A look at Eric Schmidt-backed startup Ishtari, which uses machine learning to virtually assemble and test war machines from computer models of each component
The former Google CEO is on a mission to rewire the US military with cutting-edge artificial intelligence to take on China.
Context & Ripple Effects
Ishtari is the latest step in Eric Schmidt's long conversion from Google executive to self-appointed liaison between Silicon Valley and the Pentagon — a role that began formally when he chaired the Defense Innovation Board through its 2019 combat-AI principles, then sharpened into an explicit bet on profiting from a US–China AI cold war, backed by more than $2B in personal AI investing per CB Insights.
What makes Ishtari distinct inside that arc is its focus: rather than fielding hardware, it applies machine learning to virtual assembly and testing of war machines from component-level computer models, attacking the slow physical-prototyping cycle of traditional weapons development. It previews a portfolio approach Schmidt has since scaled up with the White Stork military-drone venture and his OpenAI-style scientific AI lab.
First-order effects
- Pentagon programs gain a way to iterate weapons designs in software before committing to hardware, shortening the test-and-fail loop that drives cost and schedule overruns at incumbent suppliers like the traditional primes.
- Schmidt converts his advisory position and personal wealth directly into equity stakes in defense capability, blurring the line between the Pentagon adviser who wrote the rules and the investor selling to the customer.
Second-order effects
- Defense incumbents whose margins rest on long physical development cycles face simulation-first competitors, pressuring them to buy or build comparable modeling capabilities rather than cede early-stage design.
- Schmidt's ventures — White Stork poaching engineers from Apple, SpaceX, and Google among others — bid up the same scarce AI engineering talent that commercial tech firms need, raising labor costs across both sectors.
Third-order effects
- If the model holds, US military capability development shifts from government-directed procurement to a network of privately funded, state-compatible labs — a structure Schmidt is already replicating with his OpenAI-modeled scientific organization.
- Simulation-first design concentrates advantage in whoever owns the component models and training data, making virtual-testing platforms a strategic chokepoint analogous to software stacks in commercial computing.
The trend: US military AI capability is migrating from government-run acquisition into a privately funded portfolio of Schmidt-backed ventures spanning simulation, drones, and scientific research, all aimed at competing with China.