This is Scale & Strategy, the newsletter that's like a backstage tour of the wildest market on earth.
Here’s what we’ve got for you today:
Bezos wants AI to become the engineer behind the next generation of machines
SpaceX just gave AI IPOs their first real stress test
Bezos wants AI to become the engineer behind the next generation of machines
Jeff Bezos is betting that AI’s next major role won’t just be answering questions or writing code.
It will be designing the physical world.
Amazon’s founder just revealed more details about Prometheus, his AI startup focused on engineering, while announcing a $12 billion funding round at a $41 billion valuation.
The goal: build what Bezos calls an “artificial general engineer” that can accelerate the design and development of complex physical systems.
Prometheus launched quietly in late 2024 with Vik Bajaj, a physicist and chemist who previously helped build Verily.
The company is targeting a very specific bottleneck: engineering workflows that still look surprisingly similar to decades ago.
Bajaj argued that engineers working on some of the world’s most advanced machines, including jet engines, still rely on tools and processes that have barely evolved.
Bezos wants to compress the entire invention cycle.
Today, changing something as simple as increasing jet engine thrust by 10% can take years of simulation, testing, and redesign.
The Prometheus vision is a faster “dream-build loop” where AI helps move from idea to prototype much faster.
The ambition is basically an AI teammate for physical invention: one that can reason through designs, run simulations, test possibilities, and help engineers build things that currently require massive teams and long timelines.
Bezos also pushed back against the idea that AI will primarily destroy jobs.
His view is that AI-driven productivity gains will create more than 10x the opportunities lost, while raising living standards.
That prediction is going to face plenty of skepticism, especially in a moment where many companies are already cutting roles while increasing AI investment. Having one of the wealthiest people on Earth deliver the message naturally makes people reach for the skepticism button. Humans do love a good irony sandwich.
The bigger question is whether AI can actually move beyond digital work.
Most AI progress so far has happened inside software. Physical engineering is harder because the systems have real-world constraints: materials, physics, manufacturing limitations, and expensive testing cycles.
If Prometheus works, the impact could extend far beyond software companies.
It would give engineers a radically faster way to design everything from aircraft and robots to industrial systems and new materials.
Bezos spent decades building companies around logistics, infrastructure, and large-scale physical operations.
Prometheus looks like that same obsession pointed at the invention process itself. The bet is that the next productivity jump comes from making the people who build things dramatically more capable.
Framer helps teams design, build, and launch their marketing sites lightning fast. With the ability to publish hundreds of CMS pages in a single click, operate at a global scale with seamless localization, and even host unified content across multiple domains, teams have never been able to ship faster. Trusted by companies like Miro, Bilt, and Perplexity
Speed without chaos: ship pages and updates faster without turning the site into a fragile set of one-off hacks
Reduce dependency: shift routine brand and marketing work out of product engineering queues.
Production-grade foundation: Run real marketing systems (CMS, SEO, performance optimization) with governance and collaboration
SpaceX just gave AI IPOs their first real stress test
The AI IPO wave has officially started, and the market response was loud.
SpaceX closed its debut day up 19%, opening at $135 per share and finishing around $161. The offering valued the company at roughly $1.77 trillion, pushing its market cap above $2 trillion by the close.
Now investors are asking the obvious question: does SpaceX’s public market success create a roadmap for the next AI giants?
With OpenAI and Anthropic reportedly preparing IPO filings, SpaceX is the first major test of whether public markets are willing to pay extreme valuations for companies tied to the AI boom.
The comparison only goes so far.
SpaceX is entering public markets with a major advantage: an existing, profitable business.
Its Starlink operation has become a serious revenue engine, reportedly generating strong margins that can offset the enormous spending required for rockets, satellites, and AI infrastructure.
OpenAI and Anthropic don't have that luxury.
Both companies are carrying massive compute costs while offering expensive AI services at scale to build adoption. Training and inference costs continue rising, and both companies are reportedly exploring lower token pricing to expand usage while managing economics.
The business models look very different.
Where SpaceX gets another boost is its long-term AI infrastructure thesis.
The company’s acquisition of xAI brought in the Grok chatbot and the Colossus AI supercomputer project. A major part of the strategy is exploring orbital data centers as a way to bypass Earth’s growing constraints around energy, land, and grid capacity.
That future vision is a major part of the valuation story.
Morningstar argued that SpaceX’s valuation depends heavily on two unproven technologies: fully reusable Starship operations and commercially viable orbital AI data centers.
The firm estimates those technologies may not be proven until 2028 or later.
In its most optimistic scenario, Morningstar believes SpaceX could eventually capture 20% of projected AI computing capacity through orbital infrastructure by 2040, supporting a valuation around $154 per share. It assigns that scenario only a 7% probability.
Markets appear willing to price in some of that upside anyway.
That’s the important signal for OpenAI and Anthropic.
SpaceX’s IPO success doesn’t prove that every AI company deserves trillion-dollar valuations. The businesses are fundamentally different. But it does show investors are willing to pay heavily for companies positioned at the center of AI infrastructure and future compute demand.
The numbers are getting absurdly large.
Anthropic is currently valued around $965 billion. OpenAI sits around $852 billion. Add SpaceX’s valuation into the mix and the combined public market expectations for just three AI-linked companies would exceed the total capital raised across the entire US IPO market in 2025.
That scale guarantees one thing: the AI bubble debate is coming back.
The market is now being asked a simple question: are these companies building the infrastructure for the next computing platform, or are investors pricing in a future that still needs a lot of things to go right?
Public markets, as usual, have decided to answer by throwing billions at the question first and reading the footnotes later.
Framer helps teams design, build, and launch their marketing sites lightning fast. With the ability to publish hundreds of CMS pages in a single click, operate at a global scale with seamless localization, and even host unified content across multiple domains, teams have never been able to ship faster. Trusted by companies like Miro, Bilt, and Perplexity
Speed without chaos: ship pages and updates faster without turning the site into a fragile set of one-off hacks
Reduce dependency: shift routine brand and marketing work out of product engineering queues.
Production-grade foundation: Run real marketing systems (CMS, SEO, performance optimization) with governance and collaboration
