Report: Meta in Talks to Sell AI…

1. Key Themes

AI Infrastructure Is Becoming a Standalone Business

Big Tech's massive data-center buildouts are now being monetized as third-party cloud services — a structural shift that creates new revenue streams and new competitive dynamics.

"Meta is in talks to lease AI computing power to Anthropic in a deal worth up to $10 billion over two years, potentially giving Meta a new way to monetize its massive data-center buildout as Anthropic scrambles for scarce capacity."

"SpaceX is reportedly in talks to sell the Defense Department billions of dollars' worth of AI computing capacity."

Meta is even hiring AWS infrastructure talent to formalize this: "Meta plans to hire longtime AWS executive Dave Brown to help lead its data-center buildout as the company weighs renting excess AI computing capacity to Anthropic and potentially turning its massive infrastructure spending into a cloud business."

The GPU Dominance Is Being Challenged

Alternative compute architectures — ASICs and non-GPU inference chips — are attracting serious capital as buyers seek cheaper, faster, and more energy-efficient options.

"AI inference cloud startup General Compute has secured a $400 million loan from Upper90 backed by inference chips, a first-of-its-kind financing that reflects growing demand for cheaper alternatives to Nvidia GPUs as open-source models become more competitive."

"General Compute raised $15 million to build the world's first ASIC cloud. By removing the GPU bottleneck, it runs frontier LLMs up to 16x faster than standard GPU clouds. ASIC silicon is also far more energy efficient, so it deploys in air-cooled data centers, making growth and expansion dramatically easier."

Nuclear Energy Is Becoming a Serious AI Infrastructure Bet

Venture capital is flowing into next-generation nuclear power specifically to serve the energy demands of AI data centers.

"Valar Atomics, a three-year-old startup based in El Segundo, CA, that is developing small modular nuclear reactors to power AI data centers, is reportedly in talks to raise a $1 billion round led by Sequoia at about a $6 billion valuation."

Private Market Resilience vs. Public Market Volatility

Private high-growth companies have significantly outperformed their public counterparts on key efficiency metrics through the 2021–2024 correction.

"Among high-growth public cloud companies, the number clearing the Rule of 40 fell from 14 to just 1 between 2021 and 2024. Across private markets, the same fast-growing group only slipped from 38% to 25% of companies. Public signals aren't private signals."

AI Wealth Concentration Is Creating a Philanthropic Backlash — or Vacuum

The AI boom is generating historically concentrated wealth at a moment when elite philanthropic commitments are collapsing, creating social and political risk.

"Neil Rimer said he has 'a strong sense that there will be some sort of a redistribution.' He continued on. 'It'll either be voluntary or it'll be involuntary, but it'll happen, and I hope it's voluntary,' he told me, adding that he thinks tech leaders 'can play a leading role in seeing that through.'"

"One hundred and thirteen families signed [the Giving Pledge] in its first five years, then 72, then 43, then just four in all of 2024...underscoring how out-of-fashion philanthropy has become among some of the richest people in tech."

2. Contrarian Perspectives

The Giving Pledge is effectively dead — and no one in tech seems to care. The standard assumption is that wealthy tech leaders will eventually channel fortunes toward philanthropy. The data says otherwise. The Giving Pledge decline from 113 families in its first five years to just 4 in all of 2024 suggests a generational or ideological break from Buffett/Gates-era norms, at exactly the moment when AI is creating more billionaires faster than any previous technology wave. Rimer's warning that redistribution "will happen" — voluntarily or not — implies he sees this as a systemic risk, not just a cultural curiosity.

Meta may be building the next AWS without intending to. The conventional read on Meta's data-center spending is pure AI product investment. But the Anthropic deal suggests an emergent cloud business: Meta spending aggressively on infrastructure, then leasing excess capacity to competitors. This mirrors Amazon's accidental discovery that its internal logistics and compute could become standalone businesses — potentially more valuable than the original products.

"Meta is in talks to lease AI computing power to Anthropic in a deal worth up to $10 billion over two years, potentially giving Meta a new way to monetize its massive data-center buildout."

China's open-source AI push is a strategic geopolitical move, not altruism. Xi Jinping calling for open-source AI and global governance at a public conference — while Western AI leaders (OpenAI, Anthropic) favor closed systems — positions China as the champion of openness. This is a direct competitive strategy: open-source models commoditize the infrastructure layer where U.S. companies currently dominate, leveling the playing field.

"Chinese President Xi Jinping called for more open-source AI and global AI governance at a Shanghai conference, contrasting China's increasingly open model strategy with the closed systems favored by OpenAI and Anthropic."

3. Companies Identified

Meta | Social media / AI infrastructure | Pivoting toward AI cloud services by leasing compute to Anthropic and hiring AWS leadership

"Meta is in talks to lease AI computing power to Anthropic in a deal worth up to $10 billion over two years, potentially giving Meta a new way to monetize its massive data-center buildout."

Anthropic | AI lab | Customer in the Meta compute deal; facing a constrained GPU supply environment; criticized by Microsoft's Nadella for over-moderation

"Anthropic scrambles for scarce capacity." "Microsoft CEO Satya Nadella criticized Anthropic's Fable model as too 'editorially controlled,' telling engineers its tendency to block some user requests makes little sense."

Databricks | Data analytics and AI infrastructure software | Raising a $3B round at a $188B valuation, signaling continued confidence in AI infrastructure plays

"Databricks...signed a term sheet for a strategic funding round led by previous investor Coatue at a $188 billion post-money valuation."

Valar Atomics | Small modular nuclear reactors for AI data centers | Reportedly raising $1B led by Sequoia; early signal that nuclear is a credible AI energy solution

"Valar Atomics...is reportedly in talks to raise a $1 billion round led by Sequoia at about a $6 billion valuation."

General Compute | ASIC-based AI inference cloud | Secured a $400M asset-backed loan and claims 16x speed advantage over GPU clouds; flagship challenger to Nvidia's dominance

"It runs frontier LLMs up to 16x faster than standard GPU clouds. ASIC silicon is also far more energy efficient, so it deploys in air-cooled data centers, making growth and expansion dramatically easier."

SpaceX | Aerospace / AI infrastructure | Selling AI compute to the Pentagon; stock down ~$1T from peak post-IPO

"SpaceX is reportedly in talks to sell the Defense Department billions of dollars' worth of AI computing capacity." "SpaceX shares fell another 5% today...wiped out roughly $1 trillion in market value from its June peak."

Crypto.com | Cryptocurrency exchange | Raised $400M from Citadel Securities at a $20B valuation — institutional legitimization of crypto infrastructure

"Crypto.com...raised a $400 million round from Citadel Securities at a $20 billion valuation."

Kalshi | Prediction markets | Expanding into pharmaceutical trial outcomes; also flagged alleged insider trading by a White House aide, demonstrating active market surveillance

"Kalshi launched prediction markets for pharmaceutical trials and FDA approvals...despite concerns from bioethicists and researchers that the markets could encourage insider trading, manipulation, or interference with clinical studies."

Shein | Fast fashion e-commerce | Targeting Hong Kong IPO at $40–50B valuation after failed U.S. and London attempts

"Shein has reportedly secured Hong Kong stock exchange listing-committee approval for an IPO that could launch as soon as late August, targeting a $40 billion to $50 billion valuation."

Index Ventures | Venture capital | Generated ~$9B in exits in one year (Figma IPO, Wiz acquisition) on $15B raised over its history

"Last year's exits including Figma's IPO and Google's purchase of the cybersecurity firm Wiz reportedly netted Index roughly $9 billion."

Kind Designs | 3D-printed climate-resilient marine infrastructure | Raised $10M Series A at $70M valuation; backed by Mark Cuban and Kyle Kuzma; intersection of climate tech and coastal infrastructure

"Kind Designs...builds 3D-printed seawalls and shoreline structures that protect coastlines while creating habitat for marine life."

Hyperion Robotics | Robotic microfactories for concrete infrastructure | Raised $7.4M; deploying near project sites for utilities, energy, and water — a capital-light manufacturing model

"Hyperion Robotics...operates robotic microfactories that produce concrete infrastructure parts near project sites for utilities, energy, water, and carbon capture projects."

Mandrake Bio | Programmable gene editing enzymes | Raised $1.7M pre-seed; dual-use for crop improvement and genetic disease treatment — very early but high-optionality biotech bet

"Mandrake Bio...designs programmable gene editing enzymes for seed companies and therapy developers to improve crops and treat genetic diseases."

4. People Identified

Neil Rimer | Co-founder, Index Ventures | Cited as a rare establishment VC voice warning that AI wealth concentration is politically and socially unsustainable

"He has 'a strong sense that there will be some sort of a redistribution. It'll either be voluntary or it'll be involuntary, but it'll happen, and I hope it's voluntary.'"

Dave Brown | Longtime AWS executive, incoming Meta hire | Signals Meta's seriousness about building a real cloud infrastructure business

"Meta plans to hire longtime AWS executive Dave Brown to help lead its data-center buildout as the company weighs renting excess AI computing capacity to Anthropic and potentially turning its massive infrastructure spending into a cloud business."

Satya Nadella | CEO, Microsoft | Publicly criticized Anthropic's model safety guardrails as over-restrictive — notable given Microsoft's own AI investments and competitive positioning

"Microsoft CEO Satya Nadella criticized Anthropic's Fable model as too 'editorially controlled,' telling engineers its tendency to block some user requests makes little sense."

Xi Jinping | President of China | Used a public forum to advocate for open-source AI and global governance, positioning China against the closed-model approach of leading U.S. AI labs

"Xi Jinping called for more open-source AI and global AI governance at a Shanghai conference, contrasting China's increasingly open model strategy with the closed systems favored by OpenAI and Anthropic."

Eric Bahn | VC | Flagging how AI note-taking tools are changing the dynamics of investor-founder meetings and raising legal exposure questions

"VC Eric Bahn tells The Wall Street Journal he now assumes founder meetings will be recorded, part of a broader AI note-taking boom that has some investors changing their Zoom names, warning about legal risks, and questioning whether every conversation really needs to become searchable."

5. Operating Insights

Assume every investor meeting is being recorded — and build your information-sharing strategy accordingly. The proliferation of AI note-taking tools has fundamentally changed the privacy expectations around private conversations. Founders and investors alike need clear policies on what gets shared, logged, and made searchable.

"VC Eric Bahn tells The Wall Street Journal he now assumes founder meetings will be recorded, part of a broader AI note-taking boom that has some investors changing their Zoom names, warning about legal risks, and questioning whether every conversation really needs to become searchable."

Over-moderation is a product liability, not just a safety choice. Anthropic's Fable model is taking fire from a major enterprise customer (Microsoft) for blocking too many user requests. For AI product builders, over-restriction is not a neutral default — it has real customer satisfaction and competitive consequences.

"Microsoft CEO Satya Nadella criticized Anthropic's Fable model as too 'editorially controlled,' telling engineers its tendency to block some user requests makes little sense."

Infrastructure overcapacity can become a business — plan for it. Meta's potential $10B compute leasing deal with Anthropic illustrates that aggressive infrastructure investment, even if initially sized for internal demand, can generate substantial third-party revenue. Operators building capacity-heavy platforms should model monetization of excess capacity from day one.

"Meta is in talks to lease AI computing power to Anthropic in a deal worth up to $10 billion over two years, potentially giving Meta a new way to monetize its massive data-center buildout."

6. Overlooked Insights

Asset-backed lending against AI chips is an emerging new financing category. General Compute's $400M loan from Upper90 — collateralized by inference chips — is described as "a first-of-its-kind financing." This suggests a new structured finance market is forming around AI hardware, which could unlock capital for compute infrastructure companies that don't fit traditional VC or debt profiles.

"AI inference cloud startup General Compute has secured a $400 million loan from Upper90 backed by inference chips, a first-of-its-kind financing that reflects growing demand for cheaper alternatives to Nvidia GPUs."

MLB's AI ban in dugouts reveals how fast AI decision-support is penetrating real-time operations — and how institutions are scrambling to respond. This is a minor sports story on the surface, but the underlying signal is significant: teams were already using live generative AI to guide in-game decisions like substitutions and pitch calling. The speed of adoption — and the institutional panic response — is a preview of similar debates coming in finance, medicine, and law.

"Major League Baseball has banned teams from using custom apps on league-provided dugout iPads after finding that some clubs were feeding live data into generative AI tools to guide substitutions, pitch calling, and other in-game decisions."