Tuesday, May 26, 2026

Xanadu: Building a Quantum Computer Out of Light, and Against the Clock

This is the story of Xanadu, built from Xanadu’s own filings and investor materials. The tone runs largely bullish, and on the science it should: this is my first pass at the bull case, told from the company’s own side of the record, where every story wears its best clothes. But I let the capital reversal speak plainly, because it is the part the best clothes cannot hide. In future notes, I will challenge this version of the story, dig deeper into the science underneath it, and revise the thesis where the facts demand it. This is not a valuationEstimating what a company is actually worth, typically from its projected future cash flows., and it is not financial advice. The story, as I tell it here, is my own opinion. Do your own research, and decide for yourself whether you agree with me.

I have been reading Xanadu’s filings, and the company hands you its whole drama in two numbers that sit a few lines apart. It has $272.5 million of cash, and it has told its own investors it will take up to a billion dollars to build the machine it exists to build.¹² Everything interesting about Xanadu lives in the space between those two figures: a genuinely world-class science story on one side, and a capital plan that is not yet funded to its own finish line on the other. This is a company that has solved problems most of its field cannot, and has not yet solved the one that pays the bills.

So let me tell it as it is: a decade of light, a public debut that raised far less than advertised, and a race between a physics breakthrough and a burn rate.

A whitepaper, and a wager on light

Xanadu was founded in 2016 by Dr. Christian Weedbrook, who started it, in his own telling, “with a whitepaper and a conviction that photonicsA qubit made from particles of light (photons). Photonic machines can run at room temperature. was the right path to a scalable quantum computer.”¹ That conviction is the entire company. Almost everyone else in quantum computing builds with matter held desperately still: superconducting circuits chilled to a hair above absolute zeroThe coldest temperature physically possible (about -273.15 Celsius), where atomic motion almost stops., ions trapped in electromagnetic fields, neutral atomsA qubit made from an uncharged atom held in place by focused laser beams ('optical tweezers'). pinned by lasers. Xanadu builds with light. Its qubitsThe basic unit of a quantum computer. Like a 'bit' in a normal computer, but instead of being only 0 or 1 it can be 0, 1, or a blend of both at once., its gates, and its measurementsReading a quantum system, which forces it out of its blend of possibilities into a single definite 0 or 1 and ends its quantum behaviour. run at room temperature, on photonic chips made in the same silicon foundries that already serve the telecom industry.³

Weedbrook is worth pausing on, because Xanadu is a founder’s company and he is an unusual founder. He took years off after high school, went back to university for mathematics, and came out the other side with a PhD in photonic quantum computing and postdoctoral stints at Toronto and MIT.²¹ When the company lists, management says, he becomes the only founder-physicist running a publicly traded quantum-computing company.² His narrative is a blend of two of the classic founder types: the Charisma founder, whose authority is the original scientific vision, and the Experience founder, validated by a decade of peer-reviewed results rather than a pitch deck. He describes himself, with a fondness for Lincoln, as a little engine that would not quit.² Read uncharitably, that is a man in love with a hard problem. Read charitably, it is exactly the temperament a fifteen-year science project requires.

And the science is real, which is the part a skeptic has to concede up front. In 2022, Xanadu’s Borealis machine performed a sampling taskA computing job that produces outputs drawn from a complex probability pattern, often used in supremacy demonstrations. in roughly two minutes that the company estimated would have taken the Fugaku supercomputer on the order of seven million years, one of the first demonstrations of quantum supremacyA demonstration that a quantum computer performed a specific task no classical computer could do in any reasonable time, even if the task itself is not useful. by any company anywhere.³ In early 2025 it went further, with a system called Aurora: by Xanadu’s account the first modular, scalable, networked photonic quantum computer, assembled from 35 integrated photonic chips and providing 12 physical qubitAn actual piece of qubit hardware. On its own it is fragile and makes frequent errors. modes at each clock cycle, with real-time error-correction decoding, and published in Nature.³ Alongside the hardware sits PennyLane, the open-source software layer Xanadu started in 2017 (as Open QML) and released in 2018, which has become one of the most widely used ways to program quantum computers of any kind, with around 200,000 downloads a month, more than 35,000 active users, and use across 143 universities.²¹

The change in fortune, and the archetype

Here is the arc. For nine years Xanadu was a private research company turning venture money and government grants into Nature papers. In March 2026 it became something else: a public company, listed on Nasdaq and the Toronto Stock Exchange under the tickerThe short symbol that identifies a stock on an exchange, such as IONQ or QBTS. XNDU, the first pure-play photonic quantum computer to trade anywhere, having merged with a special-purpose acquisition company called Crane Harbor.¹⁵

What kind of story is this? The cleanest archetype is the Better Mousetrap, and Xanadu makes the argument explicitly. It does not claim to have invented quantum computing; it claims photonicsA qubit made from particles of light (photons). Photonic machines can run at room temperature. is a structurally better way to build it. Light runs at room temperature, so there is no chilled chandelier of cryogenicsEquipment and techniques for chilling hardware to extremely low temperatures, required by superconducting qubits.. Light travels down ordinary fiber, so quantum computers can be networked together the way classical servers are. And photonic chips can be made on existing silicon-photonics lines, so Xanadu can be fablessA company that designs chips but outsources the actual manufacturing to partner foundries. and lean on partners like UMC, Tower, and Applied Materials rather than building a foundryA factory that manufactures chips. Owning one lets a company iterate its own hardware faster than outsourcing. of its own.²³ On Xanadu’s own modalityThe underlying physical approach a quantum computer is built on (trapped ions, superconducting circuits, photonics, etc.). Choosing a quantum stock is partly a bet on which modality wins. comparison, that approach delivers about 99.99% two-qubit gateA gate that acts on two qubits at once, such as the CNOT. Much harder to perform accurately than a single-qubit gate, and the real test of a machine. fidelity, all-to-all connectivityWhen every qubit can interact directly with every other qubit, rather than only with its immediate neighbors., and an order-of-magnitude lower error-correction overheadHow many physical qubits you must spend to build one good logical qubit (for example 100-to-1). Lower overhead means a smaller, cheaper machine. than the electron-based approaches.³ There is a Missionary note layered on top, in the mission statement to “build quantum computers that are useful and available to people everywhere,” and there is an Underdog’s balance sheet underneath, because the company chasing this is far smaller than IonQ or the heavily funded, secretive PsiQuantum.¹² But the engine of the story is the better mousetrap: light as the missing modality.

The most important technical claim follows from networking. Weedbrook frames the path to a useful machine as two problems, scaling and errors, and says Xanadu has solved the first. Aurora demonstrated that you can connect photonic processors with fiber and grow the system rack by rack without the loss compounding exponentially, which turns scaling into what he calls a copy-and-paste exercise.² If that holds, it is a genuine structural advantage, because every electron-based rival still has to convert its qubitsThe basic unit of a quantum computer. Like a 'bit' in a normal computer, but instead of being only 0 or 1 it can be 0, 1, or a blend of both at once. into photons to network them at all, and no one has done that at scale.²

The reversal: light is the easy part; time is the hard part

A story with no friction is a sales pitch, and Xanadu has two genuine sources of friction. The first is scientific, and the company is refreshingly candid about it. The second is financial, and it is the one that should worry a reader most.

Start with the science, because Xanadu names its own weakness. PhotonicsA qubit made from particles of light (photons). Photonic machines can run at room temperature. has an Achilles heel: loss. Photons get absorbed or scattered as they travel through the optical components, and loss is what causes the errors that stand between Aurora today and a fault-tolerant machine. Management’s metaphor is a dial: all the error-correction machinery is already built into Aurora, but the loss has to be driven down to a threshold before that machinery actually starts correcting rather than merely detecting. The plan is to get there by 2029 through better chips and better theory, and the company shows a multi-year curve of loss falling toward the line.² This is the honest center of the technical bet. Networking is solved; the dial is not yet at the threshold. If loss has a floor above where fault toleranceThe milestone where a quantum computer can run long calculations correctly despite ongoing errors. It is the field's holy grail. requires, there is no utility-scale machine, no matter how elegant the architecture.

Now the harder reversal. A quantum computer is a bet on physics; a public quantum computer is also a bet on the patience of capital, and on this front Xanadu’s debut went materially worse than advertised. When the Crane Harbor deal was struck, the pitch was roughly $500 million of gross proceeds and about $459 million of net cash at closing, assuming no redemptions.³ Redemptions did not stay at zero. They were close to total among the public shareholders: holders of more than 19 million shares redeemed at about $10.35 each, draining roughly $201 million from the trust, so the deal delivered around $302 million of gross proceeds rather than $500 million.⁵ The balance sheet tells the rest: $272.5 million of cash at the end of the first quarter.¹

Hold that against what management says the machine costs. On the Analyst Day, Xanadu put the price of the full quantum data centerA facility full of networked quantum machines working together. Xanadu's end goal. at up to $1 billion, of which going public was meant to supply up to $500 million, with the rest hoped for from governments and, if need be, from selling more stock.² So the company arrived on the public market with roughly a quarter of what it says it needs, not the half it planned for. The bridge it is building across that gap has three planks, and none is yet nailed down. The first is up to about $285 million from the governments of Canada and Ontario for a program it calls Project OPTIMISM, which the company is careful to say remains “subject to the completion of due diligence and the execution of final agreements.”¹ The second is government research money, chiefly the DARPA QBIA DARPA program checking whether an industrially useful quantum computer can be built by 2033. program, worth up to $300 million across its stages if Xanadu makes it through all of them, plus a Canadian quantum-champions program that has so far paid about $23 million.² The third is a $300 million synthetic at-the-market facility, which is to say the right to sell more Class B stock over time, diluting existing holders.¹

And there is a flag on the whole thing that an honest reader cannot skip. In the merger filings, Xanadu’s auditor, KPMG, included an explanatory paragraph stating that the company’s recurring losses and negative operating cash flows raise substantial doubt about its ability to continue as a going concernAn auditor's formal warning that a company may not have enough money to survive the next year..⁴ The de-SPACThe step in which a SPAC merges with a real operating company, completing its path to public markets. cash answers that doubt for now. It does not retire it.

The founder keeps the keys

One more structural feature deserves daylight, because it shapes who bears the risk of all that dilutionWhen a company issues new shares, shrinking each existing shareholder's slice of the company.. Xanadu is a dual-class company. Each Multiple Voting Share carries ten votes; each Subordinate Voting Share carries one.⁴ The founder, management, and the venture backers hold the multiple-voting shares; the public holds the subordinate shares, and the synthetic ATM issues more of them.

Source: Form F-4.⁴ The public’s Class B shares are roughly 14.5% of the economics but only about 1.7% of the votes. Control stays with the pre-IPO base even as the company sells stock to fund itself.

I do not read this as sinister; founder control is common in research-stage deep tech, and a fifteen-year physics bet arguably benefits from a founder who cannot be voted off it mid-experiment. But it is a fact with consequences. The people who decide when and how much stock to sell are not the people who get diluted when it is sold.

The 3P test

Run the story through the three escalating gates.

Is it possible that Xanadu builds a useful, fault-tolerant photonic quantum computer? Yes, without much strain. The supremacy result is real, Aurora is real and peer-reviewed, PennyLane is genuinely adopted, and the architecture’s advantages (room temperature, networkable, foundry-compatible) are documented rather than promised.³²

Is it plausible? This is the gate that bites, and it bites twice. The scientific condition is that optical lossLight getting absorbed or scattered as it travels through components. It is the main source of errors in photonic quantum computers. falls to the fault-tolerance thresholdThe error level low enough that adding more qubits finally reduces errors instead of adding them. Crossing it is what makes error correction actually work. on something like the 2029 schedule; the company has shown the curve bending the right way but has not reached the line.² The financial condition is that roughly $700 million of additional capital, beyond today’s cash, arrives without the dilution becoming ruinous, which means the Canada and Ontario agreements have to sign, the government programs have to keep paying, and the stock has to stay high enough that the ATM is a tool rather than a tourniquet.¹² Name those two conditions out loud and you have named the whole investment debate. The science clears plausible on its own evidence; the financing clears it only if you believe the bridge gets built.

Is it probable, the single most likely outcome among the plausible ones? Here I will be honest rather than promotional. That Xanadu remains a scientifically elite, well-regarded photonic quantum computing company that keeps shipping milestones and keeps raising money is probable. That it specifically reaches utility scale on schedule, fully funded, without a dilutive scramble, and emerges as a dominant winner of the quantum era is not, in my judgment, the single most likely path, because the loss problem is unsolved and the capital plan is a quarter complete. The honest reading is that Xanadu has bought itself runway and optionality, and a genuine shot, not a probable coronation.

The story, told in value drivers

A story earns its keep only if every claim points to a number. Here is the bridge from this narrative to the drivers a valuationEstimating what a company is actually worth, typically from its projected future cash flows. would need.

What would change this story

A Break, the kind of event that decimates the narrative, would be optical lossLight getting absorbed or scattered as it travels through components. It is the main source of errors in photonic quantum computers. proving to have a floor above the fault-tolerance thresholdThe error level low enough that adding more qubits finally reduces errors instead of adding them. Crossing it is what makes error correction actually work., which would turn the elegant architecture into a dead end; or the financing bridge collapsing, the Canada and Ontario agreements failing to sign while the stock falls far enough that the ATM cannot fund the burn, which against a going-concern backdrop is the existential risk; or a rival, PsiQuantum or IonQ or a large lab, reaching useful fault toleranceThe milestone where a quantum computer can run long calculations correctly despite ongoing errors. It is the field's holy grail. first and rendering the photonic head start moot.

A Change, a reshaping that does not kill the story, would be the Project OPTIMISM money signing and a loss milestone landing on schedule, which would re-rate Xanadu from a research-stage hopeful to a credible, government-backed scale player and complete the move from Underdog to contender. The opposite Change is just as plausible: the utility-scale machine slips, and Xanadu settles into being a photonics-IP-and-software company, monetizing PennyLane, packaging, and components like the RTX quantum-sensing chips, which is a smaller and more honest identity than the quantum-data-center dream.²

A Shift, the ordinary quarterly evidence, is the loss-reduction curve per chip run, PennyLane’s active-user and download trend, the pace of the cash burn against the up-to-$1-billion need, the status of the government negotiations, and whether management holds its line on refusing easy non-recurring-engineering revenueThe total money a company brings in from sales, before any costs are subtracted..

I will end where I began, with the two numbers. Xanadu has done the genuinely hard scientific thing, which is to make a quantum computer out of light, network it, and publish the proof, and that is more than most of its field can say. What it has not done is fund the finish line, and it arrived on the public market with roughly a quarter of what it says the finish line costs, carrying an auditor’s note about its ability to keep going. The science is the bull case and it is real. The capital is the reversal and it is also real. For a company this early, the most useful way to hold Xanadu is to keep those two truths in view at once: the light is working; the clock is running.

Sources & notes

1. Xanadu first-quarter 2026 results press release (May 14, 2026): the Nasdaq and TSX listing under XNDU as the first publicly listed pure-play photonic quantum computing company and the completion of the Crane Harbor business combination; Dr. Christian Weedbrook’s founding of Xanadu in 2016 “with a whitepaper” and the room-temperature, light-based, fault-tolerant approach; cash of $272.5 million and Q1 revenue of $2.8 million (up 4x year over year) against a net loss of $20.6 million; the up-to-$285 million (CAD $390 million) funding discussions with the governments of Canada and Ontario for Project OPTIMISM, subject to due diligence and final agreements; PennyLane reaching over 35,000 active users and roughly 200,000 monthly downloads; the AMD collaboration (a 20-qubit, 35-million-gate quantum computational-fluid-dynamics demonstration with 25x workflow acceleration); the appointments of CFO Michael Trzupek and CLO Natalie Wilmore; the planned $300 million synthetic at-the-market facility issuing Class B shares; and more than $500 million raised to date.
2. Xanadu Analyst Day transcript (March 4, 2026): Christian Weedbrook’s biography and “little engine” framing and the company’s description of him as the only founder-physicist CEO of a publicly traded quantum-computing company; the vision of hundreds of networked server racks forming quantum data centers; networking solved by Aurora while optical loss remains the central unsolved problem, with the loss-reduction “dial” targeted to reach the fault-tolerance threshold by 2029; the GKP-qubit approach and the fabless, silicon-photonics manufacturing model with partners including UMC, Tower, and Applied Materials; the up-to-$1 billion cost to build the quantum data center, with up to $500 million from going public and the remainder sought from governments (DARPA’s QBI program, worth up to $300 million across its stages, and Canada’s quantum-champions program, from which about $23 million had been received); the roughly 70-75% of spending directed to hardware; the Series C 2022 valuation of about $1 billion stepping up to about $3 billion at the merger; PsiQuantum as the cryogenic photonic rival; the post-closing board’s semiconductor, defense, and audit expertise and CFO Michael Trzupek’s roughly decade at Intel; PennyLane’s history (Open QML in 2017, PennyLane in 2018) and reach (143 universities, 230 courses); the PIPE described as the largest quantum de-SPAC PIPE since 2022 with about 90% new investors; the RTX quantum-sensing chip subcontract; and management’s stated choice to forgo non-recurring-engineering revenue.
3. Xanadu and Crane Harbor investor presentation (November 2025): the “first and only publicly listed pure-play photonic quantum computing company” and “missing modality” framing; the pro forma enterprise value of about $3.1 billion, the $275 million PIPE, roughly $224 million of trust cash, and about $459 million of net cash at closing assuming no redemptions; PennyLane’s approximately 47% quantum-user penetration and 120 approved and pending patents; the 2022 Borealis quantum-supremacy demonstration (216 qubits, about two minutes versus an estimated seven million years on the Fugaku supercomputer) and the 2025 Aurora system (a modular, scalable, networked photonic quantum computer: 35 integrated photonic chips, 12 physical qubit modes at each clock cycle, an 86.4-billion-mode cluster state, real-time error-correction decoding, and publication in Nature); the modality comparison (about 99.99% two-qubit fidelity, roughly 100:1 logical overhead, all-to-all connectivity, room-temperature operation); the four-pillar commercial model (cloud access, system sales, software, and photonic devices and IP); the customer case studies (Volkswagen, Mitsubishi Chemical, Rolls-Royce); and the BCG estimate of $450 to $850 billion of value creation at full fault-tolerant scale.
4. Xanadu Form F-4 / proxy statement-prospectus: KPMG’s audit opinion containing an explanatory paragraph that recurring losses and negative operating cash flows raise substantial doubt about the company’s ability to continue as a going concern; the dual-class structure under which each Multiple Voting Share carries ten votes and each Subordinate Voting Share carries one; the roughly 254.7 million Class A Multiple Voting Shares and 43.3 million Class B Subordinate Voting Shares outstanding; and Christian Weedbrook’s founder offer letter dated April 2017.
5. The Crane Harbor business combination was approved on March 19, 2026 and closed on March 26, 2026, with the combined company beginning to trade on March 27, 2026; holders of 19,428,395 Crane Harbor public shares redeemed at about $10.35 each (roughly $201 million), so gross proceeds came in near $302 million rather than the roughly $500 million the deal had originally contemplated assuming no redemptions. TipRanks and Canadian press coverage, March 2026, tipranks.com and bnnbloomberg.ca.