The key race: a widening cast proves software-defined mass at sea and in the air, while the funding cliff and the autonomy rulebook stay unresolved
The clearest news of the past week is not a policy fight, it is what three newly tracked companies just did with hardware. On July 12, three of Saronic's 24-foot Corsair unmanned surface vessels struck a docked Iranian submarine and a ship-maintenance facility at Bandar Abbas under US Central Command, the first combat use of American sea drones and a maritime echo of Anduril's own first live air-to-air weapons shot two days earlier. Shield AI's X-BAT vertical-takeoff combat jet lit off its F110 engine and thrust-vectoring nozzle in July, cleared the path to a targeted first vertical flight by the end of 2026, and picked up a fresh $50 million Navy and Defense Innovation Unit investment on August 13. General Atomics Aeronautical Systems, whose FQ-42A Collaborative Combat Aircraft was ordered into production alongside Anduril's FQ-44A in June, is running its production line toward six aircraft a month. None of that changes the two Washington deadlines hanging over the whole sector: Golden Dome director Gen. Michael Guetlein's warning that the program could grind to a halt when fiscal 2027 begins on October 1 because roughly 97 percent of its $17.5 billion request depends on a reconciliation bill Congress has not passed, and the still-unissued rewrite of DoD Directive 3000.09, the policy governing how much autonomy a weapon may exercise, due around September 3 under National Security Presidential Memorandum 11. The FY2027 NDAA that would authorize the rest of the sector's weapons programs remains stuck behind both: the House passed its version 216 to 212 in July, the Senate has not moved past a failed cloture vote, and no conference or floor date has been scheduled as of mid-August.
The autonomy-policy deadline picked up its first real congressional pushback this cycle. Senator Ruben Gallego sent Defense Secretary Pete Hegseth a letter questioning what safeguards the Pentagon is building into a 90-day rewrite of a document that normally takes years to clear multiple offices with law-of-war equities, directly naming the risk to Americans and allies if the revised directive loosens autonomy rules faster than testing can validate them. That tension, a policy rulebook racing to keep pace with hardware that is now shooting live weapons and striking real targets, is the throughline connecting Golden Dome's funding risk, the CCA program's expanding roster, and Saronic's own combat debut.
Who's ahead: Anduril and Palantir's NGC2 roles get corrected, three new entrants graduate onto the roster, and Helsing closes in on a Japanese order
The Army's Next Generation Command and Control program clarifies a division of labor this brief had blurred. The Army named Anduril lead integrator for NGC2's common data layer baseline in June, with Anduril's Lattice serving as the tactical edge layer and Palantir's Foundry contributing the cloud data mesh alongside partner Raft; fielding of the full stack to the 4th and 25th Infantry Divisions is funded at $3.78 billion in the Army's FY2027 request, with all 11 divisions targeted within five years. Palantir's other Pentagon relationship keeps compounding on paper, a record $1.94 billion Q2 revenue up 93 percent year over year, and on August 5 the Pentagon appointed Army Col. Molly Solsbury as Maven Smart System program director, a concrete step toward the Maven Smart System's September 30 deadline to become a formal program of record. Anduril, meanwhile, added a satellite to its own resume: ANDURIL-216 returned first calibrated images from all three of its sensors on August 17, less than two years after clean-sheet design, commanded on orbit for the first time using Lattice software. Its reported talks to raise fresh capital at roughly a $100 billion valuation, first reported July 24, remain unresolved as of mid-August, with Anduril saying no decision has been made.
Three companies this brief previously treated as sector context are now tracked on their own pages, and each sharpens a different part of the sector's structure. Saronic followed its Bandar Abbas combat debut by opening a dedicated on-water testing site at the Port of Gulfport, Mississippi in August, logging hundreds of hours of trials on its first 180-foot Marauder hull ahead of an at-sea Navy MUSV Marketplace demonstration running to October 2026; the company is separately building the $3 billion-plus Port Alpha shipyard in Brownsville, Texas. Shield AI and General Atomics Aeronautical Systems sit on opposite sides of the same program: GA-ASI builds the FQ-42A airframe under a production contract awarded alongside Anduril's FQ-44A in June, while Shield AI, Anduril and RTX Collins Aerospace compete in a funded, six-month mission-autonomy phase for the software that will fly on both, with a primary-provider down-select due by summer 2027 and Collins' rival Sidekick software already flying on the YFQ-42A. Shield AI carries its own scrutiny alongside that opportunity, a Reuters investigation into V-BAT crash rates published in June, even as its valuation sits at $12.7 billion. Helsing's push into Japan advanced a step further: the company enlisted e-commerce and finance firm Rakuten in August to help broker a full sales deal for its HX-2 strike drone with Japan's Ground Self-Defense Force, which is field-testing the system through the end of September.
Europe keeps building its own alternative: Helsing's KIRK constellation, NATO's spending math, and a Japan deal that would be its first Asian order
The distrust of American software that is costing Palantir contracts in Germany and France keeps funding the continent's own champions. Munich-based Helsing's $1.8 billion Series E in July, at an $18 billion valuation, Europe's largest-ever defense-startup round, is converting into hardware ambitions beyond its fielded HX-2 drones: the Kongsberg-Hensoldt-Helsing partnership formed in December 2025, joined by OHB in May under the 'KIRK' joint venture, is combining OHB's satellite systems, Hensoldt's space-qualified sensors and Kongsberg's small-satellite and ground-station expertise into a sovereign European ISR and targeting constellation targeted for 2029, with Isar Aerospace named preferred launch partner. A converted Japan deal would be a first for Helsing outside Europe and a direct data point in the same US-versus-Europe software contest playing out in NGC2 and Maven.
The money behind Europe's shift keeps scaling to match. NATO members, Spain exempted, committed at the 2025 Hague summit to raise defense and security spending to 5 percent of GDP by 2035, and European Allies and Canada already raised combined defense spending by nearly 20 percent in real terms in 2025 alone. That is a large enough pool that European challengers like Helsing, and the sovereign platforms France and Germany are now formally underwriting, can credibly compete for a growing share of it instead of ceding the continent's own procurement to American suppliers by default.
What decides it: whether an autonomy rulebook can keep pace with weapons that are already shooting, and whether mass production actually scales
Saronic's Bandar Abbas strike and Anduril's live AMRAAM shot both happened before any rewritten autonomy policy exists to govern the next generation of systems behind them, Shield AI's Hivemind-flown X-BAT and General Atomics' CCA fleet among them. That is exactly the gap Senator Gallego's letter to Hegseth is pointing at: DoD Directive 3000.09 was last substantially updated in 2023, a 90-day rewrite deadline lands around September 3, and the systems it needs to govern are no longer prototypes, they are fielded weapons with combat records. Whether the rewritten directive tightens or loosens the autonomy latitude those systems already operate under, and whether it does so before or after more of them see combat, is now a live policy question rather than an abstract one.
Ukraine's manufacturing base remains the sector's clearest real-world proof that procurement structure, not just technology, decides who can field weapons at scale. The country's domestic drone-manufacturer base grew from seven companies in 2022 to more than 500 by 2026, and Ukrainian officials describe current output around 8 to 10 million drones a year with a stated ambition of 20 million if partner nations commit enough capacity. The European Union's July 2026 drone-production deal with Ukraine, backed by a 1 billion euro pledge inside a larger 90 billion euro support package, is an explicit bet that Ukraine's decentralized, unit-level procurement model is the template American and European suppliers need to match if mass is going to matter as much as capability in the next conflict, a lesson General Atomics' six-a-month CCA line and Saronic's multi-site USV production are both being built to answer.
The money and the rules: a record funding year keeps adding new platform companies, and three deadlines still close within weeks of each other
Defense-tech venture funding remains on a record pace, more than $14.6 billion raised in the first five months of 2026 alone, already ahead of 2025's full prior-year record, and the roster of companies absorbing it keeps widening: Shield AI's $12.7 billion valuation and Saronic's $9.25 billion valuation sit alongside Anduril's own reported $100 billion talks, while newcomer Heaviside Industries raised a $60 million Series B at a $600 million valuation in August, just three months out of stealth, and partnered with Norway's Nammo on precision-munitions warheads, a sign the funding boom is still minting new entrants, not just re-rating the largest ones. Defense primes are participating directly too, with BAE Systems, Lockheed Martin and Airbus together putting a record $4.1 billion into startup rounds this year. Set against that private-capital boom is a public-spending story that increasingly bypasses competition altogether: an August memo directing up to $243.9 million in no-bid Pentagon spending on Palantir services is the most visible recent example, part of the same pattern as Anduril's sole-source Pulsar-L jammer notice and a run of Golden Dome and counter-drone awards moving through urgency and sole-source authorities rather than open bidding.
Three deadlines still land within roughly four weeks of each other, and each will say something different about how much slack the sector's growth actually has: the NSPM-11-mandated rewrite of DoD Directive 3000.09 due around September 3, the Maven Smart System's program-of-record deadline on September 30, and the start of fiscal 2027 on October 1, the date Golden Dome's own director says the program halts without a reconciliation bill Congress has not yet passed. None of the underlying growth data, record VC funding, record CCA and USV production ramps, record Palantir and Anduril contract pipelines, changes if any one of the three slips; what changes is how much of it was actually paid for and how much was simply promised.
What to watch through 2027
Watch whether Congress passes a second reconciliation bill before October 1, 2026, the single event that decides whether Golden Dome's $17.5 billion FY2027 request exists at all, and watch the FY2027 NDAA's stalled Senate cloture vote convert into an actual conference and floor passage, still unscheduled as of mid-August. Watch the rewritten DoD Directive 3000.09 land around its September 3 deadline under NSPM-11, how much autonomy latitude it grants systems like Anduril's Fury, Shield AI's Hivemind and Saronic's Corsair, and whether Senator Gallego's questions to Hegseth produce any public safeguards. Watch whether Palantir's Maven Smart System clears its September 30 program-of-record deadline, and watch NGC2 fielding reach the 4th and 25th Infantry Divisions on the Anduril-led, Palantir-supported data-layer baseline.
On hardware, watch Shield AI's X-BAT attempt its first vertical flight by the end of 2026 and General Atomics deliver its first production FQ-42A, the events that will show whether the CCA program's two airframe makers can hold their timelines, and watch the summer 2027 mission-autonomy down-select between Anduril, Shield AI and Collins Aerospace, the decision that hands one company the software layer on both aircraft. Watch Saronic's Marauder pass its Navy at-sea demonstration by October 2026 and watch Port Alpha break ground, and watch whether Anduril's reported $100 billion funding talks actually close. In Europe, watch whether Helsing's Japan deal converts from Rakuten-brokered talks into a signed GSDF contract once field testing ends in September, and whether the KIRK constellation produces a first satellite order. And watch Ukraine's 8-to-20-million-drone annual production range, still the clearest real-world test of whether the sector's mass-production thesis, on either side of the Atlantic and now at sea as well as in the air, actually holds.
Sources
- Tech Times - Guetlein warns Golden Dome halts in October without reconciliation funding
- Naval News - CENTCOM deploys Saronic Corsair USV against Iranian submarine at Bandar Abbas
- US Army - Army and industry align on common data baseline for NGC2
- DefenseScoop - Senator questions Pentagon's plan to revise autonomous weapons policy
- Bloomberg - Rakuten, German startup Helsing test military drones for Japan
- The Register - Palantir could receive $244M Pentagon no-bid contract


































