Anti-dilution protection adjusts the rate at which preferred stock converts into common when the company later sells shares below the price that preferred paid. It does not stop dilution from new shares being issued, which ordinary pro rata arithmetic governs; it specifically compensates price-based dilution, the loss of value from a cheaper round.

The mechanism is the conversion price. Preferred shares convert to common at a ratio of original price over conversion price; lowering the conversion price raises the number of common shares received. Full ratchet resets the conversion price to the new round’s price, however few shares that round sold, as if the protected investor had invested at the lower price all along. Weighted average adjusts proportionally, blending the old price and the new price by the relative sizes of the existing capitalization and the cheap issuance; the broad-based variant counts the fully diluted capitalization in that blend, which softens the adjustment, and is the standard.

Carve-outs matter: option pool grants, conversion of existing instruments and similar issuances are customarily exempt, so routine operations do not trigger the clause. Because every extra as-converted share for protected preferred comes out of common, the founders absorb both the down round and the adjustment; that double hit is why the form of this clause deserves negotiation time that founders usually spend elsewhere.

Backlog and pipeline answer different questions. Backlog is the value of work that is contracted but not yet delivered or recognized: signed, funded, owed to the company. Pipeline is the value of opportunities still being pursued and not yet won. Backlog is a near-certainty (subject to delivery and cancellation terms); pipeline is a probability-weighted hope. Treating them as one number is the most common revenue overstatement diligence corrects.

The clean presentation keeps them in separate columns and qualifies each. Backlog: contract value, expected delivery schedule, cancellation and acceptance terms, and whether the customer’s funding is in place. Pipeline: staged by likelihood (qualified, proposal, verbal, etc.), with a blended conversion assumption and an honest age (a deal that has sat at “proposal” for a year is not really pipeline). The quality questions are familiar from quality-of-revenue work: how concentrated is the backlog in one customer, are the contracts cancellable for convenience, and does the pipeline conversion assumption match the company’s own history.

For a deep tech company the trap is double. Much of what gets called backlog is really conditional: pilots that convert only if a milestone is hit, contracts contingent on a grant landing, orders with long acceptance gates. And much of the pipeline is non-commercial interest (research collaborations) mislabelled as sales. The credible founder reports a small, clean backlog, a realistically staged pipeline, and the gating conditions on both, rather than a single headline number that diligence will immediately decompose.

A bridge round is financing raised between priced rounds to extend the company’s runway to an event that should improve its price: a technical milestone, a commercial proof, a better market. It is usually structured as convertibles (SAFEs or notes) rather than priced equity, most often from existing investors, at terms anchored on the last round (same cap, or a modest discount to the next round).

The instrument is neutral; the reason is everything. A good bridge is an investment case: the milestone is named, the amount is sized to reach it with margin, and the insiders writing it can say why the milestone changes the next round’s price. A defensive bridge, money to postpone a hard conversation, shows up in the next diligence as exactly that, and stacks one more cap onto the convertible pile the Series A must digest.

The signaling cuts both ways. Insiders bridging at the prior cap reads as conviction if the milestone story holds, and as a quiet markdown if it does not; outsiders joining a bridge strengthens it. The honest comparison a founder must run before bridging: against a smaller priced round at a lower valuation today, which path reaches the value-creating proof with less total dilution and cleaner governance? Sometimes the down round is the cheaper bridge.

The burn multiple, popularized by David Sacks, measures capital efficiency of growth: net burn over a period divided by net new ARR added in that period. A multiple of 1 means the company burned a dollar to add a dollar of recurring revenue; on Sacks’s scale under 1x is amazing, 1 to 1.5x great, 1.5 to 2x good, 2 to 3x suspect, above 3x bad, and the ratio rises sharply as a company stalls. Its appeal is that it captures in one number whether growth is being bought cheaply or expensively.

The structural caveat for deep tech is that the metric is defined only when there is recurring revenue to put in the denominator. A pre-revenue quantum company has net new ARR of zero, so the burn multiple is either infinite or undefined; computing it is meaningless, and dressing burn up as a multiple against a sliver of NRE or pilot revenue is worse than not reporting it. The metric belongs to the post-traction phase, after a repeatable revenue line exists.

The honest pre-revenue substitute is to express efficiency against milestones rather than revenue: how much capital is consumed to reach the next proof that re-rates the company (a demonstrator, a fidelity threshold, a qualified system). That figure is the deep tech cousin of capital efficiency, it lets a board compare planned against actual spend per unit of de-risking, and it sets up the burn multiple to be used properly once recurring revenue finally arrives. Reporting milestone cost pre-revenue and the burn multiple post-revenue is the credible sequence; forcing the multiple early is a tell that the metrics were chosen to flatter rather than to inform.

Burn rate measures cash consumption per month, in two flavors that must never be confused in the same sentence. Gross burn is total cash out: payroll, rent, equipment, services. Net burn subtracts reliable cash in (grant instalments, tax-credit refunds when received, early revenue), and is the figure that divides into cash to give runway. A company stating “burn” without the adjective in a board document creates exactly the ambiguity diligence exists to catch.

Reading burn well means normalizing it. One-off items (an equipment purchase, a legal bill, an annual insurance premium) belong in a separate line, not in the monthly run rate; a founder who lets a cryostat purchase sit in the March burn figure shows a 2x spike that means nothing. The useful presentation is a run-rate burn with one-offs called out, trended over quarters.

The management discipline is tiering: which costs are committed (leases, notice periods), which are controllable within a quarter, which are discretionary today. That tiering is what turns a downside scenario from a spreadsheet exercise into an executable plan, and it is the first question a serious board asks when a milestone moves.

The Crédit d’Impôt Recherche is France’s research tax credit and one of the most generous R&D incentives in Europe. It grants a credit equal to a percentage of a company’s eligible research expenditure, the headline rate is 30% on eligible R&D up to a high threshold, with a lower rate above it. The base covers research salaries, depreciation of research equipment, and certain outsourced research to approved bodies. The exact rates, base and any special treatments are set in the annual finance law and change, so a live claim should be sized against the current year’s rules. It is a cornerstone of how French deep tech, including quantum, finances its science-heavy early years.

Whether the credit is cash or a tax offset depends on the company. For SMEs, and in particular for a Jeune Entreprise Innovante or a young company, the CIR is refundable: if it exceeds the tax owed, the difference is paid out, which is what makes it true non-dilutive cash for a pre-revenue startup. For larger, profitable companies it reduces corporate tax and, if unused, is carried forward and becomes refundable after three years. The JEI status often sits alongside the CIR, adding social-charge relief on R&D staff.

Two operator realities shape the cash plan. First, timing and financing: the credit is claimed for a fiscal year and normally paid the following year, but banks and Bpifrance routinely pre-finance the receivable, advancing most of the expected credit so the company does not wait, which is why a French runway model can treat a well-documented CIR more like near-term cash than a Canadian SR&ED refund. Second, documentation and audit: the eligible R&D must meet a genuine standard of scientific or technical uncertainty and be documented to survive a tax audit, the CIR is generous but scrutinized, and an aggressive, thinly-supported claim is a liability rather than an asset.

A convertible note is debt that intends to become equity. The investor lends money today; the principal, plus accrued interest, converts into shares when a qualified financing closes, usually defined as a priced equity round above a negotiated threshold. Conversion terms mirror a SAFE’s: a valuation cap, a discount on the round price, or both, with the investor converting at whichever price is lower. The conversion math is the same; the base differs. A note’s cap is conventionally pre-money, so, unlike a post-money SAFE, it does not lock the holder’s final ownership: other instruments converting at the same round dilute the note too. The $8,000,000 cap in the worked example is a pre-money cap.

Three features separate it from a SAFE, and all three protect the investor. It accrues interest, typically 5 to 8% and most often simple, which converts with the principal rather than being paid in cash. It sits on the balance sheet as a liability until conversion. And it has a maturity date: if no qualifying round has closed by then, the loan is due, and the realistic outcomes are an extension, a negotiated conversion, or a default no early-stage board wants to test.

Notes persist where SAFEs are awkward. Some investors want the seniority of debt in a liquidation; some jurisdictions handle loan instruments more cleanly than SAFE-style contracts (France, for instance, evolved its own BSA-AIR rather than adopt the SAFE). For the founder the practical reading is simple: a note is a SAFE plus a clock plus a coupon. Price the clock honestly against the technical roadmap before preferring one to the other.

Corporate venture capital is venture investing done by an operating company’s dedicated arm, IBM, a defense prime, a chip maker, a telecom, rather than by an independent fund. CVCs invest for a blend of financial and strategic return: the parent wants exposure to a technology that matters to its business, a window onto innovation, a possible future supplier, partner or acquisition. That dual motive is what makes CVC different from a pure financial VC, and it cuts both ways.

The upside for a deep tech company can be substantial. CVCs often have deeper and more patient balance sheets than fund-constrained VCs, they bring technical people who can genuinely validate the science, and the corporate can become a design partner, a first customer, or eventually the acquirer. For a quantum company whose natural buyers are a handful of large platforms, a CVC on the cap table can be a real strategic asset.

The risks are specific and worth pricing. Signalling: an investment from one corporate can make its competitors reluctant to engage, narrowing the future customer and acquirer set, the opposite of what a startup wants. Information and IP: a strategic investor gets a close look at the roadmap and technology, which is sensitive when the parent could build or buy a competitor. Alignment and durability: CVC mandates and champions change with corporate reorganizations and strategy shifts, so the patient strategic partner of this year can go quiet next year. And terms: some CVCs seek rights (rights of first refusal on an acquisition, exclusivity, board influence) that constrain the company’s options. The discipline is to understand the strategic thesis behind the cheque, negotiate away the most constraining rights, and treat the strategic upside as a bonus rather than the basis of the plan.

The data room is the document set behind a fundraise, today a structured cloud folder with access logging rather than a physical room. Its canonical sections: corporate (incorporation, bylaws, board minutes, prior financings), equity (cap table, every SAFE and option agreement), IP (assignments, patents and applications, licenses), finance (statements, the model, grant agreements), commercial (contracts, letters of intent, pipeline evidence), team (employment and IP-assignment agreements, key hires) and technical documentation appropriate to the stage.

Staging is legitimate and expected. A first-meeting room can hold the deck, a summary cap table and headline metrics; the full room opens at term sheet. What is not legitimate is inconsistency: two documents disagreeing on the share count, an outdated model next to a newer deck, a “final” folder with three finals. Diligence analysts are trained to read version hygiene as management hygiene.

The working discipline is to maintain the room continuously rather than assemble it in the panic week after a term sheet: every new contract, grant or board decision files into its section on signature, dated. The marginal cost is minutes per week; the alternative costs weeks of diligence delay at the exact moment leverage is decaying.

The discount is the simpler of the two conversion mechanics on SAFEs and convertible notes. At the next priced round, the holder converts at the round price reduced by the discount rate, typically 10 to 25%, with 20% the most common single number. When the instrument also carries a valuation cap, the investor converts at whichever price is lower: the discounted round price or the cap-implied price. The two terms are a floor and a ceiling on the same risk premium, and they never combine (no discount on top of the cap price).

The economic logic is compensation for early risk: the discount hands the early investor a better price than the round investors who waited for more evidence. Its weakness is symmetry. A 20% discount pays the same whether the priced round closes in eight months or in three years, and whether the company’s value multiplied or merely survived. The cap exists precisely to repair that: it converts patience into ownership when the company outperforms.

In a term sheet negotiation the discount is rarely the battleground; caps carry the real economics. The founder’s discipline is simply to model both paths at signature: at what round price does the discount bind rather than the cap, and what does each scenario cost in fully diluted ownership.

A down round is an equity financing at a price per share below the price of the previous round. The comparison is per share, not headline valuation: a company can raise at a higher post-money and still price down if the share count grew through pools and conversions.

The direct mechanics are three. Anti-dilution provisions on existing preferred adjust conversion prices downward, shifting extra dilution onto common. New money buys more of the company per dollar, compounding that dilution. And options granted at the old fair value sit underwater, which is a retention problem precisely when retention is hardest.

The indirect mechanics often cost more. A down round resets the reference price every future negotiation anchors on, tests investor relationships (pay-to-play pressure, board renegotiations frequently ride along), and reads as a signal to employees, customers and the next fund unless the narrative is controlled: what was repriced, the market or the execution, and what the record shows.

The decision discipline is to compare the down round against its real alternatives (bridge, tranche, cost cuts extending runway to the milestone) on one axis: which path gets the company to its next value-creating proof at the least total dilution and the least damage to the people who must build it. Protecting a headline number is not on that axis.

A drag-along right lets a specified majority of shareholders, on approving a sale of the company, compel the remaining holders to sell on the same terms. Its purpose is to keep an exit executable: a buyer usually wants 100% of the company, and without a drag a small minority could refuse to sell and block or extract a premium on a deal the majority wants. The clause “drags” the minority into the approved transaction.

The mechanics that matter are the trigger and the protections. The trigger defines whose approval activates the drag: often a majority of the preferred, or of the preferred and common voting together, sometimes with board and a threshold of common. A founder should know exactly what coalition can trigger it, because it determines who can force a sale. The protections for the dragged holders are the other half: they should receive the same price and form of consideration as the approving majority, should not be asked to give representations or indemnities beyond their own shares, and ideally have their liability capped at their proceeds. A drag without these protections can force minority holders into a deal on worse effective terms.

For founders the drag-along intersects with the liquidation-preference stack in a way that can sting. Because preferences pay out first, a sale that satisfies a deep preference stack can leave little for common, and a drag can compel the founders and employees to accept exactly that outcome. This is why the exit math should be modelled when the clause is negotiated, not at the exit: a reasonable drag (clear trigger, equal terms, capped liability) is standard and healthy, but its interaction with preferences determines what the people building the company actually receive when the drag is finally pulled.

Dual-use funding supports technologies useful for both civilian and defense or security purposes. Because quantum technologies sit squarely in this category, sensing for navigation and detection, quantum-secure communication and key distribution, computing with cryptographic implications, defense agencies and security-oriented programs are a significant source of non-dilutive money for quantum companies. This pool is often deep and patient, funding long-horizon research that civilian markets are not yet ready to pay for, which suits the deep tech timeline well.

The trade-off is a heavier set of strings than a civilian grant carries. Accepting defense or dual-use funding can trigger export-control obligations, the technology may be classified under regimes that restrict where it can be sold and to whom; constraints on ownership and personnel, some programs limit foreign ownership, control or influence and restrict who can work on the funded effort; and government rights in the resulting intellectual property, which can include licenses for state use or step-in rights (the US calls these march-in rights; the exact mechanism varies by jurisdiction). Each of these can later interact with a fundraise (a foreign lead may be complicated by ownership rules) or a cross-border acquisition (export-controlled IP can narrow the set of eligible buyers).

For an operator the discipline is to go in with eyes open: map the obligations before signing, not after. The questions are which export-control regime applies, what ownership and hiring restrictions attach, and what rights the funder takes in the IP. Dual-use funding is genuinely valuable, often the natural home for early quantum work, but it shapes the company’s future optionality in ways a simple civilian grant does not, and those constraints belong in the data room and in the strategic plan, not as a surprise discovered during the next round’s diligence.

Due diligence is the verification phase of an investment: the period, mostly between term sheet and closing, when the investor tests what the pitch asserted; serious deeptech leads front-load the technical workstream before they price. The standard workstreams are corporate and legal (incorporation, contracts, litigation, IP ownership), financial (accounts, runway model, cap table), commercial (market, customers or letters of intent), team (references, background) and, in deeptech, technical diligence with domain experts.

Two findings categories matter differently. Confirmatory findings adjust detail: a contract to amend, a number to restate. Red flags reprice or kill: IP not actually assigned to the company, an undisclosed liability, a claimed result that does not replicate, a cap table that does not reconcile. Most diligence failures trace to surprises, not to weaknesses; a weakness disclosed early with a plan reads as maturity, the same weakness discovered in week five reads as concealment.

Diligence runs in both directions. The founder is entitled to references on the fund: how it behaved in its last down round, whether reserves exist for follow-on, what its board members are like under stress. The investor who resents reverse diligence is answering it.

The EIC Accelerator, run by the European Innovation Council, funds individual high-risk, high-potential SMEs developing breakthrough or deep tech. Its distinctive feature is blended finance: a grant component for development and demonstration activities, combined with an optional direct equity investment made through the EIC Fund for scale-up and market deployment. The grant covers work up to roughly market-readiness; the equity is meant to crowd in private investors for the capital-heavy commercialization phase that pure grants cannot fund.

It targets companies whose technology has progressed past early research, the program is oriented to the mid-to-upper technology-readiness band, with the grant aimed at moving from a validated prototype toward a market-ready system. Quantum, photonics and other deep tech fit its mandate well, which is why it is a prominent line in European deep tech funding. The amounts are substantial relative to most grants, which is part of why it is sought after.

Two cautions belong in any plan that includes it. First, it is not purely non-dilutive: if the equity component is taken, the EIC Fund becomes a shareholder, so the instrument sits between grant and equity, and the cap-table effect must be modelled. Second, it is intensely competitive and procedurally heavy, multi-stage applications, interviews, long timelines, and low success rates, so a disciplined company treats it as a deliberate, well-prepared bet with a real opportunity cost, never as committed runway. Won, it can fund a major step; assumed, it is the kind of optimistic inflow that makes a runway model fragile.

A follow-on is an additional investment by an existing shareholder into a later financing round. Funds plan for this: alongside the capital they deploy in first cheques, they hold reserves to back their winners again in subsequent rounds, and a meaningful share of a good fund’s returns comes from concentrating capital into the companies that are working. For a company that raises repeatedly, the follow-on behaviour of its existing investors is a structural feature of the cap table, not an afterthought.

The signal value is high in both directions. When existing investors follow on, especially the insiders closest to the company, it tells a prospective new lead that the people with the most information are choosing to put in more, which de-risks the deal and often anchors the round. When insiders conspicuously do not follow on, the new lead reads it as a warning: either the informed money sees a problem, or the existing funds lack reserves, and either way the round gets harder and the implied price softer. This is why an insider follow-on can be worth more than its dollar amount.

For founders the practical move is to treat reserves as a diligence item on the fund, asked early and politely: how much does the fund typically reserve for follow-on, does it have dry powder at the company’s stage, and what is its track record of supporting portfolio companies through later and tougher rounds. A fund that writes a strong first cheque but cannot or will not follow on leaves a gap that must be filled by new investors at exactly the moments that gap is hardest to fill. Mapping which existing and prospective investors have the reserves and the appetite to follow on is part of planning a multi-round deep tech financing rather than just the next round.

Freedom to operate is the ability to make, use and sell a product without infringing the valid intellectual-property rights of others. It is distinct from owning IP: a company can hold a strong patent on its own invention and still infringe a third party’s patent on a technique it must use to ship. FTO is assessed by searching the relevant patent landscape for claims that read on the intended product, then clearing them by designing around, licensing, or forming a view that the claims are invalid or do not apply.

In quantum the risk is elevated because the patent landscape is both dense and foundational. Large incumbents and research institutions have spent two decades filing across qubit modalities, control methods, error correction and fabrication, so a startup’s core approach may sit close to, or squarely within, existing claims. The danger is asymmetric: the time to discover an infringement problem is before scaling, not after a product is in market and an incumbent has both the patent and the lawyers.

For an investor, FTO is a specific diligence line, not a footnote to “we have patents”. The questions: has a freedom-to-operate analysis been done, by competent counsel, and what did it surface; are the blocking patents licensed, designed around, or genuinely absent; and how exposed is the company to the big portfolio holders in its modality. A clean, documented FTO position is a real asset in deep tech; its absence is a latent liability that can surface at the worst possible moment, typically when the company is finally large enough to be worth suing.

A full-stack quantum company builds the entire system, from qubits up through control electronics, software and applications, aiming to deliver a working computer or a complete solution. An enabling-technology company builds one layer of the stack (cryogenics, control systems, lasers and photonics, error- correction software, fabrication) and sells it to others, often to the full-stack players themselves. The choice shapes everything downstream: capital intensity, time to revenue, customer set, competitive dynamics and the kind of acquirer that eventually buys the company.

The trade-offs are close to opposite. Full-stack captures the biggest potential value and controls its own destiny, but carries the largest burn and the most binary risk, success requires every layer to work and the integration to hold. Enabling technology, the picks-and-shovels position, has a clearer near-term market (every full-stack effort is a potential customer), smaller individual upside, and less all-or-nothing risk, but lives with the threat that a large platform vertically integrates the layer and removes the market overnight.

For an investor the term forces three checks. First, which layer does the company truly occupy, because “full-stack” is a popular claim and a company that buys most of its stack from others is really an integrator. Second, does the business model match the position: enabling players are judged on near-term revenue and design wins, full-stack players on milestones and capital efficiency toward a working system. Third, what protects the position if the biggest players move, which for enabling companies turns on IP and switching costs, and for full-stack companies on integration know-how and freedom to operate. Naming the position honestly is the first step to a defensible plan.

Fully diluted capitalization counts every share that exists plus every share that could exist under instruments already promised: granted options, the unissued option pool, warrants, and convertibles (SAFEs, notes) at their assumed conversion. It answers the only question that matters for ownership: if everything outstanding became common stock today, what fraction would each holder have?

The convention has edges worth knowing. The unissued pool is normally included, which slightly understates everyone’s percentage relative to what exists today but correctly states the post-grant future. Convertibles require an assumption (conversion at cap is standard for planning); a fully diluted table should say which assumption it uses. Expired and forfeited instruments drop out; out-of-the-money options and warrants do not, as long as they are outstanding they stay in the count (moneyness matters for accounting dilution, not for the cap table).

Price per share in a round is set on this denominator: pre-money valuation divided by the fully diluted count before the new money, including the round’s pool increase and the converting instruments. That single mechanic is why the definition of fully diluted gets negotiated in term sheets, and why an oversized pool or an unmodeled SAFE stack quietly lowers the price the investor pays. The founder’s protection is unglamorous: maintain one cap table where every percentage is fully diluted, keep the conversion assumptions written down, and never let two documents in the same data room disagree on the denominator.

A fund’s thesis is the investment strategy it raised its capital to execute, and its mandate is the harder boundary of what it is actually allowed and expected to do. Together they define the box: which sectors and technologies, which stage (pre-seed, seed, Series A, growth), which geographies, what cheque sizes and ownership targets, and sometimes harder constraints set by LPs (no defense, ESG limits, a fund-of-funds restricted to certain regions). A fund does not invest in good companies; it invests in good companies that fit its box.

For a founder this is the most under-used filter in fundraising. A large share of rejections are not judgments on the company at all, they are out-of-mandate: wrong stage, cheque too small or too large, geography the fund cannot serve, sector the fund does not touch. Reading the thesis before pitching turns a scattershot process into a targeted one. The strongest signal is a fund whose mandate not only permits an investment in you but pushes toward it, a fund that announced a quantum or deep tech focus has told its LPs it will deploy into exactly your space, and has to find deals to honour that.

The practical work is to qualify funds the way a salesperson qualifies accounts: confirm stage and cheque fit, geography, sector mandate, and where the fund is in its life (a fund mid-deployment is hungrier than one fully committed or fundraising its next vehicle). It also means reading the thesis honestly against your own raise, if you are a pre-revenue hardware seed, a growth fund’s interest is usually a learning meeting, not a real prospect. Matching the raise to the mandate is upstream of the pitch, and getting it right is most of what separates an efficient process from months of polite rejections.

A venture fund has two sides. The general partners (GPs) are the firm: they raise the fund, source and pick investments, sit on boards, and decide follow-ons and exits. The limited partners (LPs) are the investors in the fund itself, pension funds, endowments, family offices, sovereign and corporate investors, sometimes governments, who commit capital but stay passive, with liability limited to what they commit. When a startup “raises from a VC”, the GP writes the cheque, but the money ultimately belongs to the LPs, and the GP is accountable to them.

The economics align the two sides imperfectly. GPs typically earn a management fee (often around 2% of committed capital a year) to run the firm, and carried interest (commonly 20% of profits, after LPs get their capital back and sometimes a preferred return) as the real upside. The fund has a finite life, classically about ten years, with an investment period early and a harvest period later. That clock is the part founders most often ignore: a fund must eventually return capital and gains to its LPs, so its willingness to back a long-horizon company depends partly on where it sits in that decade.

For a quantum founder the GP/LP structure has two practical consequences. First, fund age matters: capital from a fund early in its life is more patient than capital from one approaching the end, which is staring at a return deadline. Second, the LP base matters: a fund whose LPs understand and want deep tech can hold a position through a long roadmap and follow on; a fund whose LPs expect quick software-style returns will pressure the GP, and that pressure reaches the board. Asking, diligently and politely, about fund vintage, remaining reserves and LP appetite is reverse diligence that a serious founder does before taking the money.

Government and agency funding comes in two economically different forms that are easy to lump together. A grant (or non-repayable contribution) is money the company keeps outright if it meets the programme’s conditions, true non-dilutive capital with no payback. A repayable contribution is money the company must return, sometimes unconditionally on a schedule, more often conditionally, repaid only if the project succeeds, the product reaches market, or revenue crosses a threshold, occasionally as a royalty on future sales. Some programmes blend the two, part grant, part repayable.

The distinction matters because it changes the real cost and the balance-sheet picture. A grant is the cheapest capital there is. A repayable contribution, even an interest-free, success-contingent one, is a form of soft financing: it may be recorded as a liability, it can reduce the non-dilutive benefit on a present-value basis, and crucially it can be treated as debt-like in a future financing or acquisition, reducing the equity value at exit in the same way other debt-like items do. Money that felt free when it landed can re-appear as a deduction in an EV-to-equity bridge years later.

The operator discipline is to read every term sheet for the repayment trigger and classify the funding accordingly, not by the marketing label the programme uses. The questions: is it repayable at all, under what conditions, on what schedule, with what interest or royalty, and does it convert to a grant on failure or stay owed. A clean non-dilutive stack distinguishes the kept money from the owed money, so the runway it shows is honest and so a future buyer finds no surprises in the contribution agreements.

The investment committee (IC) is the decision-making body inside a fund that must approve a new investment before it closes. Even when a partner is enthusiastic, the deal usually has to clear the IC, which may include the fund’s senior partners and sometimes LP representatives. The partner becomes the deal’s champion, presenting a memo and defending the case; the founder is frequently not in the room. This is why fundraising is partly a process of equipping your champion to win an argument you cannot attend.

The IC exists to impose discipline and consistency, to make sure each deal fits the thesis, to pressure-test the optimism of the sponsoring partner, and to weigh the deal against the rest of the portfolio. For a deep tech company the IC is also where the technical story meets non-technical scrutiny: the committee may have no domain expert, so it will lean on external diligence and on whether the claims are independently verifiable, and it will ask the blunt questions, who buys this, when, against whom, and what does this round actually prove.

The practical implication shapes how a founder runs a raise. Give your champion the artifacts that survive the room without you: a milestone defined so it can be verified by a third party, a clean data room, an honest competitive and market read, a crisp use of proceeds tied to the next value-creating proof, and references that will hold up. Understand that a “yes” from the partner is a “maybe” until the IC clears it, that conditions and reduced terms can appear at this stage, and that a deal can die in committee for reasons that have nothing to do with the meeting that went well. Designing the pitch for the people who were not at the meeting is what separates founders who close from founders who get enthusiastic first calls.

The lead investor is the round’s organizer: the fund that negotiates the term sheet, sets the price, runs the deepest diligence, usually takes the board seat, and writes the largest single cheque, customarily a third to a half of the round or more at early stage. The rest of the syndicate, the followers, invests on the lead’s terms with lighter diligence of their own. A round without a lead, the party round of many small cheques on identical convertibles, can close faster but leaves no one accountable: nobody priced the company, nobody owns the follow-on decision, and the next downturn finds an empty chair where conviction should sit.

Leads are qualified, not just accepted. The questions that matter: does the fund have reserves and a practice of following on; what does it do when a portfolio company misses a milestone (references from founders who lived it, including failures); who exactly takes the board seat and how do they behave in conflict; does the fund’s thesis and time horizon match the technology’s. A lead’s value concentrates in the hard moments, bridges, down rounds, re-pricings, which is precisely when a flattering but uncommitted investor costs the most.

Securing a strong lead is also the fastest way to fill a round: followers exist in quantity, conviction is the scarce input.

Assignment and licensing are two ways a company can come to use intellectual property, and they are not close substitutes. Assignment transfers ownership: the IP becomes the company’s asset, to defend, license out, or sell. A license grants permission to use IP that someone else continues to own, on whatever terms the licensor sets. Founders and investors generally want the company’s core technology assigned to it; licensing is acceptable for peripheral or genuinely third-party IP, and sometimes unavoidable when a university owns the underlying research.

When the IP arrives by license, the terms decide how much it is worth. Exclusivity is first: an exclusive license behaves much more like ownership than a non-exclusive one, which lets the licensor grant the same rights to competitors. Then scope: is it worldwide, does it cover the full field of use the business needs, can the company sublicense. Then durability: can the licensor terminate, are there diligence milestones, what happens to the license in an acquisition (some terminate on change of control, which can poison an exit). And finally cost: upfront fees, running royalties, and any equity or milestone rights that tax every future dollar.

For an investor the practical rule is simple. Assigned, owned core IP is the strongest position. An exclusive, worldwide, broad, durable license is a close second. A non-exclusive, narrow, terminable, or royalty-heavy license to the company’s central technology is a structural weakness that shapes valuation and can complicate or block an exit, no matter how good the science is. This is one of the first things a deep tech diligence reads in the data room, and one of the most expensive to fix after the fact.

The liquidation preference defines what preferred shareholders receive, ahead of common, when the company is sold, merged or wound up (a “liquidity event”, not just an actual liquidation). The standard term is 1x non-participating: the investor chooses the better of getting their money back or converting to common and taking their percentage. The aggressive variants multiply: a 2x or 3x preference returns multiples of cost first; participating preferred takes the preference and then shares in the remainder (“double dip”), sometimes softened by a cap.

Stack order matters as much as size. Across several rounds, preferences are either stacked, the latest money out first, or pari passu, all preferred sharing proportionally. A deep, stacked, senior preference pile changes who gets paid at every exit value below the stack’s total.

The discipline is to negotiate the term with the waterfall open, not in the abstract: 1x non-participating, pari passu where it can be had, and a model showing what common receives at a range of exit values. A flattering headline valuation bought with a participating 2x preference is frequently worth less to founders, at every realistic exit, than a lower valuation on clean terms.

A logical qubit is a qubit protected by quantum error correction: its quantum information is spread across many physical qubits so that errors can be detected and corrected faster than they accumulate. It is the unit that matters for large, useful computation, because raw physical qubits are too noisy to run deep algorithms. Error correction works only once physical error rates drop below a fault-tolerance threshold; above that threshold, adding qubits adds noise faster than correction removes it, and no amount of scale helps.

The number that decides a hardware roadmap is the overhead: how many physical qubits are needed per logical qubit, which today can run from dozens to many thousands depending on the modality and the target error rate. This is why physical qubit counts, the figure most often headlined, can be misleading: a machine with a thousand noisy physical qubits and an error rate above threshold has zero logical qubits and cannot run a useful fault-tolerant algorithm. Fault tolerance is the regime where logical error rates can be driven arbitrarily low by scaling; reaching it is the central milestone of most serious hardware programs.

For an investor, the logical-qubit lens reframes the entire technical claim. The questions are: how many logical qubits, at what logical error rate, with what physical-to-logical overhead, and how far below threshold the physical devices operate. A roadmap stated in logical qubits with honest overheads is fundable; a pitch that leans on physical qubit counts while staying quiet on error rates and fault tolerance is quoting the flattering number and hoping the diligence does not ask the next question. It will.

A letter of intent records that two parties intend to do business, typically ahead of a binding contract. In a fundraising context it is the artifact a pre-revenue company uses to evidence demand: a prospective customer states an intention to pilot or purchase, a partner an intention to integrate. Most of an LOI is expressly non-binding; only ancillary terms (confidentiality, exclusivity, sometimes a no-shop) bind. A memorandum of understanding is the same idea with even less force.

What separates a strong LOI from theatre is specificity. A useful LOI names a scope, an indicative value, a timeframe, and the conditions that must be met to convert it into a contract (a successful pilot, a milestone reached, budget approved in the next cycle). It is signed by someone who can actually commit spend, not by a friendly contact in a research group with no budget authority. Weak LOIs omit all of this: no number, no date, no named buyer, no conversion path, which is precisely why they can be collected cheaply and why diligence discounts them heavily.

The honest use is as a leading indicator with a stated conversion assumption, not as quasi-revenue. A founder who presents ten LOIs should also present the realistic conversion rate and the gating conditions, because the investor will model it that way regardless. Overweighting LOIs (treating intent as booked demand) is one of the fastest ways to lose credibility in a deep tech diligence, where the gap between intent and a signed, funded contract is wide and well understood.

Most favored nation is a contractual promise of non-discrimination: if the company later issues a convertible instrument with terms more favorable to the investor (a lower cap, a discount, interest), the MFN holder may elect to amend their own instrument to those terms. The clause typically lives in uncapped SAFEs, where it substitutes for pricing: the investor accepts no cap today in exchange for inheriting whatever better deal the market later extracts. Y Combinator publishes a standard capless MFN SAFE for exactly this use.

Mechanically, the right is usually one-shot and expires at conversion: once the SAFE converts at a priced round, the MFN no longer reaches forward. It also reaches only across comparable instruments, later convertibles, not the priced round itself.

For the founder, MFN is invisible until it is expensive. The clause turns the worst-priced convertible in the stack into the effective price of every MFN instrument signed before it. The discipline is bookkeeping: keep a register of every outstanding instrument’s cap, discount and MFN status, and before signing anything cheaper than the stack, recompute the fully diluted outcome as if every MFN holder elects the new terms, because they will.

Milestone-based financing structures a round so the capital is released in tranches, with each tranche conditioned on the company reaching defined milestones, rather than wired in full at closing. The first tranche funds the work toward the next milestone; hitting it unlocks the following tranche, often at a pre-agreed valuation or on pre-agreed terms. It is a way to bridge the gap between an investor who wants to commit and an investor who is not ready to fund the entire plan on trust.

The logic suits deep tech. Investors are pricing de-risking events, so tying money to milestones lets them put more capital behind a company while limiting exposure to the risk that the science does not progress; for the company, a milestone structure can secure a larger total commitment and a better headline than it could raise unconditionally today. Used well, it aligns both sides around the technical proofs that actually move value.

The danger is asymmetry of timing, and it is acute in quantum. Technical milestones slip, sometimes for reasons outside the team’s control, and a tranche gated on a milestone that slips can be frozen precisely when the company needs the cash to keep working toward it, a self-reinforcing trap. The protections a founder negotiates determine whether the structure is fair: the milestone must be defined objectively and be independently verifiable (vague milestones invite disputes about whether the money is owed), there should be cure periods and reasonable tolerance bands, and there should be clarity on what happens if a milestone is partially met or met late. The interaction with signing and closing also matters. Whether a later tranche is a firm obligation or an investor option is itself negotiated: many carry IC re-approval or a material-adverse-change out, so the discipline is to confirm the tranche is genuinely committed, not merely expected, and that the milestone is defined tightly enough that “met” cannot itself be disputed. Milestone financing is a powerful tool for funding a long roadmap in fundable steps, but the value of each tranche to the company depends entirely on how cleanly, and how fairly, the trigger is written.

NISQ, introduced by John Preskill in 2018, stands for Noisy Intermediate-Scale Quantum: in his original framing, machines with roughly fifty to a few hundred physical qubits, a label now stretched to larger devices, too noisy and too few error-corrected to support full quantum error correction. It names the current era of quantum hardware, the gap between today’s devices and the fault-tolerant machines that error correction will eventually enable. NISQ devices can run shallow circuits before noise overwhelms the signal, which bounds what they can usefully do.

The commercial significance is that NISQ sets the near-term ceiling. Whether any economically valuable computation can run on noisy, pre-fault-tolerant hardware is an open and contested question; some applications (certain optimization, simulation and machine-learning heuristics) are explored on NISQ devices, but durable, defensible advantage in the NISQ era has been hard to prove. A company’s exposure to this debate depends entirely on its thesis.

For an investor, the decisive question is which era the business plan lives in. A model that claims value from NISQ-era machines is betting that useful work is possible now, a near-term, testable, but contested claim. A model that only works once fault-tolerant, error-corrected machines exist is betting on a later, more capital-intensive horizon, and should be priced and paced as such. Many pitches quietly assume fault tolerance while showing NISQ hardware, or claim NISQ-era value without a defensible application. Pinning down which era the revenue depends on, and when that era arrives, is how the timeline and the required capital get honestly sized.

Non-dilutive funding is capital that does not require giving up ownership: grants and subsidies, R&D tax credits, repayable advances, soft loans, prizes, and government or corporate co-development contracts. Against equity it has one decisive advantage, founders and existing investors keep their shares, and several costs, the money is often earmarked for specific work, arrives on the funder’s schedule rather than the company’s, and comes with eligibility rules, reporting obligations and sometimes constraints on IP or on how it stacks with other support.

For a deep tech company the strategic role is to fund the science-heavy, pre-revenue years that equity finds expensive to price. A well-run quantum company assembles a stack, a federal tax credit, a provincial or national grant, an agency contribution, on top of equity, so that each equity dollar goes further and the dilution per milestone falls. The non-dilutive layer can be the difference between reaching the next technical milestone on one round or needing a bridge.

The modelling discipline is to separate three states of this money and never blur them: received (in the bank, true runway), committed (signed but not yet paid, a receivable with timing risk), and prospective (applied for or hoped for, not runway at all). Tax credits in particular are reimbursed with a lag, often a year or more after the spend, so the cash and the accrual are different numbers. The strongest plans show the non-dilutive stack explicitly, dated and tiered by certainty, rather than folding optimistic grant income into a single runway figure that diligence will immediately pull apart.

The Industrial Research Assistance Program, run by the National Research Council of Canada, supports small and medium innovative companies with two things: funding and advice. The funding takes the form of non-repayable contributions toward the cost of a defined development project, typically cost-shared (the program covers a portion of eligible salaries and contractor costs, the company covers the rest). The advice comes from an Industrial Technology Advisor assigned to the company, who helps scope projects and connects firms to the wider innovation network.

Unlike SR&ED, IRAP is project-based and forward-looking: it is negotiated and approved before the work, against a specific plan and budget, and paid as the project progresses rather than reimbursed after year-end. That makes it useful for timing, the cash can arrive closer to when it is spent, but it also means the funding is bounded by the approved project and the program’s budget, and is not an entitlement the way a tax credit is. Eligibility centres on being an incorporated, profit-oriented Canadian SME with the capacity to execute.

The interaction with SR&ED is the point most often missed. Because an IRAP contribution is government assistance, it reduces the SR&ED-qualified expenditures on the same costs, lowering the tax credit that would otherwise be earned. The two are complementary but not additive: an operator models them jointly, choosing how to allocate costs so the combined non-dilutive take is maximized, rather than assuming the dollars stack cleanly. Used well, IRAP plus SR&ED plus an equity round is the backbone of the early Canadian deep tech capital stack.

Non-recurring engineering is work a customer pays for once to design, build, adapt or qualify something specific to them. It is common at the hardware and deep tech frontier, where the first deployments require bespoke engineering before any standard product exists. NRE is genuine revenue and genuine validation (someone paid real money), but it has two properties that change how it should be read: it does not repeat by itself, and it often consumes the scarce engineering capacity that would otherwise build the scalable product.

In diligence, NRE is separated from recurring revenue and rarely earns the same multiple. The questions that matter: does each NRE engagement move the company toward a repeatable offering, or does it just pay this quarter’s bills; does the work productize (the bespoke build becomes a reusable module) or does every customer need a fresh from-scratch effort; and critically, who owns the IP created during the NRE. A contract that funds development but assigns the resulting IP to the customer can leave the company poorer in the only asset that matters.

The strategic read is whether NRE is a ladder or a treadmill. Used well, early NRE is customer-funded R&D that de-risks the roadmap and seeds the first product, and the company deliberately reuses what it builds. Used badly, it becomes high-touch consulting with a deep tech logo, growing headcount linearly with revenue and never reaching a scalable product. A credible plan states how much NRE there is, what fraction is productizing, and when the recurring product revenue is expected to overtake it.

The option pool is a block of shares reserved under an equity incentive plan for future grants to employees, advisors and sometimes directors. Grants out of the pool (options, RSUs, or local equivalents) vest over time; the pool itself is counted in the fully diluted share count whether or not it has been granted, which is why its size moves every ownership percentage on the cap table.

The negotiation that matters is the pool shuffle. Investors customarily require the pool to be created or increased before their money converts into shares, in the pre-money. Placed there, the dilution from the new pool falls entirely on existing holders, founders first, while the investor buys into a company already carrying the reserve. The same pool placed post-money would dilute everyone, investor included, which is precisely why term sheets do not write it that way.

Two habits keep the shuffle honest. First, translate any proposed pool into its real cost: pool points demanded pre-money are equivalent to a lower pre-money valuation, and can be negotiated as such. Second, drive the size from a named hiring plan rather than a convention; an oversized pool is not a safety margin, it is dilution warehoused today that a future round will demand be topped up anyway.

A pay-to-play provision requires existing preferred investors to participate in a future financing, typically up to their pro rata share, or suffer a penalty. The classic penalty is conversion of their preferred stock into common (or into a weaker class), which strips the liquidation preference, anti-dilution protection and other rights that came with the preferred. In effect it says: keep backing the company when it raises again, or lose the protections you negotiated. The term may be written into the original financing documents or introduced at the moment of a difficult round.

The provision exists to align investors with the company through hard times. In a down round or a rescue financing, some investors will decline to put in more money; pay-to-play pressures them to participate by making non-participation expensive, which rewards the investors who do step up and concentrates support among the committed. From the company’s perspective it is a tool that keeps insiders engaged precisely when outside capital is hardest to find, and it can make an otherwise unfundable round come together.

For founders the term is double-edged and worth understanding in advance. Where it helps: it can force a supportive insider syndicate to refinance the company and can clean up a cap table by converting passive or absent preferred into common. Where it hurts: it is itself a signal of distress, it can be used aggressively by the investors leading a punitive down round, and the restructuring it triggers (conversions, new senior preferences) can heavily dilute or subordinate those who cannot participate, including earlier backers and sometimes founders. Because quantum’s long, milestone-driven roadmaps raise the odds of at least one tough round, knowing whether pay-to-play is in the documents, and on what terms, is part of reading how durable the cap table will be when the company hits a rough patch.

Pre-money is what the company is valued at before the new investment; the post-money is that figure plus the money coming in. The investor’s ownership is investment divided by post-money, and the price per share is pre-money divided by the fully diluted share count before the new money, including the round’s pool increase and the converting instruments. The two conventions describe the same transaction, but every negotiated number must declare which one it uses.

The classic ambiguities cluster in three places. Headline shorthand (“on 12”) omits the convention. The option pool sits in the pre-money by custom, so a pool increase lowers the effective price per share without touching the headline. And converting instruments blur the boundary: a stack of post-money SAFEs converts into the capitalization before the new money, so the founders’ real post-round position can only be read off a fully diluted table that models the conversions explicitly.

In negotiation, the pre-money is the number that carries the story: what the team, the IP and the milestones reached so far are worth. The post-money is the number that carries the math: dilution, ownership targets, and the next round’s reference point. Fluency means moving between the two without rounding errors, and never letting a deck, a model and a term sheet disagree.

Pre-revenue traction is the bundle of evidence that a market wants what a company is building, gathered before recognized revenue makes the case on its own. For a deep tech company it is the substitute for the ARR chart a software investor would read: paid pilots, design partnerships, letters of intent, government or corporate co-development contracts, a waitlist of credible counterparties, and the conversion of one stage into the next over time.

The discipline is to rank signals by how much skin the counterparty has in the game. A pilot the customer pays for, even a small one, is worth more than a free trial; a free trial outranks a non-binding letter of intent; an LOI outranks a memorandum of understanding; all of them outrank a logo wall of “companies we’ve spoken to”. Slope matters as much as level: three pilots this year against one last year is a trend, a static pile of year-old MOUs is a warning. Honest traction also names its denominators (how many conversations produced how many pilots) rather than showing only the wins.

The pre-revenue trap is treating scientific interest as commercial demand. A national lab eager to co-publish is validation of the technology, not proof that anyone will buy a product; conflating the two inflates the story and collapses under the first diligence question about contract value and timing. The strongest pre-revenue narrative pairs a small amount of real cash-validated demand with a clear statement of what milestone turns that demand into contracts.

A pro rata right entitles an existing investor to participate in a future financing up to the amount that preserves their ownership percentage. It is a right to buy, never an obligation: the investor can exercise in full, in part, or not at all. On priced equity the right is often part of the standard preferred package (sometimes via statutory preemptive rights); on SAFEs and notes it usually arrives as a side letter, since the base instruments carry no such right by default.

The economics cut both ways. For the investor, pro rata is how a fund concentrates capital into its winners; for the company, every reserved allocation is a slice of the next round that the new lead cannot have. Leads price rounds expecting a target ownership; when insider rights consume too much of the raise, something gives: the round grows, the lead’s target drops, or someone waives.

The founder’s job is a clean register: who holds pro rata, on what basis it is calculated (fully diluted is the honest denominator), and what the aggregate claim represents against the next planned round. Waivers are negotiable, and far easier to negotiate before a term sheet than across the table from the lead who just discovered the round is over-allocated.

A pure-play quantum fund is a venture firm whose investment thesis is dedicated to quantum technologies, computing, sensing, communication, components, rather than a generalist or broad deep tech fund that does quantum opportunistically. The category is small but influential: a handful of specialist firms, alongside the quantum-focused sleeves of larger deep tech investors, who have built the in-house technical depth to evaluate the science directly.

Their value to a founder is qualitatively different from generalist money. A specialist can read the physics, stress-test a quantum-advantage or fidelity claim, judge a roadmap’s milestones, and price the company on de-risking rather than on revenue it does not have. That makes them the natural lead and a powerful validator: when a credible quantum fund prices a round, generalist and strategic investors take the signal. They also bring a network of the right hires, customers, and co-investors, and the patience that comes from understanding the timeline.

The limits follow from the small universe. There are not many pure-play funds, their individual fund sizes are often modest relative to the capital deep tech needs, and they cannot fill a large round on their own. Concentration also cuts both ways: a specialist may already be invested in an adjacent or competing approach, raising conflict and signalling questions. The practical play is to use a specialist as the conviction lead and validator, then build the rest of the round with generalist deep tech funds, strategic CVCs and non-dilutive sources, so the company gets both the expert pricing and the balance-sheet depth. Mapping which specialists are active, uncommitted in your modality, and early enough in their fund life to follow on is part of building a credible target list.

Quantum advantage is the threshold at which a quantum computer performs a task beyond the reach of the best classical computers. The stronger phrase “quantum supremacy”, coined by John Preskill in 2012, usually refers to beating classical machines on a benchmark chosen to be hard for them, whether or not the task is useful; “practical” or “commercial” quantum advantage refers to outperforming on a problem with real economic value. The distinction is the whole game: a supremacy demonstration is a scientific milestone, a practical advantage is a business.

For an investor the term is a claim to be tested, because its credibility moves the valuation. The diligence questions are concrete. On what problem, and does anyone pay to solve that problem at that scale. Against which classical baseline, run by whom, with how much classical optimization effort (advantages have repeatedly evaporated when classical algorithms or hardware caught up). Reproduced independently, or asserted in a press release. At what resource cost, and does the advantage survive once error rates and overheads are included.

The honest framing a strong team uses is narrow and dated: advantage on this class of problem, against this baseline, expected at this scale by this time, with these assumptions. Vague claims of imminent, general advantage are the tell of either weak science or weak commercial understanding, and they invite exactly the scrutiny that collapses them. Treating quantum advantage as a roadmap target with conditions, rather than a slogan, is what separates a fundable claim from a red flag.

A right of first refusal gives the company, or its investors, the right to buy shares a holder proposes to sell, on the same terms a third party has offered, before that third party can complete the purchase. It controls who is allowed onto the cap table: a shareholder who receives an outside offer must first present it to the ROFR holders, who can match it and take the shares themselves, or decline and let the sale proceed. It is frequently paired with a co-sale (tag-along) right, which lets investors join a founder’s sale pro rata rather than match it.

The function is governance of ownership. Companies and investors use ROFRs to prevent shares from landing with competitors, unwanted parties, or a fragmented crowd, and to keep some control over secondary transactions while the company is private. For the buyer of last resort it is also an option to increase ownership opportunistically when a holder wants out. On primary issuance the analogous concept is the pro rata right; the ROFR specifically governs transfers of existing shares.

The cost falls on liquidity and speed. A founder or early employee seeking a secondary sale, increasingly relevant when a company stays private for the long horizon deep tech requires, finds the process gated and slowed: the offer must be sourced, presented, and held open for the ROFR period before anything can close, and a willing outside buyer may walk rather than wait or risk being matched out. A ROFR is standard and reasonable in venture financings, but founders should understand its drag on early liquidity and negotiate sensible mechanics (clear notice periods, carve-outs for ordinary transfers, estate planning) so the right protects the cap table without freezing it.

Runway is cash on hand divided by net monthly burn: the time the company has left at its current spending. It is the central planning number of a pre-revenue company, and its honesty depends entirely on the burn figure used (net of reliably incoming cash, with one-offs normalized out) and on which future inflows the model dares to count.

The planning discipline runs on scenarios. A base case with committed cash only; a downside where the grant slips a quarter and the hire happens anyway; an extension case showing which costs could stop. The number that matters in each: where the cash-out date lands relative to the next milestone, because runway that ends one month after a demonstrator is not runway, it is a coin flip on a physics schedule.

The raise timing rule follows directly: the next fundraise starts while runway still covers the full process, twelve months is the comfortable software threshold and fifteen is safer on quantum diligence timelines, less than nine puts the company negotiating under visible pressure, and every investor can read a runway from a burn table as fast as the founder can. Extending runway is also not one lever but three: cut burn, add non-dilutive money, or bridge; each has a cost, and the cheapest is usually the one decided earliest.

A SAFE (Simple Agreement for Future Equity) is a convertible instrument created by Y Combinator in 2013 and rewritten as the post-money version in 2018, which is now the default form. It is not debt: there is no interest, no maturity date and no repayment obligation. The investor wires money today and receives a claim that resolves at the next equity round (shares, priced by a valuation cap, a discount, or both), at a change of control (the greater of the money back or the as-converted value), or at dissolution (the money back, ahead of common).

The post-money mechanics are what founders most often misread. The investor’s ownership equals investment divided by the post-money valuation cap, measured on the company’s capitalization including all converting securities (every other SAFE and note), granted and promised options, and the existing unissued pool, but excluding the new money and any pool increase adopted for the round. That makes each SAFE’s ownership a fixed promise at signature: issuing more SAFEs later does not dilute earlier SAFE holders, it dilutes the founders and every non-SAFE holder on the cap table.

Founders choose SAFEs for speed and simplicity: no valuation negotiation beyond the cap, no board seat, a few pages of standard text, closings that can happen investor by investor. The discipline this convenience removes has to be reintroduced by the founder: keep a running fully diluted cap table where every outstanding SAFE is converted at its cap, and read every new cap as a percentage sold, not as a flattering valuation headline.

Signing and closing are two distinct moments in a financing or acquisition. Signing is when the parties execute the definitive agreements and commit to the deal on agreed terms. Closing is when the conditions in those agreements are satisfied and the transaction actually completes, shares issued, money wired. In many venture rounds the two are simultaneous or close together, but they can be separated by days or weeks, and that gap is where deals still fall apart even after everyone has signed.

The gap exists because the signed agreement usually lists conditions precedent, things that must be true or done before closing is obligatory. Common ones: completion of confirmatory due diligence, delivery of clean IP assignments, required third-party or regulatory consents, board and shareholder approvals, no material adverse change in the business, and sometimes the achievement of a specific milestone. Until those conditions are met (or waived), the money is not owed, and a party can walk if a condition fails. Earlier, the term sheet itself is mostly non-binding, so “signed term sheet” is even further from cash than “signed definitive agreements”.

For a deep tech founder the practical rules are two. First, plan runway to the close, not to the signature: a company that lets cash run to zero on the assumption that a signed deal equals money in the bank can be caught if closing slips or a condition bites. Second, work the conditions down early, especially the ones a quantum company is prone to, the university IP chain and licenses, technical verification, key consents, because they are exactly the items that take time to clear and that a buyer or lead can use to re-trade or delay. Treating the deal as real only when the wire arrives, and managing the conditions actively in between, is what keeps the gap between signing and closing from becoming the place the round dies.

SR&ED (Scientific Research and Experimental Development) is Canada’s federal tax incentive for research and development, the largest single program of its kind in the country. It rewards eligible work, experimental development, applied and basic research, with investment tax credits on qualifying expenditures (salaries, materials, some contractor costs). A Canadian-controlled private corporation earns an enhanced rate that is refundable up to an annual expenditure limit, meaning the credit is paid in cash even when the company owes no tax; other claimants earn a lower, generally non-refundable rate. Most provinces add their own SR&ED-linked credits on top.

The exact rates and limits are set in legislation and have been substantially reformed recently (the enhanced-rate expenditure limit and the phase-out thresholds were raised, and eligibility was extended to some public corporations), so a live claim must be sized against the current year’s rules, not a remembered number. What does not change is the operator reality. The credit is claimed with the corporate tax return after the fiscal year ends, then reviewed, so the cash lands months after the spend, a lag that must be modelled honestly in the runway. The claim quality depends on contemporaneous documentation: what was attempted, what technological uncertainty existed, what experiments were run. Thin documentation invites review and reduces the claim.

Two further points matter for deep tech. First, stacking: other government assistance (an IRAP contribution, certain grants) reduces the SR&ED-eligible base, so the credits interact and cannot simply be added. Second, financing: because the refund is a predictable receivable, specialist lenders advance against it, letting a company convert next year’s refund into this quarter’s runway, at a cost. SR&ED is foundational to the Canadian deep tech capital stack, but it is a reimbursement engine, not a grant that arrives when the bills do.

Stacking (cumul, in French-language programs) is the practice of funding one project from several non-dilutive sources at once. It is how deep tech companies assemble real leverage: a federal tax credit, a national or regional grant, an agency contribution and an equity round can all support the same work. But programs are designed to avoid double-funding the same dollar, so they interact through two main mechanisms that a founder must understand before assuming the money simply adds up.

The first is base reduction. Government assistance received from one program usually reduces the eligible expenditure base of another claimed on the same costs. The classic case is Canadian: an IRAP contribution lowers the SR&ED-qualified expenditures, so the tax credit earned on those costs falls. The programs are complementary but not additive, and the combined take is always less than the arithmetic sum of each taken alone. The second is the aggregate cap. Many grant and contribution programs limit total public funding to a percentage of eligible project cost (a maximum aid intensity), so beyond a ceiling, adding another public source crowds out the others rather than increasing the total.

The operator discipline is to model the stack as a joint optimization, not a sum. That means allocating costs across programs to maximize the combined result, sequencing applications so one does not silently erode another, and respecting each program’s cumul rules and aid-intensity caps. Done well, stacking turns a thin equity round into a fully-funded milestone; done naively, a founder budgets for $550,000 of support, receives $480,000, and discovers the gap exactly when the runway is tightest.

A syndicate is the set of investors who collectively fund a financing round. It has a structure, not just a list: a lead investor who negotiates and prices the round, takes the board seat and anchors it with the largest cheque, and a group of co-investors (followers) who participate on the lead’s terms with lighter diligence. Syndicates form because few investors want, or are able, to fund an entire round alone, and because a mix of investors brings more capital, more networks and more resilience than any single backer.

For deep tech the syndicate is close to a necessity rather than a convenience. Specialist quantum funds, the ones that can price the science, are small, so they lead and validate but cannot fill a large round; generalist deep tech funds add depth; a strategic CVC can add validation and a path to customers or acquisition; and non-dilutive funding sits underneath the equity. Assembling those complementary sources is how a capital-intensive, long-horizon company actually gets funded, and a thoughtfully composed syndicate stacks networks and spreads the risk that any one investor loses appetite over the years the technology needs.

The composition carries signal and risk. A round anchored by a credible lead with committed reserves is durable; a “party round” of many small cheques and no real lead funds the company today but leaves no one accountable for the next hard decision, the bridge, the down round, the follow-on, which is exactly when conviction is scarce. The founder’s job is to build the syndicate deliberately: secure a conviction lead first, then add co-investors whose capital, networks and time horizons complement rather than duplicate, and avoid a cap table so fragmented that no one owns enough to fight for the company when it matters.

A technical milestone is a specific, verifiable achievement that reduces the key technical risk of the business: a fidelity threshold crossed, a coherence time reached, a subsystem integrated and demonstrated, a device qualified. Pre-revenue, it is the atomic unit of value creation. Investors price deep tech not on cash flows but on de-risking events, so a round is implicitly a bet that the team will convert capital into the next milestone, and the milestone is what allows the following round to be raised at a higher valuation.

A milestone is only useful if it is defined to be unfudgeable. That means three things: a target stated as a number or a binary outcome (not “improve coherence” but “T2 above X microseconds on the N-qubit device”); a measurement protocol (how it is tested, in what conditions, over how many runs); and a path to independent verification (a benchmark a third party could reproduce, or data the next investor’s experts can inspect). Vague milestones (“demonstrate progress”) are worthless in diligence because they cannot be passed or failed.

In financing, milestones become the scaffolding. They define what the current round must achieve, they set the trigger points in milestone-based financing or tranches, and they anchor the narrative to the next round: the founder shows what the last round’s capital bought in de-risking, and what the next milestone will prove. A clean milestone map, honest about which are reached, which slipped and why, is one of the strongest signals a pre-revenue team can send, because it shows the company is run against proof rather than against optimism.

The term sheet is the document that turns interest into a deal: a few pages from the lead investor summarizing the terms on which they propose to invest. It is signed before due diligence completes and before definitive documents are drafted, and it is mostly non-binding, a statement of intent rather than a contract, with two customary exceptions that do bind: exclusivity (the no-shop, typically 30 to 60 days) and sometimes expense provisions.

Its content splits into two families. Economics: valuation, round size, option pool, liquidation preference, anti-dilution. Control: board composition, protective provisions (the list of decisions requiring investor consent), information rights, founder vesting. Founders negotiate economics by instinct; experienced counsel earns its fee on control, where a clause costs nothing today and everything at the wrong board meeting.

Mostly non-binding does not mean low stakes. Signing locks the company into exclusivity while diligence runs, so the negotiation leverage peaks the day before signature and, for most companies, decays after. Re-trading, a lead worsening terms after diligence, happens; the defenses are a clean data room, references on the fund’s behavior in past deals, and not burning the runway to the point where walking away is impossible.

Technology Readiness Levels grade a technology’s maturity from TRL 1 (basic principles observed) to TRL 9 (system proven in operation). NASA developed the scale for spaceflight hardware; the European Commission adopted it across Horizon programs, an ISO standard (16290) codified it for space systems, and deeptech investors and grant agencies now use it as a shared shorthand for “how far from the real world is this”.

The bands that matter in practice: TRL 1-3 is research (principles, concepts, first proof of concept); TRL 4-6 is the lab-to-prototype climb where most quantum startups live and most deeptech funding concentrates; TRL 7-9 is demonstration to deployment. Each transition is a derisking event, and the expensive ones are rarely where outsiders expect: in quantum, the integration steps (TRL 5-6), where components that each worked must work together cold, shielded and stable, routinely consume more capital and calendar than the original physics.

Used honestly, TRL is a communication device, not a vanity metric: state the system-level TRL, show the per-subsystem breakdown, and attach evidence to the claimed level (what was demonstrated, in what environment, witnessed how). That one table preempts the most common technical diligence dispute and signals that the team knows the difference between a result and a product.

A university spin-out is a company formed to commercialize research developed in an academic lab. Because the invention was made under the university’s roof and often with public funding, the institution’s technology-transfer office typically controls the resulting intellectual property and decides how it reaches the company: by assignment (the company owns it outright) or, more commonly, by license (the company has rights to use it on negotiated terms while the university retains ownership).

The terms of that transfer are where value is won or lost. The questions that decide whether the company actually controls its core technology: is the grant an assignment or a license; if a license, is it exclusive, worldwide, and broad enough to cover the intended field of use, or narrow and shared; what does the university take in return (upfront fees, running royalties, equity, milestone payments, anti-dilution rights, board observer seats); and are there diligence obligations that let the university claw the rights back if commercialization stalls. Public-funding strings (march-in style rights, government use) can sit on top of all this.

For an investor the spin-out’s IP position is a foundational diligence line, because it determines what the company can defend and sell. A clean assignment, or a broad exclusive worldwide license with reasonable economics, is a real moat. A narrow or non-exclusive license, a university royalty that taxes every future sale, or a transfer that can be revoked, are liabilities that shape the entire investment and sometimes the whole exit. The strongest spin-outs settle the IP chain cleanly and early, because it only gets more expensive to fix once there is value to argue over.

The valuation cap is the maximum company valuation at which a convertible instrument (SAFE or convertible note) turns into shares at the next priced round. If the round values the company above the cap, the investor converts as if the company were worth the cap, ending up with more shares per dollar than the new money. If the round prices below the cap, the investor simply converts at the round price, or at the discounted round price when the instrument also carries a discount and that discounted price beats the cap price (the investor always converts at the lowest of the available prices).

A cap is not a valuation. Nothing has been appraised, no shares have been priced, and the company has not “raised at” the cap, even though founders and journalists routinely speak as if it had. It is a negotiated boundary that compensates the early investor for risk taken before the milestones existed. In practice the market still treats the latest cap as an anchor: the next lead will ask why the Series A pre-money should sit meaningfully above it, and the answer has to be the milestones bought with the SAFE money.

For the founder, the working arithmetic on a post-money cap is one division: ownership sold equals amount raised divided by the cap. Keeping the cumulative result of that division across every outstanding instrument, in fully diluted terms, is the single habit that prevents the classic surprise at the first priced round.

A warrant is a contractual right to buy a company’s shares at a fixed price (the strike or exercise price) within a defined period. It resembles an employee stock option but is granted to investors, lenders or partners rather than to staff. Warrants rarely stand alone; they are usually attached to another instrument as an extra incentive, most commonly to venture debt (a lender takes warrants alongside interest), to bridge financings, or occasionally to commercial or partnership deals as an equity kicker.

The economics are described by warrant coverage: the value of shares the warrants can buy, expressed as a percentage of the associated loan or investment. Ten percent coverage on a $2,000,000 loan means warrants over $200,000 of stock. The strike is often set at the most recent round’s price, so the warrant holder profits if the company’s value rises above that level before the warrants expire. Until exercised, warrants sit as potential shares; exercised, they convert to real equity and dilute existing holders.

For founders the key discipline is to count warrants where they belong, in the fully diluted capitalization, and to track how they accumulate. Venture debt and bridges are useful tools for a capital-intensive quantum company stretching runway to the next milestone, but each debt-flavoured instrument tends to carry warrant coverage, and several of them across a long roadmap add up to meaningful dilution that is easy to overlook because it is not a priced equity round. Negotiating coverage down, capping it, or trimming the warrant term are all levers; ignoring warrants until they are exercised is how a founder is surprised by the fully diluted number at the next round.