Open in interactive viewer → charts, metric popovers & call review

Horizon Quantum Booked a $33 Million Profit Because Its Stock Fell

A 60-person, pre-revenue quantum software shop turned a warrant remeasurement into net income — while quietly building the error-correction layer its listed peers are already pricing.
HQ · Earnings Call · 2026-11-10

The profit that is really a share-price move

Horizon Quantum reported net income of $33.4 million, or $0.61 per ordinary share, for the third quarter ended September 30, 2026 — against a $5.2 million loss, or $0.13 per share, a year earlier. That is the number a screen will surface. It is also almost entirely fictional. Included in it is a $43.1 million non-cash gain from remeasuring warrant-related derivative liabilities. When the Class A share price falls, the fair value of those warrants falls, and accounting rules book the difference as income. The company's stock going down is, for GAAP purposes, a profit center. CFO Greg Gould was blunt about the mechanics: “the decrease in the company's share price during the quarter decreased the fair value of these liabilities, resulting in the accounting gain” — Greg Gould, Chief Financial Officer · 2026-11-10. And he flagged that this cuts both ways — future moves in the share price will produce more non-cash swings with no bearing on operations. CEO Joe Fitzsimons pre-empted the confusion in his own prepared remarks, which is the tell that management knows the headline is misleading: “I want to clarify that, that includes $43.1 million non-cash gain resulting from the remeasurement of warrant-related derivative liabilities due to a decrease in share price” — Joe Fitzsimons, Founder and CEO · 2026-11-10. The number to hold onto instead is the adjusted EBITDA loss of $6.3 million, versus a $3.0 million loss a year ago. The underlying business is getting more expensive to run, and honestly so. Total operating expenses came in at roughly $10.5 million versus $4.9 million, with R&D up 122% to $4.5 million (57% excluding share-based and non-recurring comp) and G&A up 182% on an adjusted basis. Headcount hit 60 full-time employees, up from 42 a year ago. Net cash used in operations ran $7.2 million for the quarter, versus $3.3 million a year earlier and $5.1 million in Q2, taking the nine-month figure to $16.4 million against $7.0 million in the first nine months of 2025. That is a burn rate roughly doubling year over year against a cash balance of $102.8 million, down $10.5 million sequentially from $113.3 million. The company started paying for trapped-ion systems it expects to receive next year, which is now bleeding through the P&L and the cash flow statement. Call it ten-ish quarters of runway at the current pace — adequate, not lavish, and shortening as the public company cost base compounds.

The deliberate refusal to have revenue

The most contrarian thing said on this call was a defense of having almost no revenue. Gould: “we do not pursue proof of concept or POC services work, which while potentially generating nominal revenue, we believe would serve as a distraction for our team and slow progress on our road map” — Greg Gould, Chief Financial Officer · 2026-11-10. Instead, the model is to convert early access users to paying customers on a cost-per-use basis — but only later. That "later" is the whole bet. The trigger is quantum advantage: the moment quantum hardware can solve problems classical machines cannot. Until then, Horizon is a pre-revenue story selling optionality on a date nobody can name. This is a real divergence from the listed peer set, and it shows up in the language. IonQ's most recent quarter leaned on Superion 256 and a quantum foundry business; IQM-adjacent reporters were quoting Order intake; D-Wave talked about average deal size. Horizon is running the opposite playbook — refusing to book small checks now in exchange for not diluting engineering focus. It is a defensible strategy and an uncomfortable one for anyone who marks the stock against revenue multiples.

Crosstalk, and the mathematics of the actual news

The quarter's genuine technical content was about characterizing imperfection. Using Rigetti's 108-qubit Cepheus processor, Horizon reduced what would have been more than 5,000 pairwise experiments to measure ZZ crosstalk — unwanted interactions between qubits — into just 7 distinct experiments. It did the analogous work on control crosstalk using its own Ember 1 testbed, where signals driving one qubit bleed into others in ways no vendor spec documents.

when I talk about a reduction to only 7 distinct experiments for a 100-qubit system, that would go only to 20 for 1 million qubit system and to 30 for 1 billion qubit system

Joe Fitzsimons, Founder and CEO · 2026-11-10
The scaling is the point. Pairwise characterization scales linearly or quadratically; this scales with the logarithm of qubit count. Horizon validated the multiplexed results against gold-standard pairwise measurements and saw close agreement. This is multiplex characterization as a competitive moat — a way to squeeze performance out of hardware that everyone else is also buying. Separately, the company integrated OQC's Toshiko 32-qubit QPU into Triple Alpha with pulse level control, letting developers run T1 and T2 characterization experiments through a web API and standard programming languages. The global keyword set has Quantum Hardware surfacing for the first time in the current quarter — Horizon's work sits precisely on that seam between hardware and software.

Entropica: buying the missing step in the compiler

The strategic headline was a non-binding letter of intent to acquire Entropica Labs for approximately $27 million in Class A shares — a stock deal, as Fitzsimons clarified under questioning, so it will not touch the $102.8 million cash balance. The technical rationale is more interesting than the price. Triple Alpha can already compile high-level programs down to a discrete fault-tolerant gate set. What it cannot yet do is encode those operations into specific error correction codes. Fitzsimons, asked how Entropica slots in: “The place we haven't focused to date is on the encoding into specific error correction codes, and that's something where Entropica has done a lot of work” — Joe Fitzsimons, Founder and CEO · 2026-11-10. The two teams have already begun joint compilation work — but still at a very early stage, by his own admission, and the LOI is subject to diligence, audits and definitive agreements. Note the peer echo: quantum error correction is now the sector's common tongue, with error correction code and logical qubit appearing across other recent reporters. Horizon is not early to the theme; it is buying its way into it after the fact.

The ladder nobody has climbed

New this quarter is a five-rung partnership framework. Level 4 means a hardware vendor can offer Triple Alpha to its users; Level 5 is default software on a given system. Fitzsimons: “We're starting to have Level 4 language in a number of the contracts. We're not ready to announce any Level 5 partnerships at this point” — Joe Fitzsimons, Founder and CEO · 2026-11-10. That is the honest boundary of the story — the Level 4 language is contract boilerplate, not a win. What also went unanswered: the Beryllium compiler update Richard Shannon asked for. The reply was about debugging tooling and compile speed, closing with “we would certainly hope to be able to have more to say later in the year” — Joe Fitzsimons, Founder and CEO · 2026-11-10. When a requested milestone migrates to "later in the year," that is a slip wearing a roadmap's clothing. Strip the warrant noise away and this is a sub-$700 million market cap company with 60 employees, no revenue by choice, about ten quarters of cash, a genuinely novel characterization technique, and a $27 million stock-funded option on the one piece of the compiler stack it lacked. The stock falling was the only reason the quarter looked profitable — and the only reason the option is cheap.