DOE Just Put a 700-Word Gate in Front of Its SBIR Program. The Genesis Mission Phase I Pitch Closes September 10.

The Department of Energy has opened the first Phase I competition under the Genesis Mission, its AI-for-science initiative, and the money is not the interesting part. DOE intends to make approximately 40 Phase I awards totaling $10 million — up to $250,000 each, or $256,500 if you request Technical and Business Assistance funds — representing up to $44 million of cumulative SBIR/STTR funding across the lifecycle of those projects. Pitches are due Thursday, September 10, 2026, at 2:00 p.m. ET.

The interesting part is that you can no longer apply.

Not directly. Under the redesigned program, run by DOE's new Office of Technology Commercialization in partnership with the Partnership Intermediary ConnectWerx, applicants submit a short structured pitch first. Only companies whose pitches are accepted receive an invitation to submit a full application. The pitch is capped at roughly 700 words across four open-text questions. That is it. That is the entire artifact standing between a small business and the DOE SBIR pipeline.

For a program that has historically asked for 20-page technical volumes reviewed by subject-matter panels, this is not a tweak. It is a different sorting mechanism with different failure modes, and companies that plan their year around DOE's traditional release calendar are about to run into it at speed.

What is actually on offer

Four topic areas, all of them AI applied to a DOE science mission:

1. Scaling the Biotechnology Revolution — AI for biochemicals, bioproducts, genome engineering, and biomanufacturing.
2. Realizing Quantum Systems for Discovery — AI for quantum computing and networking, including error correction, decoherence mitigation, and quantum resource optimization.
3. Designing Materials with Predictable Functionality — AI-enabled inverse materials design.
4. Achieving AI-Driven Autonomous Laboratories — robotics and experimental automation in the loop with models.

Phase I awards run six to twelve months and are meant to establish technical feasibility plus a commercialization plan. Progression is governed by milestone-based go/no-go decisions rather than a fresh open competition, which is the quiet upside of the redesign: a subset of Phase I winners are advanced to Phase II on performance, not on their ability to survive a second full proposal cycle.

Companies may submit up to three pitches. Registration requirements are the familiar ones plus a new one: SAM.gov, the SBA Company Registry, and an account on the DOE SBIR/STTR Application Hub at sbir-sttr.connectwerx.org. If you have only ever used DOE's old eXCHANGE portal, budget real time for the new hub — portal registration failures are the single most common cause of missed federal deadlines, and this portal is weeks old.

The two eligibility walls nobody is talking about

Read the fine print and two doors close.

First: there is no Direct to Phase II pathway for new applicants. If your company has never held a DOE SBIR or STTR award, you cannot skip the feasibility stage no matter how mature your technology is. Companies that built a strategy around entering federal SBIR at Phase II scale — common for firms with commercial revenue and a working product — have to run the $250,000 Phase I lap first.

Second: the $147.4 million Phase II competition, deadline September 25, 2026, is restricted to prior DOE awardees. That is where the real money in this cycle sits, and it is a closed room. The practical effect is a two-tier program: incumbents compete for $147 million; everyone else competes for $10 million and the right to become an incumbent.

This is worth stating plainly because the aggregate headline — "DOE announces $157 million in SBIR funding" — obscures a 93-to-7 split by prior-relationship status. If you are new to DOE, your addressable pool this cycle is $10 million and roughly 40 slots.

Why the pitch format changes what you should write

A 700-word pitch is not a shortened proposal. It is a different genre, and the most common mistake will be compressing a technical volume into it.

The four questions are approximately equally weighted:

• Summary and mission alignment (≤100 words)
• Technical promise (≤200 words)
• Commercialization potential (≤200 words)
• Team qualifications (≤200 words)

Note the ratio. Commercialization gets exactly as much room as the science. In a conventional DOE SBIR review, a technically brilliant proposal with a thin market section still scores respectably because the technical volume dominates the page budget. Here, a 200-word commercialization answer carries a quarter of the decision. Deep-tech founders who instinctively spend their best words on the physics will be outscored by teams that write 200 disciplined words about who buys this, at what price, and through what channel.

Three tactical implications:

Lead with the mission-fit sentence, not the background. With 100 words for summary and alignment, there is no room for a paragraph explaining why AI matters. Name the topic area, name the specific DOE bottleneck you attack, and name the output. Reviewers screening dozens of pitches will decide in the first sentence whether you belong in the topic at all.

Make the technical claim falsifiable. In 200 words you cannot prove feasibility, so do not try. State the specific mechanism, the specific baseline you beat, and the specific number that would demonstrate success within a 6-to-12-month Phase I. "We will improve quantum error correction using machine learning" is invisible. "Our decoder cuts logical error rate by 3x at code distance 7 against the standard MWPM baseline, validated on published surface-code datasets in month four" is a proposal.

Treat team as evidence, not credentials. 200 words on qualifications is enough for three specific proof points — a prior deployment, a named lab partnership, a relevant prior award — and nothing else. Lists of degrees consume the budget without moving the score.

Who should actually pitch

Strong fits: small businesses with an existing model, dataset, or automation stack that plugs into one of the four topics; companies with a national-lab collaboration already in motion; teams whose AI work has a physical-science validation loop rather than a pure-software story.

Weak fits: general-purpose AI tooling with a DOE use case bolted on. The Genesis Mission topics are written around scientific bottlenecks — inverse design, decoherence, autonomous experimentation — and a horizontal platform that could serve any vertical will read as unaligned in a 100-word alignment answer.

If you are university-affiliated, the STTR track lets you formalize the research partner relationship, but remember the small business must still lead and hold the majority of the work.

The calendar

• Now through September 10, 2:00 p.m. ET — pitch window. Register on all three systems immediately; SAM.gov renewals alone can take weeks.
• September 25, 2026 — Phase II deadline, prior DOE awardees only.
• Fall 2026 — DOE has signaled a second Phase I release. If you miss September 10, that is your next entry point, and the pitch you write now is reusable.

The broader pattern matters too. DOE is the second major agency this year to move toward a fast, low-burden front gate — a shift we traced across the research portfolio in our analysis of NSF's move to rolling submission across 12 foundational research NOFOs, and one that sits alongside the $5 billion, 278-project Genesis Mission buildout already underway. Lower submission burden sounds like a gift to small applicants. In practice it raises application volume, which raises the screening ratio, which makes the first 100 words of your pitch the most valuable words your company will write this quarter.

Write them last, after the other three answers are finished. And write them for a reader who has 40 slots and several hundred pitches.