Newsletter · 2026-08-29 · Ashutosh Agarwal

America Lost the Drone to China as the Real Race Moves to AI - Aerospace & Defense Weekly - Week of August 29, 2026

Aerospace & Defense Weekly for the week ending August 29, 2026. Podcast synthesis on a new US tariff on foreign-made drones and the admission behind it that DJI owns a category America invented, a former Pentagon AI official arguing the export-control lead over China is closer to a month than a year, and Amazon quietly flying an in-house American drone fleet toward 500 cities.

Aerospace & Defense Weekly

Week of August 29, 2026: America Lost the Drone to China as the Real Race Moves to AI

After a fortnight of loud headlines, the $100-billion-a-year venture flood, the aerospace IPO window, robot fighters carrying live missiles, the big set-piece conversations took a breath. What filled the gap was smaller, less glamorous, and in some ways more revealing: the plumbing underneath the boom. Not "is the money coming" (settled) or even "who captures it" (last week's fight), but the two most basic questions of all. Who actually builds the machines? And who controls the software that will fly and fight them?

Both questions got uncomfortable answers this week. On one show, a blunt reckoning with the fact that America invented the small drone and then lost the entire industry to a single Chinese company. On another, a former Pentagon official arguing that the real weapon in this race is not a missile at all but an AI model, one that, he says, already outclasses anything the West can defend against. The through-line is the same anxiety that runs through every bullish story in this sector: the demand is real, but the capacity, meaning factories, chips, code, and the people who make them, is not where it needs to be.

1. The most patriotic-sounding rant of the week: how America lost the drone

If you want the emotional core of the whole "cheap mass" debate, the one every defense investor keeps circling, it came, of all places, from a markets-and-comedy show. On DHUnplugged (August 19), the hosts, pundits rather than operators, stopped mid-episode to vent about a new U.S. tariff, "initiated last week, on imports of foreign-made drones." What set them off was less the tariff than what it admits.

"I'm embarrassed personally, and I think we all should be, as patriots," one host said, "that we invented the drone, and then we completely fumbled it, and now the largest drone producer is a company over in the East." That company is DJI, and the way they tell it, the loss was self-inflicted. "Many of the people that were involved in some of the early drone companies in California ended up basically having a reverse brain drain over to China," he said. The American approach, in his sardonic telling, was culturally lazy: "it's very American-coded... let's just put four propellers on it. And instead of solving any of the hard math, let's just add more propellers." DJI, meanwhile, "has like conquered the place."

Then it got worse. A finding "two years ago," the host said, was that DJI drones "had spy software," the ability, while streaming video over your Wi-Fi, to "send that footage somewhere else." The Pentagon's response was to ban them: "we can't use it in the military now because we can't trust DJI." Which leaves the U.S. in the awkward spot of needing drones for a drone war while having "no domestic drone production that's anywhere." The tariff, they argued, is the same playbook being run on solar panels, where America's share of global polysilicon (the raw material for solar cells) "fell from about 50% in 2005 to under 2% in 2024," and Washington's answer is a surcharge to force buyers back onshore.

And here is the part that should make investors pay attention rather than laugh. The hosts pointed out, pointedly, who profits from the drone tariff: "Junior holds a large position in Unusual Machines," the domestic drone maker (ticker UMAC), "and he's also on the board." (They meant Donald Trump Jr.) "Coincidentally, that stock was up nicely with this announcement. But I'm sure it's nothing." Their dry verdict on the alignment of interests: "They try to keep the market up, you know, because they're playing it."

Why it matters: strip out the jokes and this is a real thesis. The West's entire "flood the sky with cheap drones" doctrine, the one Rick Scott and the CCA operators were preaching a fortnight ago, runs straight into a supply-chain hole: the cheap drones are made in China, and the Chinese ones can't be trusted on a battlefield. Tariffs are the crude tool for rebuilding a domestic base from almost nothing, exactly as with solar and steel, and the solar comparison is not encouraging about the odds or the speed. For stock-pickers, the uncomfortable footnote is that the policy tailwind for domestic drone names like Unusual Machines is entangled with who owns them, a governance flag worth holding in view before chasing a tariff-driven pop. This is a pundit's riff, not an operator's data, but the underlying facts (DJI's dominance, the military ban, the reshoring push) are the hard constraint under every "cheap mass" pitch.

2. "A digital nuclear weapon": a former Pentagon official on the race that actually matters

The single most substantive conversation of the week wasn't about hardware at all. On The Peter McCormack Show (August 20), Gregory Allen, a former Pentagon AI official and now a think-tank analyst, so an insider-turned-expert rather than a current operator, made the case that the defining national-security contest of this decade is the AI race, and that Washington already treats it that way.

His best evidence is a policy that's now so familiar people forget how radical it was: the export controls blocking Nvidia and others from selling advanced AI chips to China. "This policy has been with us for four years now," Allen said, "so for some it's a little difficult to remember how radical it was. But that policy overturned two decades of trade and technology policy towards China." The logic, in his telling, was a bet on exponentials: "AI is in its infancy today, but... it's going to be crazy powerful. It would be really good if we had it and they don't. And it is worth us blowing up 20 years of U.S. and China trade and technology policy." The Trump administration's July 2025 AI Action Plan, he noted, explicitly "compared the current AI race with China with the space race with the Soviet Union," and Allen goes further: "I would say it's probably more important than winning the space race." He reached back for the quote everyone reaches for: Putin, September 2017, "whoever controls AI will control the world."

Then came the claim that gives the episode its title. Allen described an AI model, which he called "Mythos," that, "according to Senator Mark Warner," reportedly "autonomously hacked into the NSA's most secure servers in a matter of hours." His framing of what that means: imagine "a monopoly on this chainsaw that can cut into any digital system anywhere in the world instantaneously and just take control of it autonomously. It's not crazy to think of that kind of cyber power in the same terms as nuclear weapons." He noted that "the director of the CIA said recently that it is comparable to a digital nuclear weapon," and that the founder of China's leading cybersecurity firm, 360 Security Technology, called it "a cyber nuclear weapon." (These are claims relayed on the podcast, not things we can independently verify, so treat them as Allen's account rather than settled fact.)

The part with the sharpest investment edge is his read on the scoreboard. The whole export-control strategy, he argued, "was not just a reasonable but an admirable policy if you assume the United States has a year or so lead on China. Unfortunately, the United States does not have a year or so lead. I would say, apples to apples, we have a month's lead." He blamed "bungled" implementation by both the Biden and Trump administrations, and warned of the tail risk that underwrites every defense-electronics thesis: "China might invade Taiwan to try and get control of the chip factories as part of a way to one-up the United States in the AI race." He also described the U.S. and Anthropic "rolling out access" to the most powerful model to allies on a "priority list," securing the biggest banks in London, Seoul, and Tokyo, a picture of AI capability being treated, quite literally, like a strategic weapon to be distributed among allies.

Why it matters: for a defense investor, this is the argument for why the chip-and-AI layer is the defense-industrial base now, not an adjacency to it. Export controls, foundry security, and model access are national-security policy, which is why Nvidia, TSMC, and the AI labs sit inside the same risk map as the primes. Allen's "one-month lead" claim, if anywhere near right, reframes the whole China-deterrence trade: the export-control moat that the market has been pricing as durable may be far thinner than assumed, which cuts both ways (more urgency and spending on one hand, less strategic advantage on the other). And the Taiwan-invasion scenario is the ultimate off-ramp risk sitting under every semiconductor and defense-electronics position. This is an expert's assessment, heavy on relayed claims, so weight it accordingly, but it is the clearest statement this week of why "who controls the AI" now belongs in an aerospace-and-defense briefing at all.

3. Meanwhile, the civilian drone fleet quietly hit real scale

For contrast, and a reminder that the "who can actually build and fly these at scale" question has a very different answer in the commercial world, Crain's Daily Gist (August 19) reported that Amazon is expanding its Prime Air drone delivery to "cities in at least five more U.S. states in the coming months." Deliveries will start "by the end of the year" from warehouses in suburban Chicago and Atlanta, plus metro Cleveland and Syracuse, New York, adding to the "11 U.S. locations" it already flies from, including the Houston and Dallas areas and suburban Phoenix.

The numbers are the point. Amazon says the service will eventually reach "almost 500 cities and towns in the U.S., a six-fold increase from the current footprint." Its workhorse, the MK-30, "designed and built in-house," a six-propeller craft that takes off and lands like a helicopter with a range of "about seven and a half miles," has "made hundreds of thousands of deliveries this year," a dramatic jump from a program that spent a decade "churning through drone designs and suffering a number of crashes." Jeff Bezos first floated the idea back in 2013 and guessed it would take five years; it took about twelve.

Why it matters: it's a marginal item for a defense letter, and it comes from a news read rather than an operator, but the juxtaposition with the two sections above is the whole story. In the same week the West is fretting that it can't manufacture military drones domestically or trust the Chinese ones, a single American retailer is quietly operating an in-house-designed, made-in-America drone fleet at a scale of hundreds of thousands of autonomous flights a year and heading for 500 cities. The commercial autonomy stack, meaning airframes, autonomy software, air-traffic integration, and above all the industrial muscle to churn out units and iterate through crashes, is maturing fast. The recurring theme of this sector is that defense keeps discovering it needs exactly that commercial speed and manufacturing base. Amazon isn't a defense play. But the capability it just demonstrated is precisely the capability the Pentagon spent this week admitting it lacks.

The bottom line

Peel back the capital-flow headlines and you're left staring at capacity, the machines and the code, and who controls the means to make both. The drone story is a confession: America invented the category, let the talent and the manufacturing drain to a Chinese champion it now can't use in a war, and is reaching for tariffs to rebuild from near-zero, with the awkward twist that the beneficiaries include the President's own family. The AI story is a warning from someone who sat inside the building: the real weapon in the U.S. and China contest is a model, the export-control moat protecting the American lead may be measured in weeks rather than years, and Taiwan's fabs are the fault line under all of it. And the Amazon story is the quiet rebuttal to the gloom, proof that when an American company is actually motivated to build and iterate, it can put a home-grown autonomous fleet in the sky at scale.

Put together, the week argues that the next phase of this sector is less about funding the arsenal and more about building it: reshoring the drone base, holding the chip lead, and turning commercial-grade autonomy to military ends before an adversary does. The demand was never the debate. The question that keeps surfacing, from a comedy markets show to a former Pentagon official to a Chicago news brief, is whether the West can build fast enough to matter.