The Pentagon gave the defense industry 21 days to fix a supply chain it spent thirty years hollowing out. The memo calls it a production problem. It is a demand problem, and commerce solved that one a decade ago.
On August 5, Deputy Defense Secretary Steve Feinberg sent the primes a letter with a clock attached. Twenty one days to submit plans for faster delivery and expanded production across sixteen critical programs. Years-long development cycles, he wrote, are not acceptable. The Pentagon confirmed the memo is real and said it will shape the FY2028 budget.
The memo asks for more than schedules, it asks the primes to name the capital investments they are prepared to make themselves, the facilities they would expand, the risk they would carry before Congress guarantees a dollar of funding. The Pentagon’s phrase for this is skin in the game. Hold that thought; it matters later.
It reads like a production directive: make more, make it faster. That framing mistakes the symptom for the disease, and the difference is the entire story.
Start with the arithmetic that produced the memo. In the first month of the Iran war, the United States expended more than 850 Tomahawks, more than 1,000 Patriot and THAAD interceptors, and over 1,300 Army tactical ballistic missiles. CSIS estimates the Patriot inventory fell from roughly 2,330 before the war to somewhere between 759 and 827. That is a decline of at least 65 percent. THAAD dropped by nearly 40. The magazine emptied in weeks, and it will take years to refill, because the binding constraint was never the assembly line. It is solid rocket motor capacity, concentrated in two suppliers, sitting on an energetics base that is thinner still.
Retail buried a generation of companies between 2010 and 2020, and the autopsy was the same every time
That is the crisis everyone is describing. It is not the crisis that matters.
The failure is in the feedback loop
Retail buried a generation of companies between 2010 and 2020, and the autopsy was the same every time. The companies that died were not the ones with bad products. They were the ones whose demand signal and whose production had come unbolted from each other.
SEARS reordered on seasonal guesses against a customer that had already moved. Department stores bought two quarters ahead of a demand curve they could no longer see. Meanwhile retailers like Zara compressed the loop from concept to shelf to roughly two weeks, and Shein compressed it to a mere number of days, cutting micro-batches against live sell-through and reordering only what the signal confirmed. The winners did not have better warehouses; they had shorter, more predictive loops. When demand moved, their systems saw it move. When demand moved on the losers, the shelf was either empty or buried in markdowns, and the P&L found out a year later. The traditional defense industrial base is a version of the losing company.
For thirty years, the interceptor model optimized for one comfortable assumption. A small number of exquisite, expensive rounds. Drawn down slowly against a small number of high-value threats. Replenished on a peacetime cadence measured in fiscal years. Every incentive pointed toward fewer, better, and costlier. Margin lived at the top of the stack. Capacity was sized to the reorder rate, and the reorder rate assumed the magazine would never actually empty.
Then the demand signal changed shape, and the architecture could not see it.
The exchange ratio is the demand signal
The signal is a number, and the number is a ratio. A Patriot interceptor costs on the order of four million dollars. The threats now saturating the airspace, the FPV quadcopters of Ukraine and the Shahed-class loitering munitions of the Middle East, cost between a few thousand and a few tens of thousands each. Carbon fiber, injection-molded plastic, commercial motors, off-the-shelf guidance. Sourced from the same supply chain that ships consumer electronics, and producible at effectively unlimited scale because nothing in them is exotic.
Every time a four million dollar interceptor kills a fifty thousand dollar drone, the defender wins the engagement and loses the war, because the attacker builds a hundred more before the defender refills a single tube. That is not a marginal inefficiency. It is a structural cost inversion, and it is the actual content of the Feinberg memo whether the memo knows it or not. The magazine did not empty because production was slow; it emptied because the defender answered cheap, distributed mass with expensive, centralized scarcity. That is an architecture decision, not a manufacturing one.
Commerce has a name for a business that answers distributed demand with centralized inventory. It calls it bankrupt.
The reorder loop, in public
Look at how the department has tried to replenish so far, because it proves the diagnosis better than any memo could.
The Pentagon has spent months announcing framework agreements with primes and startups to expand supplies of interceptors and low-cost munitions. These are nonbinding. They signal intent to buy, pending congressional funding. Tom Karako at CSIS put it plainly: they are agreements to agree. Almost nothing has been contracted.
Translate that into commerce terms. The buyer is issuing letters of intent instead of purchase orders, against inventory that is already gone, on a funding cycle that resolves in years. There is no reorder signal in that system because there is no order in it. And the skin-in-the-game ask completes the picture: the channel is asking its suppliers to build capacity and carry the risk ahead of committed demand. Retail ran this exact play on its vendor base in the 2010s. The vendors who complied on faith got crushed. The vendors who answered by going direct, owning their own demand signal, and sizing production to real sell-through became the DTC generation.
A 21-day memo is what it looks like when an organization discovers in public that its reorder logic is three budget cycles behind its own sell-through.
Cheap, many, attritable
The answer commerce arrived at was not to build the central warehouse faster, it was to distribute the magazine. Push cheap, replaceable, demand-responsive capacity to the edge. Tolerate loss at the unit level and win at the level of the system. The Pentagon has a word for this already: attritable.
And watch what the department is actually funding underneath the memo’s production panic.
Replicator 2, the counter-small-UAS line of effort, made its first acquisition on January 11, 2026. AI-driven interceptor drones using radar and tethered nets to detect and capture small UAS, fielded through a joint interagency task force built specifically to move counter-drone capability from a community of interest to a community of action.
Drone Dominance, a roughly one billion dollar program, plans to run production through four competitive gauntlets beginning in early 2026. The first phase alone puts twelve vendors on contract to deliver 30,000 one-way attack drones at a unit cost of 5,000 dollars.
Thirty thousand units at five thousand dollars. Set that against a magazine of 827 Patriots at four million. The department is quietly, under a different budget line, building the distributed magazine. It is doing offensively what the interceptor base cannot do defensively. Match cheap mass with cheap mass, and make the exchange ratio run the other way.
So the memo and the gauntlet are the same organization holding two contradictory theories of its own supply chain, one in each hand. One says refill the expensive magazine faster; the other says the expensive magazine was the wrong architecture.
Only one of them is right, and the interceptor math already delivered the verdict.
Demand-driven design, applied to the arsenal
This is where a commerce operator sees something the acquisition system structurally cannot, because the acquisition system was built to optimize the exact thing that is now the problem.
The bottleneck everyone names, solid rocket motor capacity, is real and it is also downstream. The upstream failure is that the system has no live loop between what gets expended and what gets built. It reorders on the fiscal-year cadence of a peacetime demand curve, against a wartime demand curve that moves in weeks.
The fix is the thing distributed commerce spent a decade building. A production architecture whose default is cheap, plentiful, and close to the signal. Capacity sized to real expenditure instead of to a comfortable assumption. A feedback loop tight enough that when the magazine starts to empty, the line already knows. Attritable systems are not just cheaper interceptors. They are a supply chain that can see its own demand, because the units are cheap enough to build at the rate they are spent.
That is demand-driven design. It rebuilt retail. It is now the unspoken logic of the Pentagon’s most-funded drone programs. And it redraws the map of who matters. The twelve vendors on the Drone Dominance gauntlet are running the DTC playbook against the primes: short loops, cheap units, capacity sized to expenditure, demand signal in hand. The primes are the department stores in this story. Some of them will figure out that the shelf has moved. The ones that internalize distributed production, that stop defending the exquisite magazine and start owning the distributed one, are the ones that still matter in 2030. The rest will keep signing agreements to agree.
The verdict
The memo will be read as a production story. The arsenal ran low, the department cracked the whip, and the primes will surge. That reading is comfortable but it is wrong. The arsenal ran low because it was the wrong arsenal. A centralized, expensive, slow-feedback magazine answering a distributed, cheap, fast-moving threat. Twenty one days of accelerated production refills the wrong magazine faster.
The war did not expose a manufacturing gap. It exposed an architecture that lost the plot the way legacy retail lost it, by optimizing for the demand curve it wished it had instead of the one it got. The signal has already spoken. It is a ratio, it is inverted, and no volume of surge production turns a four million dollar answer to a fifty thousand dollar problem into a trade worth making.
The magazine problem is a demand problem. Commerce solved it a decade ago by innovating out of it. The arsenal is about to learn the same lesson, on a 21-day clock, in front of Congress.
Research and Analysis by Web Smith