Jeff Bezos is returning to day‑to‑day leadership. On November 17, 2025, the Amazon founder emerged as co‑CEO of Project Prometheus, a new artificial intelligence venture focused on engineering and manufacturing.
Reports say the startup has raised $6.2 billion (C$8.7 billion), a scale rarely seen at formation, and is recruiting specialists across computing, autos, and aerospace. Early descriptions point to tools and models that work with physical systems, not just software. Project Prometheus lists about 100 staff and a short mission line, AI for the physical economy. Bezos shares leadership with scientist and builder Vik Bajaj.
Early funding signals scale and risk
The funding size matters for infrastructure. Models that plan factories, design parts, and steer robots need enormous compute, power, and cooling. That means more high‑density data halls, faster networks, and long‑lead equipment orders.
If Project Prometheus moves quickly, suppliers of chips, liquid cooling, and substation gear could see firm demand. The $6.2 billion figure, sets expectations for rapid pilots and hardware buys in 2026. It also flags execution risk if proofs do not mature into paid deployments, given today’s tight capital and supply chains for advanced computation, a point underscored when media noted the raise as one of the largest for an early‑stage firm.
Focus on real‑world engineering
Project Prometheus targets sectors with heavy assets and strict safety rules. That includes chip packaging lines, automotive plants, and spaceflight assembly. These are complex, regulated environments. Any AI that touches them must fit with quality systems, worker training, and certification.
Speaking about the broader market mood on October 3 in Turin, Bezos offered a blunt frame, saying, “This is a kind of industrial bubble,” a view that sets a cautious tone even as investment builds.
Vik Bajaj brings lab‑to‑market experience from Alphabet’s X lab and Verily, a mix that fits Prometheus’s physical‑world pitch. Ties to Blue Origin may help with aerospace workflows, though governance and data segregation will matter. Early trials will likely sit inside existing procurement rules, with phased milestones and off‑ramps.
Still, the ambition is clear. “The benefits to society from AI are going to be gigantic,” Bezos said, outlining a long view that leans on patient capital and measurable gains on factory floors.
What to track next
Three signals will test momentum. First, siting of computation and the power deals behind it, since energy use and water risk are now board‑level issues. Second, named pilots with delivery schedules in autos, aerospace, or electronics, which would show a path‑to‑value beyond demos. Third, hiring pace in controls, simulation, and safety engineering, roles that turn models into certified tools.
If those elements line up, procurement programs could move from trials to framework agreements in 2026. If not, the story stays in research, and the capital waits for clearer proofs.