Autonomous Drones & Defense
A defense-tech roll-up scaling revenue fast on a large contracted backlog — funded by heavy share issuance, which is the whole argument.
Personal notes published for reference. Figures are sourced from company filings, releases, and press coverage as of the date above — the interpretation is mine and may be wrong.
The hardest of the four to underwrite here. Management is running an aggressive acquisition strategy — ten acquisitions closed in 2026 alone, with an eleventh (Aran Defense) pending — which demands integration skill on top of operating skill. They have consistently raised guidance, which is a point in their favor, but the track record is short.
Drone warfare and counter-UAS, which has gone from speculative to a procurement priority. Management points to a pipeline of roughly $4.3B across 45+ program submissions, and the $982M Army IDIQ validates relevance to long-term US drone procurement. Modern conflict has made this category structurally larger.
About $1.4B in cash and short-term investments with minimal debt as of June 30 ($658M cash, $727M short-term investments, $8M restricted; some data sites quote only the $658M cash line) — genuinely funded, though about $325M went to closing DZYNE and Cyberhawk in Q3 and September's acquisition spree put more to work on top of a ~$50M/quarter EBITDA burn. The company is still posting substantial losses as expenses scale with the business. Management expects EBITDA positivity during 2026; that has not happened yet.
The strongest of the four criteria for this name. 2025 revenue of $50.7M was up 605% YoY. Q2 2026 revenue hit a record $83.8M against a $68.5M consensus, and full-year guidance was raised repeatedly — from $375M to $390M to more than $525M.
The bull case is the order book. Backlog is what turns a story about drone demand into contracted, visible revenue.
$175M of new orders in Q2, then a faster Q3: on Oct 7 Ondas disclosed more than $165M of new orders between mid-August and Oct 1, on top of ~$105M earlier in the quarter. That's over $270M booked since June 30, more than 50% above Q2's pace. (The $56M European order announced Oct 5 appears to be separate, though the release doesn't spell that out.) Backlog has climbed from roughly $457M earlier in the year to $613M reported, and ~$757M pro forma once the Q3 acquisitions are included.
Five acquisitions totaling $557M completed in 2026 — Mistral, World View Enterprises, Omnisys, then DZYNE and Cyberhawk in Q3 — expected to contribute roughly $230M of 2026 revenue. DZYNE alone adds a stated $1.5B pipeline and ~$111M of backlog.
Then it kept going. On Sept 14, 2026, Ondas closed GATE Technologies (Israel) and Bron Technologies (Poland) — electronic safe-and-arm and fuzing tech already embedded in dozens of weapon systems, including ~100 loitering-munition configurations — for $105M cash, ~10.7M shares, a $25M working-capital adjustment, ~$22.5M more due within nine months, and up to $185M in stock earn-outs. Management expects the pair to contribute more than $130M of aggregate adjusted EBITDA through 2028. On Sept 23 it added Insignito (acoustic drone detection), Ottopia Defense (remote vehicle control), and CaribouLabs (GPS-denied navigation) for ~$56M plus up to $32M in earn-outs, and it has agreed to buy Aran Defense for ~$33M. The GATE/Bron piece is notable because it's EBITDA-positive — the first deal that pushes toward profitability instead of away from it.
Three weeks after closing, on Oct 5, Ondas announced a $56M order for electronic safe-and-arm devices for a European loitering-munition program, produced in Poland on Bron's lines. It's the first order since the acquisitions and roughly a quarter of what Ondas paid in cash and stock up front — the cleanest evidence yet that the roll-up is buying revenue, not just backlog slides. The stock rose ~5% to ~$7.63, and Citizens initiated coverage the same day at Market Outperform.
Collaboration with Palantir on AI-enabled defense systems, and the LADOS command-and-control platform debuting alongside a suite of autonomous defense products.
On Oct 12, 2026, Ondas and Palantir announced Regna AI, a new independent company funded by Ondas, which owns it at the start and is its first customer. Its models will orchestrate autonomous missions at the edge, with Palantir supplying model-training infrastructure and delivery help. It starts in intelligence, surveillance, and reconnaissance and is meant to expand into security, critical infrastructure, and industrial markets. No contract terms, outside customers, funding amount, or revenue were disclosed, and reports of the stock's reaction conflict (up ~4% early, down 5–6% in at least one account). Read it as strategic positioning and another call on Ondas cash, not a revenue line.
Gross margin actually declined to 43.1% in Q2 from 49.2% in Q1, not held near 50% as the integration pace might suggest. Adjusted EBITDA loss widened nearly 5x quarter-over-quarter, from $10.9M to $50.6M, as operating expenses scaled with the acquisitions. The stock fell 5.4% on the print despite the revenue beat — the market reacting to cost, not demand.
This is the reason the share price hasn't followed the revenue. The growth is real; so is the cost of funding it.
Over the last year the stock fell 12% ($7.06 to $6.21) while market cap rose 50% ($2.40B to $3.61B), because the share count grew 71% (340M to 582M). Year to date it's worse per share: the stock is down 44%, market cap down 20%, shares up 43%. From the Jan 8 closing high of $14.01 the stock is down 56% and market cap down 37%. Anyone watching market cap alone would think Ondas had a decent year; a holder's return is the price line. In fairness, the new shares weren't free: they funded a $1B raise and the acquisitions behind the backlog, so some of the market cap growth is real capital at work. The test is whether revenue per share outgrows the share count. The count stops at Sept 22 (582.3M) and is held flat after that, so the current market cap is likely a little higher.
Shares outstanding rose roughly 324% in a single year — diluted average share count around 462M as of March 2026. At the May 2026 annual meeting, shareholders approved raising authorized common stock from 800M to 1.2B shares, with over 50% already issued and nearly 25% reserved for higher-strike warrants. A $229M ESOP-related shelf was filed alongside.
So the company both has diluted heavily and has explicitly created room to keep doing it. That's the bear case in one paragraph.
Shares outstanding were 582.3M on Sept 22, 2026 — up about 26% from the ~462M diluted average in March, in six months. The Sept 23 resale registration (7.8M shares from the World View, Insignito, CaribouLabs, and Ottopia deals) brings no cash to Ondas; it just lets sellers exit. Earn-outs across the new deals are payable mostly in stock, so the share count keeps growing even if Ondas never runs another offering.
On cash specifically: the January $1B raise (~$959M net) is what's funding all of this. Q3 cash outlays — DZYNE, Cyberhawk, $105M+ for GATE/Bron, ~$56M for the September trio, ~$33M for Aran — plus a ~$50M/quarter adjusted EBITDA loss are eating into it fast. No new raise has been announced, and there's still likely several hundred million to a billion left, so it isn't urgent. But at this pace of deals, another raise in 2027 is more likely than not unless EBITDA turns positive. With 1.2B authorized and 582M issued, they have the room to do it.
Revenue growth has outpaced share issuance. On a per-share basis the business is still expanding — and the acquisitions bought with that equity are what produced the backlog. Forward EV/revenue has compressed from ~21x to single digits as estimates caught up. Whether that's cheap or a value trap depends entirely on whether the backlog converts.
Roughly 40–44% of the float has been reported sold short — an unusually crowded bet against the dilution thesis, which makes the stock reflexive in both directions.
Two things: (1) does backlog convert to recognized revenue on schedule, and (2) does share count stabilize now that the cash pile is large. If both happen, the per-share math turns quickly. If issuance continues while conversion slips, the bears are right.
The most direct check on the dilution thesis above: are the same institutions who'd have to absorb the new shares actually buying them, or is the float being sold mostly into retail and short-sellers?
Institutional shares held rose roughly 49% quarter-over-quarter — about 105M shares added — with 383 institutions increasing positions against 88 reducing, and 181 new holders initiating versus 125 exiting entirely. That's a lopsided buy/sell ratio for a name this contested.
BlackRock disclosed a 7.2% passive stake in July 2026. The top 5 holders control roughly 31% of disclosed institutional shares, top 10 about 47% — concentrated, but not dominated by any single fund.
Roughly 40–44% of the float is reported short. Institutions net-buying into that setup while the company keeps issuing shares is a real data point for the bulls — it means the money absorbing the dilution isn't just the company's own ATM desk finding retail buyers.
The market's own reaction to Q2 is the cleanest summary: record revenue and a raised guide, and the stock still fell — because the integration bill showed up faster than the backlog conversion.
If this plays out: ~$4
Record $83.8M revenue and a raised guide still sent the stock down 5.4%, because adjusted EBITDA losses nearly quintupled quarter-over-quarter ($10.9M to $50.6M) and gross margin fell from 49.2% to 43.1% in the same three months. That's the roll-up risk showing up exactly where you'd expect — six-plus acquisitions integrating at once while opex outruns even genuinely strong revenue growth. Layer on the existing 40–44% short interest and last year's 324% dilution, and this needs the backlog to convert cleanly and the margin trend to reverse, with little room for another quarter like Q2's cost overrun.
If this plays out: ~$15
Revenue and backlog keep beating and guidance keeps rising ($525–550M full-year now) — the top-line half of the thesis is still working. What the market re-priced is integration cost discipline, not demand. Institutional ownership climbing (56%+ and still rising) even through this margin scare suggests professional money is underwriting the backlog-conversion story more than the quarterly EBITDA print. Expect the stock to keep trading on the gap between revenue beats and margin trend each quarter until one clearly wins.
If this plays out: ~$22.75
If Q2's margin and EBITDA slippage prove to be a one-quarter integration spike — plausible, given five acquisitions closed inside two quarters — then a company guiding to $525–550M in 2026 revenue on a $757M pro forma backlog looks meaningfully mispriced at about $6. Analysts already see it that way: Strong Buy consensus, average target above $19, Ladenburg's $22.75 implying nearly four times from here. With 40–44% of the float short against exactly that conversion thesis, one clean quarter — revenue beat and margin improvement together — could move the stock sharply on the squeeze alone.
Roughly $6.0–6.35 on Oct 15–16, 2026 (about $6.35 at the Oct 15 close and $6.04 on Oct 16; sources differ), market cap ~$3.6B, against a 52-week range of $4.95–$15.28. That's down from $6.73 at the Oct 9 close, ~$7.42 on Oct 5, and $9.51 in early September, so roughly 36% in six weeks, in a market where small-cap and AI-adjacent names are being sold hard. The $4.95 low is only about 18% below here, so I've lowered my bear case from ~$5 to ~$4: the stock had already moved most of the way to the old one. The prior week it fell about 8% despite $270M+ of quarterly orders and the $56M ESAD order. Good order news isn't moving the stock; the market is waiting to see margins and cash burn in Q3 results, expected Nov 12. Short interest is ~41.6% of the float, the highest of any US company above $2B in market cap. Citizens' new Outperform comes with a $14 target. That's against a Strong Buy consensus averaging roughly $19.42–$20.13 (Ladenburg's high at $22.75, raised Aug 17, 2026). That's the widest price-to-target gap on this watchlist — the market is pricing in more execution risk than the sell side is. The scenario targets above are my own rough, directional estimates — not formal price targets — anchored to the sell-side range rather than a full model.
Ten acquisitions closed in 2026, five of them in September alone, across the US, Israel, and Poland, is a lot of moving parts. Roll-ups fail on integration far more often than on demand.
Q2 adjusted loss of $0.19/share missed the $0.10 consensus, gross margin fell to 43.1% from 49.2%, and adjusted EBITDA loss widened nearly 5x to $50.6M. Revenue beats haven't yet translated into the bottom line — if anything, the trend moved the wrong way this quarter.
Revenue is heavily tied to defense procurement timing and budget cycles — lumpy by nature, and politically exposed.