The Great Superconducting Qubit Consolidation
Google acquires Atlantic Quantum, D-Wave acquires Quantum Circuits Inc. What this tells you about the first wave of quantum hardware M&A, winners in the quantum computing race, and what it doesn't.
“Who is left in the startup world building superconducting qubits?” is a question investors seem to ask about the superconducting qubit modality, with a tone that implies that superconducting is dead, that neutral atoms and trapped ions have won, and that the M&A is the tombstone.
I don’t think that’s what the data says. A more illuminating question is: why is superconducting the modality consolidating first, when neutral atoms and ions are still raising, and even new startups spinning out in these modalities?
In October 2025, the Nobel Prize in Physics went to John Martinis, Michel Devoret, and John Clarke, the three physicists whose 1984 and 1985 experiments demonstrated macroscopic quantum tunneling in electrical circuits and laid the foundation for superconducting qubit quantum computers.
A few days before, Google had announced that Atlantic Quantum, the MIT spin-out from Will Oliver's Engineering Quantum Systems group, whose fluxonium-transmon-coupler architecture set records for superconducting single- and two-qubit gate fidelities, was joining Google Quantum AI (Google Quantum AI, Oct 2, 2025).
Three months later, in January 2026, D-Wave acquired Quantum Circuits Inc. (D-Wave press release), the Yale spin-out co-founded by Nobel laureate Michel Devoret and Rob Schoelkopf, inventors of both the transmon and dual-rail qubit, for $550 million (note: Devoret was already at Google Quantum AI by the time of the QCI acquisition).
What actually happened
Let’s be precise about the two deals, because the details matter.
Atlantic Quantum was acquired by Google Quantum AI . Atlantic was founded in 2022, spun out of MIT’s Engineering Quantum Systems Group under Will Oliver, with a team drawn from MIT and Chalmers. Their bet was on fluxonium qubits — a superconducting architecture with much higher anharmonicity than the transmon. Anharmonicity means the qubit’s energy levels are unevenly spaced, so the two you use to compute are cleanly separated from the ones you don’t. Higher anharmonicity means you can drive the qubit faster without accidentally exciting the wrong levels, so, in principle, faster gates with less leakage error. The trade-off is complexity: fluxonium is harder to fabricate than the transmon. Atlantic Quantum was one of 18 companies selected for DARPA’s Quantum Benchmarking Initiative earlier in 2025. On the acquisition, Google did not disclose terms, but rumors are approximately the pricing of the last funding round. The team integrates into Google Quantum AI under Hartmut Neven, with the stated goal of scaling Google’s roadmap from ~10² to 10⁶ physical qubits.
Quantum Circuits Inc. acquired by D-Wave . QCI was founded in 2015 in New Haven, out of Yale, by Rob Schoelkopf, Michel Devoret, and Luigi Frunzio. Their approach centers on dual-rail qubits, which encode a logical qubit across two physical modes in a way that turns the dominant error (photon loss) into an error where you know when it happens and where. Flagged errors are much easier to correct than unknown ones, which in theory means fewer physical qubits per logical qubit. D-Wave paid $550 million: $300M in D-Wave common stock and $250M in cash. The deal closed January 20 (D-Wave, Jan 20, 2026). The QCI team stays in New Haven as a new R&D center; D-Wave now markets itself as the only company with annealing and gate-model superconducting platforms under one roof.
There’s a detail in the QCI deal that I think is an interesting fact in the entire story: Michel Devoret has been Chief Scientist for Quantum Hardware at Google Quantum AI since 2023. He retired from Yale's full-time ladder faculty in mid-2024 (taking emeritus status). By the time D-Wave bought QCI, one of QCI’s founding scientists had been inside the main superconducting competitor for roughly two and a half years.
The D-Wave move: two reads
Spend five minutes on any D-Wave earnings call from the last two years and you will hear CEO Alan Baratz explain — with increasing confidence as each quarter passed — that annealing is the only quantum approach delivering commercial value today.
While in 2023–early 2024, Baratz described annealing and gate-model as complementary, as different tools for different problem classes, and that D-Wave was building both (Inside Quantum Technology, 2024), things changed.
On the Q2 2024 call, he told analysts that “the world is waking up to the fact that annealing quantum computing is here now” and that D-Wave was “single-handedly creating the market for commercial quantum computing.” On the Q2 2025 call, he called characterizing annealing as “niche” “misleading and ill-informed” and said DARPA’s gate-model focus was “a huge mistake on the part of DARPA and the US government.” On the Q4 2025 call, he was even more direct: “the annealing quantum systems are the only ones that can actually deliver any commercial value today. They are the only ones that are used in production by customers today.”
Then, on January 7, 2026, D-Wave announced the $550M Quantum Circuits Inc. deal.
QBTS was trading around $30 at the announcement, implying a market cap near $10.5B.
There are two ways to read this. The charitable read is that D-Wave is returning to its earlier position — annealing and gate-model are complementary, and now that the technology is maturing, owning both is strategically rational.
The other is that the annealing-only narrative was getting harder to sustain in a market where Google, IBM, and multiple gate-model public companies were setting the benchmarks investors cared about, and IQM switched from qASIC to fault-tolerant messaging. Adding a gate-model platform, and doing it through the acquisition of a team with deep Yale credentials, repositions D-Wave as a dual-platform player, similar to what Google is doing by investing in neutral atoms alongside superconducting qubits (Google blog; Colorado Sun).
What about Atlantic Quantum?
Atlantic was more conservative with headline numbers. Their focus was on fluxonium as a physical qubit architecture and on integrated cryogenic control — the idea that you move control electronics onto the cold stage to reduce wiring overhead as you scale (The Quantum Insider, Oct 3, 2025). Their public demonstrations emphasized single- and two-qubit fidelities on small devices rather than assembled multi-qubit processors; Preskill at Q2B cited two-qubit gates with better than three-nines fidelity on MIT's fluxonium work, and the Oliver group has peer-reviewed results at 99.92% two-qubit gate fidelity (Ding et al., PRX 2023) and a world-record 99.998% single-qubit gate fidelity (Rower et al., PRX Quantum 2024; MIT News, Jan 14, 2025).
What seems to have made Atlantic attractive to Google wasn't a specific processor demonstration.
First, fluxonium has strong small-scale evidence of outperforming transmon, and if that advantage survives scaling — which is unproven in either direction and was exactly the question Preskill flagged at Q2B — then whoever owns the IP owns a next-generation physical qubit. Google has the fab access to actually try.
Second, the integrated cold-stage control architecture directly addresses Google's stated scaling problem of getting from 10² to 10⁶ qubits without the wiring nightmare becoming unmanageable — a problem Google's own Nature Electronics paper in May 2025 flagged as one of the major remaining obstacles to fault tolerance and their collaborations with industry on cryogenics work have shown.
Third, the deal brings Google a tight pipeline into MIT and the broader Engineering Quantum Systems group. In a field where the global pool of frontier superconducting talent is small, a few thousand people at most, buying a team that is trained in one of the two or three best labs in the world is a long-duration benefit as the talent war heats up.
The DARPA signal
DARPA’s Quantum Benchmarking Initiative has three stages, and the superconducting picture depends on which stage you’re looking at.
Stage C (Validation and Co-Design): Microsoft and PsiQuantum announced February 2025 via the US2QC predecessor program (DARPA, Feb 6, 2025). Microsoft is pursuing a topological / Majorana qubit approach — technically superconducting, but a very different architecture from transmons. Worth flagging that the Microsoft Nature paper on Majorana 1 in February 2025 carries an editorial note from the reviewers stating the results do not constitute evidence for Majorana zero modes themselves (Aaronson, Shtetl-Optimized, Feb 2025). DARPA spent two years evaluating Microsoft’s program with a large team and selected them anyway, but the scientific community has not fully accepted the claim, and it’s worth separating “DARPA thinks this is worth verifying” from “this is verified physics.” We will probably see new announcements in the coming months on companies moving to stage C.
Stage B: 11 companies advanced on November 6, 2025 (DARPA Stage B announcement). The superconducting player is IBM (modular superconducting transmons), plus Nord Quantique (superconducting qubits with bosonic error correction). Everyone else on the Stage B list is neutral atom, trapped ion, silicon spin, or photonic: Atom Computing, Diraq, IonQ, Photonic Inc., Quantinuum, Quantum Motion, QuEra Computing, Silicon Quantum Computing, Xanadu.
Stage A: Google joined Stage A in September 2025 (Google Blog). Atlantic Quantum was in Stage A before being absorbed by Google, Oxford Ionics was also in Stage A before being absorbed by IonQ. Rigetti, Alice & Bob, and HPE were in Stage A but did not advance to Stage B in this cohort. DARPA has said staggered timelines mean more teams may advance.
On March 10, 2026, DARPA reopened Stage A with a new solicitation, explicitly inviting novel architectures that have not yet been evaluated (DARPA, Mar 10, 2026; abstracts due July 31, 2026). The program has new leadership under Micah Stoutimore and is actively looking for approaches it hasn't seen.
So is superconducting dead, or not?
Look at the actual global roster before answering. 45 companies and government-affiliated entities are currently building or have recently built superconducting quantum computers, across 15 countries on 6 continents. The US leads with 14. China has 10. Japan has 4. There’s a growing European cluster anchored in Finland, the UK, France, the Netherlands, and Spain. Roughly 70% of these companies use the transmon.
(Full interactive map is linked beyond the paywall)
Three patterns are worth naming up front before any of the other analysis:
First, the consolidation is US-specific. The 45-company roster grew. Peak Quantum (WMI/TUM Garching, 2024), QuamCore (Israel, founded 2022, emerged from stealth 2025), LogicBit (Zhejiang University spin-off, Pre-A October 2025), QpiAI (India, 25-qubit QpiAI-Indus launched April 2025 under NQM), SDT (Korea, 20-qubit Kreo), Anyon Technologies (Singapore/Emeryville), and Qolab (Martinis + Ho, 2022) are all active.
Second, China runs three superconducting tracks across different funding models and degree of control. The USTC/Zuchongzhi pipeline is the academic-anchored one, commercialized through China Telecom Quantum Group and QuantumCTek with heavy state backing. Origin Quantum runs a separate full-stack commercial chain out of Hefei at a ~$950M implied valuation. SpinQ sells internationally — 200+ institutions across 40+ countries, including the US, Japan, Australia, and the EU, and was the first Chinese company to export a complete superconducting system abroad. When Alibaba and Baidu shut down their quantum labs in 2023-24, the equipment went to BAQIS and Zhejiang University, but the talent and IP stayed in the ecosystem. It’s assumed they are still working on this, but quietly. In China, the 15th Five-Year Plan names quantum first among the designated "future industries."
Third, the new entrants are not competing on qubit count. If you actually read what Peak Quantum, QuamCore, Qolab, Anyon Technologies, and the others are pitching, almost none of them are trying to out-scale IBM or Google at the transmon game. They’re betting on a specific architectural or integration thesis.
Fourth, the enabling-technology supply chain is its own industry now. Bluefors (roughly 35-40% of the global dilution refrigerator supply). Maybell (US) and Zero Point Cryogenics (Canada) are coming for this market. Quantum Machines (control electronics). Qblox (European, expanding to US). Low Noise Factory (cryogenic HEMT amplifiers used by nearly every superconducting readout chain). The picks and shovels partnerships are becoming more important for scale and breakthroughs. It is also, for the most part, invisible to generalist tech investors.
Also, of note: IQM announced a $1.8B SPAC merger with Real Asset Acquisition Corp. on February 23, 2026. IQM has delivered 15 systems to 13 customers including four of the top ten supercomputing centers, pitching on-premises sovereign deployment rather than hyperscaler cloud access.
The real asymmetry: fab access
Superconducting qubits are chips, and chips require a fab. The easy take is that superconducting is the most capex-heavy modality and therefore the one most exposed when capital tightens (“hairdryer power”). But that’s not quite right. Other modalities have their own costs and own challenges — laser power and noise, how to deploy systems, losses, etc.
The real asymmetry is that serious volume production at industrial yield requires compatibility with 300mm CMOS tooling because that’s what modern semiconductor fabs are built around. You can make excellent superconducting qubits on 200mm (Lincoln Lab) or smaller, but you can’t hit the yield and throughput that a million-qubit system would need.
IBM made this explicit on November 12, 2025: all future chips on the IBM Quantum roadmap will be fabricated at NY CREATES’ Albany NanoTech Complex on 300mm wafers (IBM Quantum blog; IBM press release). Jay Gambetta said IBM is uniquely positioned across software, hardware, fabrication, and error correction. You can argue with “uniquely,” but the fab argument is the strongest part of that statement.
Semiconductor people I talk to are blunt about why this matters: TSMC is not interested in turning over a leading-edge production line to print one quantum chip at a time. They run a volume business. imec has done important work — a 2024 Nature paper demonstrating 300mm CMOS-fabricated transmons with >100 μs coherence — but imec is a research foundry that partners with companies; it’s not a production line a startup can rent at scale. For modalities like diamond NV centers in quantum sensing, the process complexity is different and the volume economics are plausible, which is part of why foundries have been more interested in that direction.
The practical consequence for superconducting startups: you can (a) raise enough to build or co-develop a fab (unrealistic on venture terms), (b) partner with an incumbent and accept constraints on what you ship, or (c) get acquired by someone with fab access.
When Jay Gambetta publicly called the surface code an “engineering pipe dream” at scale. IBM shifted to quantum LDPC codes — specifically the Bivariate Bicycle “Gross code” — which encodes 12 logical qubits in 144 physical qubits versus roughly 1,452-2,028 physical qubits for comparable surface code. Ten to fourteen times more efficient. The catch: qLDPC requires non-local connectivity, which requires the 300mm fab move. So when investors ask why IBM doesn’t publish logical-qubit demonstrations the way Quantinuum does, the answer is that IBM’s code only shows its advantage at scale, and the long-range couplers needed to run it are exactly what IBM’s hardware roadmap is building toward over 2026-2029. Additionally, IBM is working with AMD for running quantum errorr correction algorithms.
The right question isn’t “who has more logical qubits today.” It’s “which fault-tolerance pathway is credible at the scale the company is actually building toward, and does the hardware roadmap match the code?” IBM’s gamble is that qLDPC on 300mm fab wins at 200+ logical qubits. Google is hedging with both surface code and color code on Willow-class hardware and acquired Atlantic Quantum and is starting a neutral atom program (oh, to have infinite money). Trapped ions and neutral-atom companies are betting reconfigurability wins.
Does one modality win?
John Preskill, summarizing the state of the fluxonium-vs-transmon question at Q2B 2024, said the open question is whether “trading simplicity for performance in superconducting qubits will ultimately be advantageous for scaling to large systems is still unclear” (quoted in HPCwire).
I’d extend that beyond fluxonium and the transmon. The broader question across the whole sector is: does any one modality win, or do multiple architectures survive because they’re suited to different problem classes?
Three things to keep in mind when thinking about this:
First, “going first” in a given year doesn’t mean winning long-term. Are you looking at an investor timeline or a 100-year timeline? The ranking keeps moving. Any investor who reads the current ordering as permanent is the same investor who told you in 2019 that photonics would have 1,000,000 qubits that year.
Second, the metrics are weird and non-comparable. When you see a neat leaderboard that ranks modalities in the same column, someone has made simplifying assumptions that quietly favor one architecture. (And see my own biased and flawed neat table with simplifying assumptions beyond the paywall.)
Third, short-term winners and long-term winners can be different companies. Near-term commercial quantum advantage — the kind that moves a revenue line on an enterprise customer’s P&L — may come from whichever modality first ships a useful workload, or a company that makes components (“the picks and shovels” very very well), as some companies are leaning on. But fault-tolerant quantum computing is a separate race on a different time scale, and the architecture that wins the near-term revenue race may not be the one that delivers FTQC in 2033.
This is also a talent and scaling war
Zoom out. Here’s what’s actually happening across the quantum hardware industry over the last six months:
September 2025: IonQ acquires Oxford Ionics
October 2025: Google acquires Atlantic Quantum
November 2025: IBM commits all future chips to NY CREATES Albany 300mm fab
November 2025: DARPA selects 11 Stage B companies
January 2026: D-Wave acquires QCI
January 2026: Quantinuum files for IPO
January 2026: IonQ announced intent to acquire SkyWater (a foundry serving multiple quantum companies including D-Wave and PsiQuantum)
February 2026: IQM announces $1.8B SPAC
February 2026: Infleqtion starts trading, $1.8B SPAC
March 2026: Pasqal announces $2B SPAC
March 2026: Google hires Adam Kaufman from JILA/CU Boulder to lead a new neutral-atoms team in Colorado
March 2026: DARPA reopens QBI Stage A for novel architectures
March 2026: Xanadu starts trading, $3B SPAC
Four different things are happening.
Teams are being acquired — Atlantic goes to Google, QCI goes to D-Wave, Oxford Ionics goes to IonQ, Kaufman gets hired directly, Michel Devoret sits inside Google Quantum AI.
Fab and supply chain is being locked down — IBM at Albany, IonQ buying SkyWater, Google running Santa Barbara.
Public markets are pricing quantum hardware very differently than private markets
For comparison, recent confirmed private-market valuations:
Trapped ion — Quantinuum at $10B pre-money (September 2025, $600M raise, with new investors Quanta Computer, NVentures (NVIDIA’s venture capital arm), QED Investors as well as JPMorganChase, Mitsui, Amgen, Cambridge Quantum Holdings, Serendipity Capital and Honeywell, upsized to $800M when Fidelity joined an oversubscribed close)
Photonic — PsiQuantum at $7B post-money (September 2025, $1B Series E led by BlackRock, with Temasek and Baillie Gifford, and new investors including NVentures and Ribbit participating)
And government demand is being signaled — DARPA Stage B narrows the field while Stage A reopens, the Genesis Mission executive order lands, and the NQI Reauthorization Act advances.
Startups have to pick, but hyperscalers don’t. For hyperscalers, it’s to build moats across every dimension simultaneously (superposition, haha). Google now has: superconducting (native plus Atlantic), a neutral-atoms team (Kaufman plus ongoing QuEra collaboration), and Michel Devoret as Chief Scientist for Quantum Hardware. IBM has its own internal program plus the 300mm fab. D-Wave now has annealing and gate-model superconducting. IonQ is rolling up trapped ions and controlling fab supply. Microsoft is going deep with in-house Majorana and long-term Atom Computing partnership. Startups have to pick. Hyperscalers don’t.
There’s an analogy to the AI industry that’s worth sitting with. OpenAI worked on ChatGPT for years. When they launched in late 2022, many people assumed they had an insurmountable moat — the team, the infrastructure, the data advantage. Months later, Anthropic shipped Claude. Google shipped Gemini. Meta open-sourced Llama. We laughed at Google, but now we’re using Claude and Gemini day-to-day, not ChatGPT. The “moat” turned out to be a head start, and that head start was eroded not by a single competitor but by the normal circulation of researchers between labs. The people with the skills exist in a small global talent pool, and money can find them.
Quantum is at the same stage, with a smaller pool. The people who know how to build a superconducting qubit at the research frontier are numbered in the low thousands globally. The companies that now have cash to pay compensation and acquire those people are Google, IBM, Amazon, Microsoft, Nvidia, and a handful of public quantum-pure-plays with a lot of money to spend.
Thesis 1: Sovereign buyers, not VC, will price the next rounds
If you built a superconducting startup, likely the next funding round will come from governments.
Read the National Quantum Initiative Reauthorization Act (S.3597) — introduced January 8, 2026 by Senators Cantwell (D-WA) and Young (R-IN), advanced by the Senate Commerce Committee on April 14, 2026. Key provisions if passed: extends the National Quantum Initiative to December 2034; $85M/year for NIST QIST research; up to three new NIST quantum centers; NASA formally included for quantum satellite communications and sensing; a supply chain resilience plan required from the Secretary of Commerce within three years; manufacturing institute and prize challenges for near-term applications; “valley of death” commercialization language specifically aimed at moving lab breakthroughs into hardware.
Read that list alongside the DARPA QBI reopening (March 10, 2026) and the November 2025 Genesis Mission executive order (DOE-led AI-plus-quantum initiative linking national labs to quantum processors). It seems the US government intends to build, procure, and operate large quantum computing systems — including, over the longer horizon, machines capable of running Shor’s algorithm at cryptographically relevant scale.
Thesis 2: How many quantum computers can we actually build?
Once quantum advantage is hit, the demand for quantum computing for the rest of the decade is going to materially outstrip the supply of actual, working, installed systems. This is a near-term bidding war.
Count the supply side. How many superconducting systems with meaningful qubit counts physically exist today?
IBM has roughly 25 machines above 100 qubits, 9 of them client-located, 2 in data centers. Google has a small fleet of Willow-class devices at 105 qubits, with select research partners getting access through the Willow Early Access Program (Google blog, Dec 2024; SDxCentral, Mar 2026).
Fujitsu/RIKEN has a 256-qubit system running at the RIKEN RQC-Fujitsu Collaboration Center plus an additional system delivered to Japan’s AIST in March 2025, with a 1,000-qubit machine planned for 2026 at Fujitsu Technology Park.
D-Wave has dozens of annealing systems deployed across cloud and on-premises, including new Advantage2 installations at Davidson Technologies (Huntsville, AL) and an Advantage2 upgrade at Jülich Supercomputing Centre (Germany), with Leap cloud access across 40+ countries.
IQM has sold 21 systems to 13 customers — including four of the top ten supercomputing centers globally — and delivered 15 to date, one of the largest publicly disclosed number among frontier quantum companies.
Everyone else is operating below or just at the 100-qubit threshold:
Rigetti’s newest flagship is Cepheus-1-36Q, a 36-qubit chiplet system built from four tiled 9-qubit chips at 99.5% median two-qubit fidelity — a 2x error-rate reduction over their earlier 84-qubit Ankaa-3, but down to a smaller number of qubits. OQC’s Toshiko is 32 qubits, deployed in London, Tokyo, and Spain, with a New York deployment and the next-generation GENESIS in progress (OQC). Origin Quantum’s Wukong is 72 qubits, with three confirmed commercial deployments in China (postquantum.com), which admits the roadmap targeting 1,024 qubits by 2025 was not achieved. QuantWare is the highest-volume QPU supplier in the world, with customers in 20+ countries, but they ship smaller processors as components for others to integrate — their 64-qubit Tenor was delivered to the University of Naples Federico II in October 2025 as the first in a broader rollout, with a 10,000-qubit VIO-40K planned for 2028 (QuantWare/HPCwire, Oct 2025; QuantWare, Dec 2025).
Add it all up globally: low hundreds of usable systems with most below 100 qubits, and a much smaller number — perhaps 30-50 globally — above the 100-qubit threshold, deployed primarily at hyperscaler data centers, national labs, and a small number of enterprise research customers.
Count the supply chain constraints. To build a new superconducting system, you need a dilution refrigerator (Bluefors has had multi-quarter lead times, sometimes longer). You need cryogenic control electronics, turbo pumps, TWPAs, wires. You need the chip itself, which requires one of a dozen facilities globally that can fabricate quality superconducting qubits at research scale, and one of five or six that can do it at production scale. And you need a team of physicists who know how to bring it all up — which brings us back to the talent war. Companies can acquire or partner with companies to prep for the supply chain.
Count the demand side. BCG counts 150+ active enterprise PoCs in 2024, up 50% since 2022, with 80% of $5M+ adopters planning to expand (BCG, 2025 end-user survey, n=109). The quantum market is about $450M today — enterprise accounts for about half — and BCG’s 2029 scenarios put the market at $2.5B if the IBM public roadmap plays out (40% likelihood), $5B if “early advantage” arrives (30%), and $10B+ in the most aggressive scenario (<20%). Add DARPA, DOE, NIST, NASA, the UK National Quantum Computing Centre, European Quantum Flagship procurement, Chinese national procurement, Japanese MOONSHOT, the Indian National Quantum Mission, the Korean $2.3B strategy, the UAE TII program, the Saudi Norma contract. The number of vendors who can actually ship systems to sovereign buyers at the quality and support level they need is a very short list, and most of that demand is from buyers with deep pockets.
What this implies. A “bad” quantum computer — one that is not best-in-class on fidelity or qubit count or benchmarks — still gets sold, because the customer does not have the option of waiting two years for a better one to start experimenting or building a system into its data center. A company that ships an operational 50-qubit system to a national lab today has a relationship that is hard to displace, even if a 100-qubit competitor ships next year, because the swap-out cost is lower. And because the supply chain itself is constrained, even the incumbents cannot scale deliveries as fast as demand wants.
Talent mobility means that no moat in the current leaderboard is permanent. A “bad” system today can be a competitive system in 18 months if the company hires the right people. Flip-chip bonding was a genuine differentiator in the 2019, but by 2022 it was table stakes.
So who’s left?
A few things I held back from this edition that will get their own treatment soon: the Qiskit-as-CUDA argument (IBM’s pitch claimed 74% developer adoption, 7,400+ dependent projects, 83x faster transpilation than the nearest competitor), the enabling-technology supply chain in depth (the Delft-Helsinki-Gothenburg corridor and the emerging US challengers), and a comparable deconstruction of how the leading neutral-atom and trapped-ion companies are positioning.
Behind the paywall: the interactive map of all 45 superconducting companies — clusterable by geography, qubit type, funding, or fidelity — plus the company-by-company fab-access tier list (who owns their path, who has privileged access, who is fabricating at research cleanrooms and calling it a supply chain). Part 3 walks through some error correction code architecture discussion, and a little more coverage on QCI dual-rail, Alice & Bob cat qubits, Microsoft Majorana 1, and examining marketing running ahead of the physics. Part 4 deconstructs the October 2025 IBM Ventures pitch claim-by-claim.