BETHOS, SCHMIDT, POWELL JOBS: THREE AI INVESTMENT PHILOSOPHY OF SILICON VALLEY OLD MONEY

Original by Deepwater TechFlow

On November 17th, 2025, 61-year-old Jeff Bessos re-established as a company CEO. For the first time since he left the Amazon in 2021, he has returned to his seat. The new company, Project Prometheus, started with $6.2 billion and focused on “physical AI”, targeting manufacturing。

Seven months ago, 70-year-old Eric Schmidt owned a rocket company called Relativity Space, and went up to be CEO. He didn't explain why this age was going to end, and maybe the answer to "Ai's important every day" was a default。

In June of the same year, the widow of Jobs, Laurena Powell Jobs, gave a rare public interview. She sat next to Johnny Ive, talking about the prototype she saw at O.C. It's an "AI device" collected by OpenAI with $6.4 billion in shares, without screens, and is said to be like a player hanging around its neck. Her evaluation of the prototype was: “It is an incredible thing to see ideas become reality”

Three men, three gestures. But they all bet in the same casino。

IN THE LAST THREE YEARS, SILICON VALLEY'S TOP MONEYBAGS HAVE ALMOST DONE THE SAME THING: PUT ALL THE MONEY FROM FAMILY OFFICES, VENTURE CAPITAL, CHARITABLE FOUNDATIONS INTO AI. SCHMIDT, BESSOS, POWELL JOBS ARE JUST THREE OF THE MOST VISIBLE. BUT IF YOU TAKE DOWN THE LIST OF THEIR LOGOS, YOU'LL FIND IT'S NOT THE SAME GAME, THEY'RE VOTING FOR THREE COMPLETELY DIFFERENT FUTURES。

SCHMIDT: CONSIDER AI THE NEXT COLD WAR

According to figures quoted by Wikipedia and The AI Insider, the family office of Schmidt Hillspire has invested more than 22 AI companies since 2019, totalling more than $5 billion. The list includes Anthropic, SandboxAQ (Quantum +AI from Alphabet), Inworld AI, Holistic AI, Altera. These are the targets that people in the industry will list。

But what really exposed him was another list。

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

Schmidt chaired the Defence Innovation Commission from 2016 and co-led the National Committee on Artificial Intelligence Security from 2019 to 2021. He was a player who treated AI policy, defence procurement, energy infrastructure as the same thing. In January 2024, Forbes disclosed that he had launched a White Stork drone project simultaneously in the United States and Ukraine, using the Ukrainian battlefield as an “AI weapons laboratory”。

Then infrastructure。

In January 2026, he and Texas Pacific Land founded a company called Bolt Data & Energy, as its own chairman. The company, which does not rent rooms or buy electricity grids, is building its own gas power plant in the open field of West Texas, filling it directly with data centres. The plan is to achieve 1 giva first and eventually 10 giva, equivalent to 7 million households. Texas Pacific Land offers $50 million and includes priority water supply rights. Schmidt's e-mail response to Fortune states, "The biggest bottleneck facing AI is not algorithms, it's energy."

In March of the same year, he held shares in Relativity Space. This company is doing a reusable rocket called Terran R, which aims at impacting SpaceX's monopoly on medium- and low-orbit launches. As of then, the order was $2.9 billion。

If we put these together, the logic is clear。

Schmidt didn't believe in the "Big Models." He believed that the success of AI would ultimately fall on three things, namely, computing power (data centres and electricity), delivery (rocket, satellite, drone), policy (defence committee and parliamentary hearings). The model company also voted, and after DeepSeek came out, he publicly wrote an article calling on the United States to increase its input, but it was just a piece on his board, not all of it。

His reaction to DeepSeek is very illustrative. In early 2025, immediately after DeepSeek's release, Schmidt wrote in the Washington Post, saying that it was “a turning point in the global AI competition”. His prescription was not retreating, but adding codes, including more open sources, more Stargate infrastructure and more training methods shared among model laboratories。

In other words, he saw AI as an interstate endurance race, and he was already standing by the track and on the organizing committee. At the age of 70, who was the CEO of Relativity, the outsider looked like he was working at the age of 100, explaining that “the moment of major change required responsibility, not withdrawal”。

Bethos: All-in-house control freaks

Bessos' play is completely different from Schmidt。

According to StatupHub, citing TechCrunch, The Information and Bezos Earth Fund, as of mid-2026, Bessos had deployed over $19 billion on AI. This number is still rising。

Open up and look。

The first is Anthropic. The Amazon started to invest $8 billion in September 2023 and committed up to $25 billion in April 2026. Anthropic ran on AWS, using the Amazon Trainium chip. This is a triangulation of Amazon cloud infrastructure, Bessos model bets, Anthropic research capacity, not just financial investment. By the time the Anthropic valuation hit over $60 billion, the Amazon had eaten the largest external piece of cake。

The second is the bulk investment of Bezos Express. Bezos Express is raising a multi-billion-dollar AI special fund to upgrade the "Bezos Personal Angel" to "institutional investors." Of these, it invested in Perplexity, an AI search company, which increased from $520 million in January 2024 to $20 billion in September 2025。

The third is Project Prometheus。

In November 2025, Bessos and former Google X executive Vik Bajaj jointly announced the start-up of the company, with $6.2 billion in start-up funding and nearly 100 people, and team members dug it out of Openai, DeepMind, Meta, with Ashish Vaswani and Jakob Uszkoreit, two authors of the 2017 article, Attention Is All You Need. The company's goal is to use AI for manufacturing, including cars, spacecraft, chips。

Why make it? Because it coincides with the rest of Bessos' business. Amazon has a Kuiper satellite constellation, manufacturing AI's first clients in his own home after landing。

Mask refers to Project Prometheus as "copycat"。

But structurally, this is not a copy。

Bethropes holds the model layer through Anthropic, the application layer through Perplexity and Figure, and the algorithm layer through Amazon. Now we're going to build another Prometheus, and we're going to integrate AI into the manufacturing industry, and we're going to put the "Physical World's Executive Level" next. This is a whole inn, from training chips to deployment to factory workshops, and each floor has its own card。

Project Prometheus quietly acquired a company called General Agents some 10 days after its launch. This company is a “computer agent” who can directly operate the entire computer. WIRED later disclosed that it took only four days from the beginning of the takeover to complete the delivery。

Donely's CEO Harsha Abegunasekara says, "General Agens, the real crack is speed, and Ace runs almost instantaneously on your computer." And his company was a competition for General Ages。

IT TOOK ONLY 18 MONTHS FOR BESSOS TO INVEST IN AN ANGEL TO CREATE A SPECIAL FUND AND TO PERSONALLY BECOME CEO. HE'S ACTUALLY BUILDING A BIGGER SYSTEM THAN AMAZON。

Powell Jobs: Low profile

PUT THE THREE OF THEM TOGETHER. POWELL JOBS IS THE LEAST LIKE "AI INVESTOR."。

According to CNBC, citing data from the Private Wealth Data Platform Fintrx, her family office Emerson Corporate has invested in at least nine AI-related start-ups since 2022, involving a total of over $1 billion in financing rounds. That's not the same number as Schmidt or Bethos。

But interesting is the list itself。

Proximie: Telesurgery Linking Platform; Agropos Health: Clinical Data AI; Formation Bio: AI Pharmaceutical `Curipod: A teaching tool for Norway; Mistral: Large Models in France, Europe's only player against OpenAI。

There are no defence, data centres and rockets。

Emerson Corporate’s official network clearly describes the direction of investment: education, energy and environment, digital health, financial science and technology, and the media. AI is just a tool to fit into these themes. She owns most of Atlantic Monthly's shares and is highly skilled in investing in Colombian soft power。

But the one she actually hit, it wasn't even on the line。

After Johnny Ive left Apple in 2019, Powell Jobs voted for his design company LoveFrom through Emerson Collective. Ive later said in an interview with Financial Times: "If it weren't for Laurene, there would be no LoveFrom." A few years later, Ive founded a hardware company called O, specializing in AI equipment, and Powell Jobs went on to pitch. In May 2025, OpenAi bought an io with $6.4 billion in full stock trading, making Ive a billionaire on the books. Emerson Collective also went ashore。

Another investment is also critical, and Emerson Corporate is one of the early investors in Mistral AI. At that time, the French company was the only remaining European fire in the big model。

Put these together, her AI bets are concentrated in two directions, either “using AI to solve specific people's problems” or “reshaping the interaction between humans and machines” (io equipment, Ive design)。

In an assessment, VC Sheet describes Emerson Collactive as “a lLC that deliberately maintains ambiguity, placing venture capital, charity, policy advocacy, art and media ownership under one roof, using any of the most effective tools in grants, policy lobbying or investment”

her philosophy is closer to the old east coast family office, and her influence is more important than return, more important in the long term than in the short term, and the microphone is more important than the spotlight。

Three investment philosophy

PUT THREE LISTS TOGETHER, YOU'LL SEE THREE SETS OF JUDGEMENTS ABOUT THE FUTURE OF AI。

Schmidt gambled on national competitions and infrastructure bottlenecks. In his world, AI was finally determined by "who has the most electricity, who has the fastest rocket, who has the strongest drone." Models are just tickets, real moats in the physical layer. So he personally ran to be the head of Relativity and Bolt, not in return, control。

Bessos gambles on the proliferation of applications at the industrial revolution level. He believed that AI would eventually infiltrate every machine, every plane, every satellite like electricity. So he locked the model layer through Amazon, locked the manufacturing layer through Prometheus, embedded through Expeditions. He's not betting on whether a company can win, but on the whole structure。

Powell Jobs bets on another thing. She'll end up in a gambling position where she can't stand the interactive model. She and Ive repeatedly stressed in interviews with the Financial Times that "human beings deserve better." She cast io, LoveFrom, Medical AI and Education AI behind the same judgement that the largest market in the next decade was “the side effect of repairing the Internet for the last decade”。

Three sets of judgment, three strokes。

Which one? No one knows. Schmidt may have overestimated geopolitical weight in the AI economy. Bessos may have underestimated the financial consumption of the heavy asset model of the “whole house”, a typical example being Prometheus, which has not yet delivered its goods and is already being sent for another $10 billion. Powell Jobs is faced with a more embarrassing problem: O’s equipment was not produced 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 last generation of the Internet collectively turned family funds to AI, it was no longer a little bit of a track. Bolt has raised 150 million start-ups, and Anthropic alone ate the $30 billion that Amazon promised. This volume of capital flows will shape the industrial geography of the next decade。

And who's gonna laugh to the end, wait till 2030 to look back. Until then, the three old Dengs were still at the table, and the chips were still on the table。