Open the Front Door: Emil Michael on Rebuilding U.S. Defense Around Startups, AI and Supply Chains

Emil Michael, formerly chief business officer at Uber and a White House fellow, explains that the Pentagon used to bundle acquisition, technology and logistics in one organization. About eight years ago it split acquisition from research and engineering because technology was moving faster than the old system. As under secretary for research and engineering, he says, he now gets to work with entrepreneurs and new companies much as he did in Silicon Valley.

His portfolio includes DARPA, which he calls some of the most advanced research in the country; the Defense Innovation Unit in Mountain View, meant to link startups and defense; and a strategic capabilities office that modifies existing systems for strategic surprise. In recent months he also took over the chief AI office for an organization of about three million people. The department spends roughly 150 billion dollars a year on technology, so the goal is to unify efforts and avoid duplication.

Sarah Guo asks how his reputation as a fast, aggressive dealmaker fits the department's slower history. Michael says the urgency and willingness to run over barriers transfer from company leadership to a bureaucracy, though more slowly. He adds that he has more resources and a mission he considers important to humanity, and says many people are asking to join his team.

Michael uses an Uber-era analogy: a Series B company with 14 product lines would be uninvestable. When he arrived, the department had 14 critical technology areas, which he says cannot all be critical and are hard for people to hold in their heads. He cut them to six after studying them and made them action-oriented by putting sprints behind each, like an engineering team.

Number one is applied AI. The department won't build a foundation model, he says, since the private sector is spending hundreds of billions, so the job is to adapt it to defense use cases. Next is "scaled hypersonics": hypersonic missiles travel at least Mach 5, are maneuverable and expensive, and the goal is to make them more producible and cheaper. Another is scaled directed energy, using high-powered microwaves or lasers, because zapping drones is cheaper than using a 5 million dollar missile against a 50,000 dollar drone.

Prioritizing across Army, Navy, Air Force and Space Force means looking for common needs and enablers. If something is useful to only one service, that service can handle it with its own labs and budgets. Michael wants economies of scale across the enterprise, and says narrowing the list took a couple of months.

Elad Gil asks how the department will shift from large crewed systems like aircraft carriers to distributed autonomous fleets. Michael says the Russia-Ukraine war started a renaissance in drones, treating robots as the new front line. In territorial battles, he says, robots can fight first and reduce human casualties, and he credits drone warfare with preventing even higher casualties in that conflict.

Carriers will still project power and protect sea lanes for some time, he says, but the mix will move toward autonomous submarines, boats, large drones and airplanes. He expects each budget year to allocate more as AI improves at sensing and controlling these systems. In ten years, he says, it wouldn't surprise him if 20 to 30 percent of the budget went there, and that fraction buys far more firepower than the same money spent on crewed systems.

Michael says even unclassified Department of War networks are secure, so employees can't simply sign up for consumer chatbots. The team had to settle policies and build a data flow so nothing returns to the pool of data used by model providers. He credits a 60-day tiger team of former Databricks, Meta and AWS engineers, plus strong collaboration from Google's Gemini team.

GenAI.mil launched to three million people, and more than a million unique users used it in the first 30 days, about a third of the enterprise on one model. It currently runs Gemini 2.5, with 3.0 coming in a couple of weeks, then Grok from xAI, and later across classification levels. Michael says it will fundamentally change how people work, and was possible because of "fast people."

Instead of white papers and committees, Michael's applied AI work targets three areas: enterprise efficiency, intelligence and warfighting. In intelligence, AI can help analysts by flagging items and fusing different types of intelligence that a single analyst usually wouldn't see. In warfighting, it can support wargaming, planning and order execution. Each area gets pacesetter projects meant to break down data barriers and show the department what AI can do.

Looking five to ten years ahead, he says autonomous systems are the clearest area, citing how AI-infused robotics are outpacing older approaches. On the enterprise side, he says people in DC are just starting and will be amazed that they no longer have to write 20-page PowerPoints or 100-page reports. Less drudgery, he says, will help attract better talent.

Michael says one untold story of DOGE is the talent it brought into every department, people with a "fixer-builder" mentality, not all of them engineers. He borrowed many of them, and they attract more. He also mentions a US Tech Force, led by Scott Kupor, meant to bring thousands of college graduates into two-year stints as service to the country as technologists rather than soldiers.

He recruits directly too, calling people who recently left jobs and pitching a year doing the coolest work imaginable, with "Recruiting Tuesdays." He also asks friends, including the hosts, for people on the bench. He describes picking off a few people every month.

On the "Arsenal of Freedom" tour with Secretary Hegseth, Michael describes the secretary's first reform as restoring the warrior ethos and objective standards, and the next as the fitness of the defense industrial base. Given recent operations in Iran, Venezuela and against the Houthis, he says the base wasn't as ready as it should have been. The message to industry is to produce, cut bureaucracy and open a wide front door to new entrants.

He describes a change in procurement: instead of a 300-page requirement document, the department now states the problem and lets companies propose solutions. Primes excel at filling out questionnaires, whereas a startup may have a novel approach. Over six days on the road in two weeks, they visited about 20 companies.

Michael says the department is more open for business than ever, citing China's military buildup as the biggest in world history. In the 1980s there were around 50 defense contractors, and mergers left about five, leaving room for new entrants. He finds it crazy that SpaceX, Anduril and Palantir each had to sue for their first contract, and says the idea now is that they don't have to.

His caution is that selling to the government is not like selling to another business, because lives are at stake and products must work. He also notes that hardware is hard in defense, so startups must prove they can produce flawless products at scale. His advice: a tech pitch and prototype are good, but the question is the plan to deliver 10,000 units.

On the "valley of death," he says several mechanisms now collapse it. DIU offers rapid contracting with about a billion dollars; another program helps companies that have a product and need to scale manufacturing. The Office of Strategic Capital has 200 billion dollars in lending authority at roughly Treasuries plus 100 basis points, and has done five critical minerals deals in four or five months. The department is also reserving part of the budget each year for in-year decisions, since Congress passes funding late.

Asked where founders should build, he points to hypersonics, larger drones resembling long-range missiles, and GPS-denied seeker technology. He also cites Golden Dome, which involves taking things out from space. He considers AI generally well funded already.

On change management for three million people, Michael says he tries not to overthink it, taking an Uber-like approach of moving forward until someone says stop. It is not as fast as private industry, but still works. He argues that bureaucracies deserve their bad reputation but individuals want to thrive, and that Department of War staff tend to be mission-oriented.

On industrial base rebuilding, he says that though he is a free-market person, outsourcing pharmaceuticals, critical minerals, semiconductors and nuclear plant building leaves the country without control of its destiny if an adversary controls them. He says the department is operating as if it must be self-sufficient in critical minerals within a year, citing the China deal's export control provisions, and notes about 200 pharmaceuticals on a shortage list.

The hosts argue that regrowth will look different from 40 years ago, driven by automation and AI. Michael adds that people will still be needed, citing welders for ships, data center electrification and housing, and says some welding jobs pay six figures in lower-cost areas. He also points out that cost arguments are overstated, since China subsidizes rare earths and dumping has crushed industries, which he links to the logic of tariffs.

The Office of Strategic Capital targets brittle supply chains where the technology exists but a company needs to produce ten times more. Michael mentions germanium, brushless motors for drones and fiber optic cable, and says some systems depend on a single nozzle supplier, whom he wants to succeed alongside two or three others. Sarah notes the supply chain is geographically distributed, with germanium in Bozeman, Montana. He says capital and research and engineering belong together, and admits there is no public to-do list for founders but agrees one would help.

Michael was a White House fellow in 2009 to 2010, during the global war on terror, when the adversary was insurgents with crude weapons like improvised explosive devices. He says the U.S. then had superior technology but was vulnerable to adversaries with nothing to lose. Today's near-peer adversary has space capabilities, long-range strike, drones and AI, so the battlefield has become far more about technological superiority.

What surprised him was how much this adversary had done in so little time, which he says only hit him once he was inside. He cites what has happened to markets such as germanium, fiberglass and Taiwan's importance, saying he now understands them viscerally and not just theoretically. He closes by promising to return to the hosts with a to-do list for founders.