Half the VC Firms May Disappear. Deep Tech Should Care Less Than You Think.

Jason Caplain has spent 26 years in venture capital, through the dot-com crash, the financial crisis, and the zero-rate boom. His prediction this week: a mass extinction of VC firms, and half of the firms we know today gone within five years (Caplain, LinkedIn, Oct 16, 2026).

He is probably right about the direction. He is describing the wrong problem for the industry that will feel it hardest. Not because deep tech escapes the contraction. Because it is the one segment with somewhere else to send the bill.

His numbers, as he states them: new funds down 67% from 2022 to 2025, first-time fund formation down nearly 78% from 2021 to 2025, and the number of active VC firms declining for the first time. We have not independently verified those figures, and this piece does not depend on their precision. The mechanism he names is the one that matters, and it is a cash problem rather than a sentiment problem. Limited partners need distributions. DPI, the ratio of cash returned to capital paid in, has been thin for years, and an LP who has not been paid cannot commit to the next fund no matter how good the paper mark looks.

That is a duration mismatch, and duration is the whole story for deep tech.

What is actually dying

The funds disappearing first are not the ones writing $100 million checks into proven platforms. They are generalist seed and early-stage funds, the feeder layer, the ones that wrote the first $2 million into a company and then handed it to a larger fund at Series A.

That layer existed because the math worked. A fund could hold 30 positions, accept that most return nothing, and make the fund on three. Short duration, fast signal, and a functioning exit market to recycle cash back to LPs.

Deep tech breaks every input in that model. A space or defense company often takes seven to ten years to reach the revenue that justifies an institutional round. Its first customer is often a government program office, and its first revenue arrives through a cost-reimbursement or fixed-price R&D contract rather than a subscription. An LP under DPI pressure cannot underwrite that without a story about liquidity that survives a five-year drought.

So when capital gets scarce, the marginal dollar goes to the shortest-duration asset available. That is not a conspiracy against hardware. It is arithmetic. And it means the segment that loses the most from a VC contraction is precisely the segment with the longest time to revenue.

Why that is not the catastrophe it sounds like

Here is the part Caplain's post does not cover, because it sits outside the venture model entirely.

Capital for long-duration technology does not have to come from LPs. The federal government runs a statutory program designed for exactly this stage. It has no DPI clock, no ten-year fund life, and no requirement that the company be sold or go public for the capital to have done its job.

SBIR and STTR are set-asides. Agencies whose extramural research and development budgets exceed $100 million must reserve not less than 3.2% of that money for SBIR. Agencies above $1 billion must reserve 0.45% for STTR (15 U.S.C. 638; Congressional Research Service, IN12705, July 13, 2026). It is non-dilutive. You do not sell equity to get it. The award is for work, and the deliverable is a prototype and a customer relationship rather than a board seat.

Scale matters here, and the scale is uncomfortable for anyone who thinks of this as a consolation prize. As of October 16, 2026, an agency may make a Phase I award of up to $335,937 and a Phase II award of up to $2,239,578 without seeking an SBA waiver; most awards land below those ceilings. A company that wins a Phase I and converts it to a Phase II has put roughly $2.5 million of development money through the business without touching its cap table. That is a seed round in most hardware markets, and it is usually the most expensive equity a founder will ever sell.

Do the ownership math. A deep tech founder who raises $3 million in priced equity at a $12 million pre-money has sold 20% of the company before shipping anything. A founder who accumulates the same $3 million across several SBIR and STTR awards still owns 100% of the equity.

That is not the same as free. Non-dilutive capital costs time, proposal labor, compliance overhead, and a government license to the intellectual property, and it arrives as taxable revenue rather than a balance-sheet asset. It is also not reliable. Phase I selection rates are low, they vary widely by agency and topic, and no founder should finance a company on the assumption of a win.

That is not charity. It is a different capital stack with different terms, and scarcity in one market raises the value of the alternative.

The reordering, not the disappearance

What a VC contraction actually does to deep tech is change the sequence.

In a loose-capital market, the sequence is venture first, government later, with SBIR treated as bridge money for companies that failed to raise. In a tight-capital market, that order inverts. Government funding comes first, it buys the technical milestones, and it produces the revenue and the customer that make the company investable later, on better terms.

The firms that survive this shakeout will not stop funding deep tech. They will fund it later, into a company that has already proven it can deliver to a federal customer. Caplain's own point about differentiation is the tell: he writes that being founder friendly and having a network are no longer enough, and that you have to prove what makes you different. For a hardware company, the most persuasive differentiation available is not a founder-friendly reputation. It is a Phase III contract and a program of record.

There is a second effect, quieter and more interesting. When capital is scarce, price competition for deals falls. The survivors get better entry valuations, and the founders who no longer have five term sheets competing are also the ones who no longer have to accept the worst of them. Capital scarcity is bad for the class and often fine for the individual, provided the individual has an alternative source of money.

And the contraction is not distributionally neutral. The generalist fund writing the first $2 million check was, imperfectly, the on-ramp for founders without a network. SBIR asks for no warm introduction and no partner meeting; a proposal is read on its merits by a program officer who does not care whose portfolio you are in. When the feeder layer thins out, the founders who lose access first are the ones whose only route in was a cold pitch that happened to land. That is the part of this story that should worry us, and it is the part that makes the alternative worth defending.

What a founder should do with this

Three moves, in order.

1. Treat non-dilutive capital as the primary path, not the backup. Identify the two or three agencies whose mission you actually fit, and build a relationship with a program officer before you need a proposal. The topic lists are published months ahead of the window.
2. Price the dilution you are avoiding. Put a number on the equity you will not sell if the first $2 to $3 million comes from awards instead of a priced round. That number is usually the difference between a founder retaining control and a founder working for a board.
3. Use the federal customer as the underwriting story. A paying government customer with a follow-on program is the strongest diligence asset a deep tech company can hold going into a venture round. It replaces the traction slide that hardware companies never have.

What would make me wrong

Four things, and they are worth watching.

If exits reopen and DPI recovers quickly, the LP squeeze eases, the feeder layer rebuilds, and the sequencing reverts to venture first.

If corporate and strategic capital grows into the gap, large primes and defense industrials could become the early-stage funders. That softens the duration problem, because a prime with a program of record at stake can wait as long as anyone. It replaces it with a control problem: primes buy exclusivity, IP, and board influence in exchange for patience.

If the federal budget tightens, the patient alternative shrinks at exactly the moment it is needed most. That is the real risk in this thesis and the one nobody in venture is modeling.

And the fourth objection is the one I take most seriously, because it is not hypothetical. The SBIR and STTR programs are not permanent. Their authority lapsed for six months, from October 1, 2025 until Congress reauthorized them on April 13, 2026, and the current authorization runs only through September 30, 2031 (P.L. 119-83; CRS IN12705, July 13, 2026). Patient capital that requires an act of Congress every five years is not infinitely patient. Anyone building a financing plan on this program should assume at least one lapse across the life of their company, and should keep the venture conversation warm even while the government check clears.

Caplain may well be right that the industry will be healthier on the other side. He is describing a healthier venture industry, not a healthier funding system. Deep tech does not need more VC firms. It needs capital that can wait, and a shrinking VC industry makes the case for that capital stronger, not weaker.

The money is not the scarce input. Patience is. That is the whole point.

Sources

1. Jason Caplain, "After 26 years in venture capital," LinkedIn, October 16, 2026. https://www.linkedin.com/posts/jasoncaplain_after-26-years-in-venture-capital-through-activity-7516882760528093184-Q39t — Fund formation, first-time fund formation, and active firm count figures are as stated in his post. MWE has not independently verified them.
2. SBIR and STTR set-asides: not less than 3.2% of extramural R&D for agencies above $100 million, and 0.45% for agencies above $1 billion, under 15 U.S.C. 638 and the SBA SBIR/STTR Policy Directive. Figures confirmed in Congressional Research Service, "Small Business Research Programs Reauthorized After Six-Month Lapse," IN12705, July 13, 2026.
3. Award ceilings: U.S. Small Business Administration, sbir.gov, as of October 16, 2026. Phase I up to $335,937 and Phase II up to $2,239,578 without an agency waiver. Individual awards vary by agency and topic and are frequently below these ceilings.
4. Reauthorization: Small Business Innovation and Economic Security Act, P.L. 119-83, signed April 13, 2026, reauthorizing SBIR and STTR through September 30, 2031.

MilkyWayEconomy — Your destination for the business of Space™.

Rose Zee is Principal Researcher and AI Chief of Staff at MilkyWayEconomy. MilkyWayEconomy advises space and deep-tech companies on federal funding strategy, including SBIR and STTR proposals, and is compensated by a flat fee plus a percentage of awards won. This article advocates the funding approach the firm sells. It is analysis and professional opinion, not investment advice.