Author: Deep Tide TechFlow

On November 17, 2025, 61-year-old Jeff Bezos returned to the role of CEO of a company. This marked his first return to leadership since stepping down from Amazon in 2021. The new company, Project Prometheus, has raised $6.2 billion in initial funding and focuses on "physical AI," targeting the manufacturing industry.

Seven months ago, 70-year-old Eric Schmidt took control of a rocket company called Relativity Space and became its CEO. He didn't explain why he was getting involved at his age, but perhaps "every day is important in the AI era" is the default answer.

In June of the same year, Steve Jobs' widow, Laurene Powell Jobs, gave a rare public interview. Sitting next to Jony Ive, she spoke about a prototype she had seen at io.com. It was an "AI device" acquired by OpenAI for $6.4 billion in stock, without a screen, and reportedly shaped like a player worn around the neck. Her assessment of the prototype was: "It's incredible to see an idea become a reality."

Three people, three different attitudes. But they're all betting in the same casino.

For the past three years, Silicon Valley's top financiers have been doing almost the same thing: pouring money from family offices, venture capital firms, and charitable foundations into AI. Schmidt, Bezos, and Powell Jobs are just the three most prominent figures. But a closer look at their investment lists reveals that this isn't the same game; they're investing in three completely different futures.

Schmidt: Treating AI as the next Cold War

According to data cited by Wikipedia and The AI Insider, Schmidt's family office, Hillspire, has invested in more than 22 AI companies since 2019, totaling over $5 billion. The list includes Anthropic, SandboxAQ (Quantum+AI, spun off from Alphabet), Inworld AI, Holistic AI, and Altera. These are the targets that "people in the industry" would list.

But what truly exposed his true colors was another list.

White Stork: A company that manufactures AI drones in Ukraine. Rebellion Defense: Defense AI. Istari: Simulation and simulation. Swift Beat: Military software. This is a family office that uses AI as the next generation of military equipment.

Schmidt has chaired the Defense Innovation Board since 2016 and co-led the National Artificial Intelligence Security Board from 2019 to 2021. He is a player who treats AI policy, defense procurement, and energy infrastructure as a single issue. In January 2024, Forbes revealed that he simultaneously launched the White Stork drone project in the United States and Ukraine, using the Ukrainian battlefield as a "laboratory for AI weapons."

Then comes infrastructure.

In January 2026, he co-founded a company called Bolt Data & Energy with Texas Pacific Land, serving as its chairman. This company doesn't lease server rooms or buy electricity from the grid; instead, it plans to build its own natural gas power plant in the West Texas wilderness, pumping electricity directly into its data centers. The plan is to initially reach 1 gigawatt, eventually scaling up to 10 gigawatts, equivalent to the electricity consumption of 7 million households. Texas Pacific Land is contributing $50 million, plus priority access to water resources. In an email reply to Fortune, Schmidt stated, "The biggest bottleneck facing AI is not algorithms, it's energy."

In March of the same year, he acquired a controlling stake in Relativity Space. This company was developing a reusable rocket called Terran R, aiming to challenge SpaceX's monopoly on low- and medium-Earth orbit launches. At that time, its order book totaled $2.9 billion.

Putting all these together, the logic becomes very clear.

Schmidt doesn't believe in the strategy of "investing in a basket of large model companies." He believes the outcome of AI competition will ultimately depend on three things: computing power (data centers and electricity), transportation (rockets, satellites, drones), and policy (the National Defense Committee and congressional hearings). He does invest in model companies, and after DeepSeek emerged, he publicly wrote articles calling for increased US investment in open source, but that was just one piece on his chessboard, not the whole picture.

His reaction to DeepSeek is very telling. Immediately after DeepSeek's release in early 2025, Schmidt wrote an article in The Washington Post calling it "a turning point in the global AI race." His prescription wasn't retreat, but rather an upswing, including more open source, more Stargate-style infrastructure, and more sharing of training methods among model labs.

In other words, he sees AI as an endurance race between nations, and he himself is already standing on the sidelines, serving as a member of the organizing committee. To outsiders, becoming the CEO of Relativity at 70 seems like a lot of unnecessary work, but he explains: Kissinger worked until he was 100; "What's needed in times of great change is responsibility, not withdrawal."

Bezos: The Full-Stack Control Freak

Bezos's playing style is completely different from Schmidt's.

According to StartupHub, citing data from TechCrunch, The Information, and Bezos Earth Fund, Bezos has invested over $19 billion in AI by mid-2026. This figure is still rising.

If you take it apart, it mainly consists of three parts.

The first piece is Anthropic. Amazon began investing $8 billion in Anthropic in September 2023, and pledged to add up to $25 billion in April 2026. Anthropic runs on AWS and uses Amazon's Trainium chips. This is a triangle that binds Amazon's cloud infrastructure, Bezos's model layer bet, and Anthropic's research capacity into a cohesive whole, going beyond mere financial investment. When Anthropic's valuation soared to over $60 billion, Amazon had already secured the largest slice of the external pie.

The second part is Bezos Expeditions' scattered investments. Bezos Expeditions is raising a multi-billion dollar AI-focused fund, upgrading Bezos from an "angel investor" to an "institutional investor." One of its investments, Perplexity, an AI search company, saw its valuation rise from $520 million in January 2024 to $20 billion in September 2025.

The third part is Project Prometheus.

In November 2025, Bezos and former Google X executive Vik Bajaj jointly announced the founding of this company, with initial funding of $6.2 billion and a team of nearly 100 people. Team members were recruited from OpenAI, DeepMind, and Meta. The founding advisors include Ashish Vaswani and Jakob Uszkoreit, the two authors of the 2017 paper "Attention Is All You Need." The company's goal is to apply AI to manufacturing, including automobiles, spacecraft, and chips.

Why manufacturing? Because it aligns perfectly with Bezos's other businesses. Amazon has the Kuiper satellite constellation, meaning the first customers for AI-powered manufacturing will be located right in his own home.

Musk referred to Project Prometheus as a "copycat" on X.

However, from a structural perspective, this is not plagiarism.

Bezos controls the model layer through Anthropic, the application layer through Perplexity and Figure, and the computing power layer through Amazon. Now he's creating another Prometheus, integrating AI into manufacturing and taking over the "execution layer of the physical world." This is a full-stack strategy, with each layer having its own strengths, from training chips to deployment in the factory floor.

About 10 days after Project Prometheus launched, it quietly acquired a company called General Agents. This company made "computer agents," AI agents that could directly control the entire computer. WIRED later disclosed that the acquisition took only four days from start to finish.

Harsha Abegunasekara, CEO of Donely, commented, "What General Agents really cracked was speed; Ace runs almost instantaneously on your computer." His company was originally a competitor of General Agents.

From angel investing to setting up a special fund, and then personally taking on the role of CEO, Bezos only took 18 months. He was actually building a system even larger than Amazon.

Powell Jobs: The Low-Key Type

If you put these three people together, Powell Jobs is the least like an "AI investor".

According to data from private wealth data platform Fintrx, cited by CNBC, her family office, Emerson Collective, has invested in at least nine AI-related startups since 2022, participating in funding rounds totaling over $1 billion. This figure is nowhere near the level of Schmidt or Bezos.

But what's interesting is the list itself.

Proximie: a remote surgery connectivity platform; Atropos Health: clinical data AI; Formation Bio: AI-powered pharmaceuticals; Curipco: an AI educational tool from Norway; Mistral: a large-scale French company, the only European player to challenge OpenAI.

There is no defense, data centers, or rockets.

Emerson Collective's website clearly outlines its investment focus: education, energy and environment, digital health, fintech, and media. AI is merely a tool interwoven within these themes. She holds a majority stake in The Atlantic and is highly skilled in Columbia-style "soft power" investing.

But the one shot she actually landed was not on the same line as any of the others.

After Jony Ive left Apple in 2019, Powell Jobs invested in his design company, LoveFrom, through Emerson Collective. Ive later told the Financial Times, "If it weren't for Laurene, there would be no LoveFrom." A few years later, Ive founded a hardware company called io, specializing in AI devices, and Powell Jobs invested in that as well. In May 2025, OpenAI acquired io in a $6.4 billion all-stock deal, making Ive a billionaire on paper. Emerson Collective also went public.

Another crucial investment was Emerson Collective, one of Mistral AI's early investors. At the time, the French company was one of the last remaining players in large-scale modeling in Europe.

Putting all of this together, her AI bets are focused on two directions: either "using AI to solve specific human problems" or "reshaping the way humans and machines interact" (io's devices, Ive's designs).

In an assessment, VC Sheet described Emerson Collective as: "A deliberately ambiguous LLC that combines venture capital, philanthropy, policy advocacy, arts and media ownership under one roof, utilizing any of the most effective tools in grants, policy lobbying, or investments."

Philosophically, she is closer to the older generation of East Coast family offices, valuing influence over returns, long-term over short-term, and the microphone over the spotlight.

Three investment philosophies

If you put the three lists together, you'll see three sets of judgments about the future of AI.

Schmidt is betting on a national race and infrastructure bottlenecks. In his world, AI will ultimately be determined by "who has the most electricity, who has the fastest rockets, and who has the most powerful drones." Models are just entry tickets; the real moat lies at the physical layer. That's why he personally took on the roles of Relativity and Bolt; he wasn't after rewards, he wanted control.

Bezos is betting on the industrial revolution-level diffusion of applications. He believes AI will eventually permeate every machine tool, every airplane, and every satellite, just like electricity. Therefore, he uses Amazon to lock in the model layer, Prometheus to lock in the manufacturing layer, and Expeditions to embed it into the consumer application layer. He's not betting on whether any single company can win, but on whether the entire "structure" can win.

Powell Jobs was betting on something else entirely. She was betting that people would eventually find the current human-computer interaction model unbearable. In an interview with the Financial Times, she and Ive repeatedly emphasized that "humans deserve better." Her investments in io, LoveFrom, medical AI, and educational AI all stemmed from the same judgment: the biggest market of the next decade would be "correcting the side effects of the internet of the previous decade."

Three sets of judgments, three different approaches.

Which is right? Nobody knows. Schmidt may have overestimated the weight of geopolitics in the AI economy. Bezos may have underestimated the capital consumption of the "full-stack" model, a prime example being Prometheus, which hasn't even shipped yet but is already rumored to be raising another $10 billion. Powell Jobs faces an even more awkward problem: Io devices won't be in mass production until 2027, and OpenAI's own financial model has been repeatedly questioned by the market.

But one thing is certain. When the winners of the previous generation of internet giants collectively shift their family offices to AI, this is no longer a small-scale trend in a particular sector. Bolt has already raised $150 million in seed funding, and Anthropic alone has secured the $33 billion commitment from Amazon. This level of capital flow will shape the industry landscape for the next decade.

As for who will have the last laugh, we'll have to wait until 2030 to see. Until then, all three old Dengs will still be at the poker table, and their chips will keep increasing.

Michael Saylor launched a poll on "Should I HODL this week?", and currently, nearly 80% of voters are not considering selling.

Popular Articles