Term Sheets in Deeptech: Clauses David Friedberg-Style Investors Care About

‍

David Friedberg gets invoked in almost every deeptech fundraising conversation I’m in because founders want to know how operators-turned-investors actually assess and structure risk.

‍

If you’re negotiating a deeptech term sheet today, the rules are not the same as a vanilla SaaS round, and the details determine whether you scale or stall.

In this guide, I break down the clauses David Friedberg-style investors care about, why they matter in science-heavy companies, and how to negotiate them without slowing your build.

I’ll cover liquidation preference, IP assignment, milestone tranches, governance, and other venture terms that are specific to deeptech, hard tech, and frontier AI.

I’ll also share field-tested tips I use with founders and investors at Capitaly.vc to close rounds faster and reduce post-close friction.

1) Why Deeptech Term Sheets Are Not SaaS Term Sheets

I’ve raised and reviewed term sheets on both sides of the table, and deeptech deals always bend the template.

In deeptech, the primary risk is technical and time-to-validation, not just go-to-market or CAC/LTV.

That shifts what clauses matter and when cash should arrive.

David Friedberg-style investors want risk matched to capital, not capital dumped into the void.

They prefer structure that pays for proof, not promises.

Here’s how I frame it with founders:

• Milestones replace pure momentum metrics.
• IP position replaces pure code velocity.
• Regulatory and safety covenants replace simple growth KPIs.
• Downside protection matters more because technical timelines slip.

When you accept that, the rest of the term sheet starts making sense.

For more on practical negotiation guardrails, see our blog post: Term Sheet Red Flags Cheat Sheet.

2) How David Friedberg-Style Investors Think About Risk Buckets

I categorize deeptech risk into four buckets when I draft or review terms.

Technical feasibility, unit economics at scale, regulatory clearance, and commercial adoption.

Investors who admire Friedberg’s approach will match capital to the riskiest unresolved bucket.

They do it with staging, protective provisions, and reporting obligations aligned to that risk.

If the big unknown is technical feasibility, tranched funding tied to lab results is rational.

If unit economics is the bottleneck, capex oversight and build-to-budget covenants are rational.

Structure isn’t a lack of trust.

It’s a map to reduce the unknowns at the lowest cost.

I tell founders to label their dominant risk first.

Then invite investors to underwrite that specific risk with bespoke terms instead of blanket constraints.

3) Liquidation Preference Mechanics That Don’t Kill The Company

Liquidation preference is where I’ve seen the most damage done to deeptech companies that later needed large follow-ons.

Investors want downside protection, but toxic stacks scare off future rounds.

The play I like is simple: 1x non-participating preferred, senior to common, pari passu across the same round.

That gives protection without creating unfinanceable overhang.

Participating preferred can be justifiable in hairy tech transfer or late pivots, but it should come with pay-to-play or caps.

Caps limit the participation multiple and keep upside aligned.

Examples I’ve negotiated include 1x non-participating, or 1x participating with a 2x cap when the technical path was more speculative.

Don’t accept stacked seniority across every round unless you must.

Pari passu within the round and clear intercreditor language reduces future friction.

For an explainer you can send to your board, see our blog post: Liquidation Preferences Explained for Founders.

4) Participating Vs. Non-Participating Preferred In Hard Tech

I get asked when participating preferred makes sense in deeptech.

My rule of thumb is to reserve participation for cases with truly asymmetric downside or when new capital is rescuing a failing technical program.

Even then, cap it.

Otherwise, push for non-participating so that founders and early believers share in upside without double-dipping.

David Friedberg-style investors will accept clean non-participating if milestones are clear and reporting is reliable.

If you need to trade, trade governance not economics.

Offer a board observer seat, tighter information rights, or milestone gates in exchange for cleaner liquidation terms.

5) IP Assignment: From Day Zero Or Don’t Raise

IP assignment is existential in deeptech.

If the company doesn’t own or control the core IP, valuation is a guess and follow-on fundraising becomes a slog.

I require assignment of inventions and work-for-hire from every contributor, including visiting scholars and contractors.

University spinouts need an executed license that’s free of poison pills like reach-through royalties that scale with revenue.

Ask for field-of-use breadth sufficient to pivot adjacent applications.

Negotiate sublicensing rights and step-in rights for prosecution.

Include clear obligations for patent maintenance and fee responsibility.

For AI, clarify ownership of models, weights, and derivative datasets.

Add a covenant that all training data licenses allow commercial use and model training.

For more on spinouts, see our blog post: University IP and Spinouts: A Founder’s Guide.

6) University Tech Transfer: The Clauses That Trip Founders

I’ve sat through tech transfer negotiations that killed deals before they started.

The biggest offenders are equity plus stacked royalties plus milestone fees on top of onerous diligence requirements.

Deeptech investors almost always push back on stacked economics that crowd out the cap table.

Reasonable terms include single-digit equity, low single-digit running royalties, and milestones tied to real regulatory or revenue events.

Ensure Bayh-Dole compliance if federal funds touched the IP, but don’t accept march-in paranoia language that spooks corporate partners.

Define practical diligence to avoid default due to unrealistic timelines.

Spell out publication review windows to protect patentability without stifling academic careers.

7) Milestone Tranches That Align With Physics And Cash

Milestone tranches are not a punishment.

They are a way to finance physics at the speed of proof.

I design tranches around unambiguous, binary outcomes.

Think “achieve X output at Y efficiency at Z conditions verified by a third party” rather than “demonstrate progress.”

Add an objective verification method and a 14-day release timeline to avoid starve-the-lab antics.

Include a cure period and a dispute resolution mechanism to prevent weaponized ambiguity.

When milestones are designed well, you de-risk faster and preserve equity.

For a blueprint you can adapt, see our blog post: Milestone-Based Financing Playbook.

8) Technical Diligence And The AI-Native Data Room

Deeptech diligence is won or lost in the data room.

I want to see TRL mapping, experiment logs, calibration protocols, raw data, and replication attempts.

For AI-heavy stacks, include training data provenance, licenses, model cards, evals, and ablation studies.

Make it AI-searchable with structured metadata, short video explainers, and a change log.

The faster I can answer my own questions, the cleaner your term sheet gets.

Investors who respect Friedberg’s operator rigor will reward this.

For a checklist, see our blog post: Deeptech Data Room: What Investors Actually Check.

9) Governance: Boards Built For Science And Scale

Governance in deeptech should optimize for technical truth and capital efficiency.

I push for smaller boards early, with one independent who understands your domain.

Observers can be useful if they bring customers, not just commentary.

Protective provisions should guard against existential changes, not daily operations.

Define a cadence where the board reviews milestone evidence, not just slideware.

Add a Safety and Ethics Committee for bio or dual-use AI to build trust with future investors and partners.

For templates and examples, see our blog post: Governance for AI and Hard Tech Boards.

10) Founder Vesting And Reverse Vesting For Scientists

Deeptech teams often form around a breakthrough, then evolve quickly as the build gets real.

Reverse vesting keeps equity aligned with contribution when roles shift.

I recommend a refresh or a re-vesting event at major funding milestones if the plan outgrows the early team’s bandwidth.

Four-year vesting with a one-year cliff still works, but add double-trigger acceleration for acquisition scenarios.

For academic founders, include leave-of-absence language and conflict-of-interest clearance to prevent later disputes.

This is not anti-founder.

It’s pro-company longevity.

11) Option Pools That Compete For World-Class Talent

You will compete with FAANG and national labs for talent.

Set the option pool for the next 18–24 months of critical hires, not forever.

In deeptech, that often means 12–18% pre-money.

Layer in an EIP (Employee Incentive Plan) that allows milestone-based refreshers for technical inflection points.

Clarify assignment of inventions in your option grant agreements.

Offer relocation or immigration support for specialized hires and bake that cost into your use of proceeds.

12) Anti-Dilution: Weighted Average Beats Full Ratchet

Full ratchet anti-dilution is a growth killer in deeptech because timelines slip and bridge rounds happen.

Push for broad-based weighted average as the default.

If an investor demands stronger protection, add a pay-to-play that forces pro rata in a down round to keep the cap table supportive.

Explain that punitive anti-dilution narrows your future investor universe.

The best deeptech funds know this and will trade for better insight rights instead.

13) Pro Rata And Super Pro Rata Rights

Good deeptech investors want the right to keep supporting you.

Grant standard pro rata to all major investors.

Reserve super pro rata only for those willing to anchor or lead and who bring hard-to-replace value like foundry access, regulatory expertise, or strategic customers.

Cap super pro rata with a clear time window to avoid blocking new leads.

Clarity here prevents last-minute signing drama when the next round comes together.

14) Information Rights And Reporting Cadence That Builds Trust

Information rights aren’t surveillance.

They are how I help you avoid expensive detours.

I like monthly technical memos, quarterly financials, and a milestone tracker with variance analysis.

For AI and bio, include a risk register of safety and compliance items.

Automate what you can from your lab notebook or MLOps stack.

When reporting is reliable, investors relax on some other terms.

15) Regulatory, Export Controls, And Dual-Use Covenants

If your tech touches defense, sensors, advanced materials, or bio, add export control and compliance language early.

Spell out ITAR/EAR classification workstreams, data segregation, and hiring constraints.

Deeptech investors prefer proactive disclosure here because surprises are costly and reputationally risky.

If you operate in healthcare, bake in clinical and quality milestones tied to capital release.

For a primer, see our blog post: Export Controls 101 for Dual-Use Startups.

16) ESG, Biosafety, And AI Safety Clauses That Win Later Rounds

I’ve seen later-round institutions ask for safety covenants that early rounds ignored.

Bake them in now and avoid renegotiation pain.

For bio, include BSL compliance, data handling, disposal protocols, and incident reporting.

For AI, include red-team procedures, model misuse monitoring, and dataset governance with opt-out protocols where relevant.

These covenants are not window dressing.

They are how you scale responsibly and unlock strategic partnerships.

17) Redemption Rights, Pay-To-Play, And Downside Discipline

Redemption rights in deeptech can become a sword of Damocles if timelines stretch.

If they must exist, push the window out 6–7 years and tie redemption to a board-majority approval.

Pay-to-play is a healthier discipline tool because it rewards continued support and penalizes free riders.

I prefer soft pay-to-play in earlier rounds and hard pay-to-play in structured bridges.

It keeps the syndicate aligned when the science requires more cycles.

18) Commercial Rights, ROFR, And Collaboration Clauses

Strategic investors will ask for commercial rights.

Make sure rights of first refusal (ROFR) and most-favored-nation (MFN) clauses don’t block future customers.

Limit field-of-use and set reasonable response timelines.

Convert perpetual rights into term-limited options with pre-agreed pricing bands.

If the strategic is a must-have partner, stage rights based on delivered value, not promises.

19) Secondary Sales And Founder Liquidity Without Handcuffs

Small secondaries can keep founders sane in deeptech because paydays take longer.

I’m fine with modest liquidity once key milestones are hit and the company is not cash-starved.

Hard caps, board approval, and right-of-first-offer (ROFO) mechanics prevent distortions.

Align the secondary with long-term incentives like refreshed vesting.

For more guidance, see our blog post: Founder Liquidity Without Losing Control.

20) Delayed Closings, SPVs, And Syndicate Engineering

Deeptech rounds often require complex syndicates because you need both capital and capability.

Use SPVs to bundle strategic checks without bloating the cap table.

Delayed closings can align grant timing or customer contracts with capital calls.

Spell out conditions to close, drop-dead dates, and what happens if specific strategic investors drop out.

Clarity wins here because operations cannot stall while lawyers negotiate logistics.

21) Drag-Along, Co-Sale, And Keeping M&A Optionality

M&A optionality matters more in deeptech because IPO windows are lumpy.

Clean drag-along with fair market value and minority protection ensures you can accept a good offer.

Co-sale rights give early supporters liquidity alongside founders during secondary transactions.

Keep thresholds practical so a single holdout can’t block a value-creating outcome.

22) Protective Provisions That Actually Protect

Not every decision needs investor consent.

Reserve protective provisions for actions that change the nature of the company.

Examples include issuing senior securities, changing charter, selling IP crown jewels, or taking on debt above a threshold.

Operational minutiae should remain with management.

For a punch list you can copy, see our blog post: Protective Provisions That Matter.

23) Reps, Warranties, And Scientific Integrity

Representations and warranties in deeptech should include scientific integrity statements.

State that data and results presented are accurate to the best of knowledge and all material negative results are disclosed.

Include IP non-infringement reps anchored in a documented FTO review for key claims.

For AI, add reps about data license scope and consent where applicable.

These reps reduce post-close surprises and speed legal review.

24) Data Rights, Open Source, And AI Licensing Pitfalls

Data is your differentiator, so get the terms right.

Prohibit training on data that forbids commercial use or derivative model creation.

Track all third-party datasets, licenses, and expiration terms.

For open source, avoid copyleft contagion in core IP layers.

If OSS is essential, ring-fence it and document compliance.

For a deeper dive, see our blog post: Data Licensing for AI Startups.

25) Manufacturing, Foundries, And Capex Covenants

Scaling deeptech often means capex before revenue.

I add covenants that tie capex drawdown to vendor bids, site permits, and pilot success thresholds.

This keeps you from overbuilding the wrong thing.

Vendor step-in rights and escrowed prepayments can protect both sides.

If you’re a foundry-dependent startup, include supply assurance and IP protection in your vendor contracts and cross-reference them in the term sheet.

26) Grants, SBIR/STTR, And Matching Clauses

Non-dilutive funding is gold in deeptech when used correctly.

Some term sheets include matching requirements or stipulate how grant funds interact with milestones.

I prefer to treat grants as bonus oxygen and not make the company hostage to unpredictable government timelines.

Disclose all grant conditions, IP obligations, and reporting so investors don’t get blindsided.

27) Convertible Notes, SAFEs, And When To Price The Round

In deeptech, I price the round earlier than in SaaS when patents, prototypes, or clinical data create a clear step change.

Notes and SAFEs work for fast bridges, but too many layers create a cap table Jenga tower.

Consolidate into a priced round when you raise institutional capital.

Watch for MFN clauses in older notes that could unintentionally import unfavorable terms.

28) Valuation, Caps, And Narrative Math

Valuation is narrative math pinned to proof.

Anchor your ask to the cost and time to reach the next objective inflection point with a credible budget.

David Friedberg-style investors respond to an instrumented plan more than comps that don’t exist.

If a cap is necessary in a SAFE, set it to reward early believers but avoid a step-up that traps the next round.

Make the plan investable, not just impressive.

29) International Subsidiaries, CFIUS, And Corporate Structure

Deeptech with defense or sensitive tech triggers CFIUS review risk if you have non-US investors or foreign subs.

Structure early to keep options open.

That can mean US TopCo, clean data paths, and governance that limits foreign control rights.

Spell these expectations out so you don’t need a painful restructure later.

30) Closing Mechanics, Conditions Precedent, And Post-Close Checklists

Deals die on logistics more than on price.

List conditions precedent clearly: IP assignment complete, key hires signed, insurance bound, and any mandatory facility permits in process.

Create a post-close 90-day plan with who does what by when.

Wire schedules, tranche release triggers, and reporting templates should be in the data room before signatures.

This is how you hit the ground running, not haggling.

FAQs

What is a fair liquidation preference in deeptech?

1x non-participating preferred, pari passu within the round, is the default I push for because it protects downside without scaring off future rounds.

When should I accept participating preferred?

Only when the deal is rescuing a failing program or the risk profile is extreme, and even then cap participation and trade for milestone-based releases.

How do milestone tranches actually work?

Define binary outcomes, objective verification, a short release window, and a cure period to avoid weaponized ambiguity.

What IP clauses matter most in a deeptech term sheet?

Company ownership of core IP, clean university licenses, assignment of inventions from all contributors, and data/model ownership clarity for AI.

How big should the option pool be?

Usually 12–18% pre-money for the next 18–24 months of critical hires, with milestone-based refreshers for technical inflections.

What anti-dilution protection is reasonable?

Broad-based weighted average is fair and keeps future rounds viable, especially when bridges are common in deeptech.

Should I give a strategic investor commercial rights?

Yes, but limit field-of-use, set response timelines, and convert perpetual rights into time-bound options with predefined pricing.

How much founder secondary is okay?

Modest secondary after hitting key milestones and with board approval is fine, especially in long-duration deeptech.

Do I need special regulatory covenants?

Yes if you touch bio, defense, or healthcare, including export control processes and clinical quality systems.

When should I move from SAFEs to a priced round?

When you have institutional interest and a technical validation that justifies a coherent price, to avoid a messy cap table later.

How do I handle open source in a deeptech company?

Use permissive licenses in non-core areas, avoid copyleft in core IP, and document compliance and provenance thoroughly.

What goes into a deeptech data room?

TRL maps, experiment logs, raw data, FTO analysis, regulatory plans, and for AI, training data licenses and evals.

If you take nothing else from this, take this: deeptech term sheets work when risk, capital, and proof are aligned.

That’s the core of how David Friedberg-style investors operate, and it’s how I’ve seen deeptech companies survive the hard parts and scale the right ones.

Design clean liquidation preference, insist on real IP assignment, stage capital with milestone tranches that reflect physics, and build governance that prefers truth over theater.

If you do that, you won’t just close this round.

You’ll make the next round easier.

For more help, my team at Capitaly.vc publishes playbooks and checklists to move faster with fewer surprises.

Subscribe to Capitaly.vc Substack (https://capitaly.substack.com/) to raise capital at the speed of AI.

And when in doubt, remember the core keyword of this piece and the mindset behind it: Term Sheets in Deeptech and what David Friedberg-style investors really care about.