July 2026 Edition

July gave us two versions of the same story.

In the Middle East, the constraint on American power was not will or platforms. It was inventory. After thirteen consecutive nights of strikes, the Chairman of the Joint Chiefs warned the President that continuing at that tempo would draw down high-end interceptor stocks far enough to compromise readiness in Europe and the Pacific. The strikes stopped on July 25.

In PJM, the grid serving 13 states, Washington DC, and the largest data center cluster on earth, the capacity auction cleared at the regulatory price cap for the third consecutive year. It still came up 6,831 MW short of the reserve margin the operator says it needs, and it attracted 525 MW of genuinely new generation.

Different domains, same lesson. You cannot surge what you have not built.

Against that backdrop, Antares Nuclear closed a $470 million Series C on July 27, eight weeks after taking a reactor critical at Idaho National Laboratory. More on the thesis over at Investing in Organized Chaos.

Cyrus is proud to support transformative companies building the physical backbone of American dominance:

Portfolio Spotlight

Arbor Energy

Arbor continues building toward commercial deployment of its supercritical CO₂ turbine platform. The company's agreement with GridMarket, announced in March, covers up to 5 GW of baseload power for data centers and hyperscalers beginning in 2029. That runs to as many as 200 modular HALCYON units at 25 MW each, against demonstrated customer willingness to pay north of $100 per megawatt hour. In a month where PJM's auction made the supply gap impossible to ignore, that order book looks better every quarter.

Icarus

Icarus continues advancing its stratospheric solar platform. Worth noting for category context: in April, U.S. Air Forces Central awarded a $270 million IDIQ for ultra-long-endurance solar-powered ISR aircraft to Kraus Hamdani Aerospace. That is a combatant command putting real procurement dollars behind persistent stratospheric ISR as a mission area, which validates the category Henry and the team have been building toward since day one.

Volund Manufacturing

The Pentagon signed Family of Affordable Mass Missiles framework agreements on July 15 with Anduril, CoAspire, and Zone 5 Technologies, targeting up to 8,000 low-cost missiles per year. Anduril separately confirmed plans to build 7,000 Barracuda-500s in 2027. Every one of those airframes needs a turbine. Eric has been direct about where the bottleneck sits. Of roughly 200 investment casting foundries in the United States, only eight or nine work with the superalloys these engines require. Volund's automated approach is aimed squarely at that chokepoint.

Investment Opportunities & Updates

Antares Nuclear

The headline of the month. On July 27, Antares announced a $470 million Series C, consisting of $370 million in equity and $100 million in debt, co-led by Paradigm and Caffeinated Capital, with Point72 Ventures, Shine Capital, and Industrious Ventures participating. Proceeds fund the Mark-1 electricity-producing reactor in 2027 and initial deployments to defense customers in 2028.

This follows the June 4 milestone, when the Mark-0 achieved zero-power fueled criticality at Idaho National Laboratory. That came a month ahead of the deadline set by Executive Order 14301, and made Mark-0 the first privately developed non-light-water reactor to go critical in the United States in over forty years. On July 24, Jordan Bramble joined the President at the White House to mark the program's results.

Bramble's framing says it best. They won the race to criticality, and the race that matters now is commercialization.

Anduril Industries

A remarkable month. On July 2, the Air Force selected Anduril and General Atomics to produce airframes for the first three lots of CCA Increment One. On July 15, the YFQ-44A Fury fired a live AIM-120, the first air-to-air weapon ever launched from a U.S. Collaborative Combat Aircraft, and the same week brought the FAMM agreement and the company's first NATO contract, with the NATO Communications and Information Agency selecting Lattice for a nine-month air command-and-control evaluation across all 32 members. On July 20, Anduril unveiled Thunder, a Group 5 autonomous attack rotorcraft.

Separately, Reuters reported on July 24 that Anduril is in discussions with investors on a new funding round that could value the company at roughly $100 billion. The size of the raise has not been determined, and terms remain fluid. A company spokesperson said no decisions have been made about any future financing.

Chaos Inc.

Chaos continues scaling its VANQUISH distributed radar business following its selection to the Army's G-TEAD Marketplace, which lets any Army Service Component Command and NATO ally procure its systems without a separate competition. In a month defined by drone and cruise missile saturation, distributed multi-static radar is exactly the right architecture.

Critical Loop

Critical Loop continues deploying modular microgrids that compress time-to-power for industrial and data center customers from years to weeks. The company raised a $26 million Series A in April led by Conifer Infrastructure Partners and Hanover, bringing total funding to $49 million. PJM's auction results are the clearest argument yet for what they are building.

Dominion Dynamics

Dominion closed a C$139 million ($100 million USD) Series A on June 30, led by Georgian. It is the largest Series A in Canadian defense history, bringing total capital to C$169 million since the company launched in June 2025. The round funds AuraNet, the sovereign command-and-control layer already field-proven with the Canadian Armed Forces during Operation Nanook-Nunalivut in the High Arctic, and Scout, an Arctic-hardened autonomous platform designed to fly alongside fifth-generation fighters. Canada has committed to 5% of GDP on defense by 2035 with a mandate to direct 70% of procurement domestically. Eliot and the team are positioned precisely where that money lands.

Gitai 

Gitai completed the flight model of its S3 robotic servicing satellite in June and demonstrated ground capture of a moving, non-cooperative target in late July, which is the hard version of the rendezvous and proximity operations problem. The company was selected in May for the Space Force's Space-Based Interceptor Program, following its award under the Missile Defense Agency's SHIELD IDIQ in December.

Hemispheric AI

Hemispheric emerged from stealth on July 14 with $52 million in funding and Descartes, a six-billion-parameter NeuroAI foundation model for decoding non-invasive brain activity, trained on more than 250,000 hours of data from over 100,000 participants. Initial applications target PTSD, mild traumatic brain injury, depression, anxiety, schizophrenia, and Alzheimer's. Six years in stealth, and the decoding-first thesis is now public.

Impulse Space 

Impulse moved its Helios kick stage to the vertical test stand in Mojave on July 14 and was added to the bidder pool for National Security Space Launch Phase 3 Lane 1. VICTUS SURGO, the first operational Helios flight, carrying a Mira vehicle to geosynchronous transfer orbit, remains scheduled for no earlier than Q4 2026.

Stoke Space

Stoke completed proto-qualification of Nova's first stage in early June, clearing 46 structural verification tests at Moses Lake. Zenith engine hot-fire footage followed in early July, and the rebuild of Launch Complex 14 at Cape Canaveral is essentially complete. Kevin Weil joined the board on July 8. Nova's first orbital flight remains targeted for late 2026.

Market Insights

Inventory became the constraint on American military power.

CENTCOM struck Iranian targets for thirteen consecutive nights before the campaign paused on July 25. The reason was not diplomatic. General Dan Caine warned the President that sustaining the tempo would deplete high-end interceptor reserves to the point of degrading readiness in other theaters, and CENTCOM's commander assessed that the campaign around Hormuz had reached the limit of its effectiveness.

The numbers behind that judgment come from CSIS, which found the U.S. expended a substantial share of its Patriot PAC-3 and THAAD inventories, with replenishment measured in years rather than months. It concluded that stockpiles adequate for Iran remain insufficient for a peer competitor. Secretary Hegseth told the Senate Appropriations Committee on July 21 that the war had cost $37.5 billion and requested $67.1 billion more, warning that training would be curtailed without it. The committee approved an approximately $87 billion supplemental on July 24, with munitions the single largest line at roughly $21 billion.

The grid ran the same experiment and got the same answer.

On July 14, PJM published results from its 2028/2029 capacity auction. This is the mechanism that pays generators to guarantee availability on the worst day of the year, locked in three years ahead. It cleared at the FERC-approved cap of $325 per MW-day across the entire footprint. Total procurement reached 138,318 MW for $16.4 billion, and still fell 6,831 MW short of the reserve margin required to maintain reliability standards. That is the second consecutive RTO-wide shortfall in PJM's history.

Without the price cap, PJM estimates the auction would have cleared at $555 per MW-day, and $777 in the ComEd zone. The auction attracted 525 MW of new resources. PJM will ask FERC for approval to run an emergency backstop procurement in September.

An eleven-fold increase in four years, three straight years pinned against a regulatory ceiling, and the supply still is not showing up. That is what a structural shortage looks like in market data.

Four microreactors went critical in thirty days.

Executive Order 14301 required at least three DOE-authorized test reactors to reach fueled criticality by July 4. Four did. Antares' Mark-0 on June 4, Valar Atomics' Ward 250 on June 18, Deployable Energy's Unity on June 30, and Aalo Atomics' Aalo-X on July 4. After decades in which advanced reactors existed mostly in slide decks, four physical machines reached criticality in a single month. The capital is now following, and Antares' $470 million round is the first major test of what commercialization funding looks like in this category.

Budget backdrop.

The FY2027 request carries a $1.45 trillion topline, a 44% increase over FY2026 enacted. The House passed its FY2027 NDAA on July 22 by 216 to 212. The request includes $53.6 billion for autonomy, drone platforms, and contested logistics, the largest such investment ever proposed.

Until Next Time

The two halves of what we invest in stopped being separate stories this month.

A President's options in the Middle East were bounded by how many interceptors American factories can build in a year. A hyperscaler's options in Northern Virginia are bounded by how many megawatts clear an auction three years out. Both are production problems. Both were decided years ago, by capital that either went into physical capacity or did not.

The companies we back exist to move those constraints. Turbines, reactors, radars, microgrids, airframes. July made the case better than any pitch deck could.

Thank you for reading.

If you're in LA, reach out. Always happy to connect.

Regards,
Jordan Yashari
Founder & General Partner
Cyrus Ventures