11-16-ibm-nighthawk-r3-520-qubits-explained.png","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/hadamard.com\/c\/wp-json\/wp\/v2\/posts\/1240","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/hadamard.com\/c\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/hadamard.com\/c\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/hadamard.com\/c\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/hadamard.com\/c\/wp-json\/wp\/v2\/comments?post=1240"}],"version-history":[{"count":0,"href":"https:\/\/hadamard.com\/c\/wp-json\/wp\/v2\/posts\/1240\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/hadamard.com\/c\/wp-json\/wp\/v2\/media\/1241"}],"wp:attachment":[{"href":"https:\/\/hadamard.com\/c\/wp-json\/wp\/v2\/media?parent=1240"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/hadamard.com\/c\/wp-json\/wp\/v2\/categories?post=1240"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/hadamard.com\/c\/wp-json\/wp\/v2\/tags?post=1240"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}