Anthropic

Controlling 60% of global compute gives Amazon, Google, and Microsoft a structural tool to throttle Anthropic and OpenAI's growth — buying time for their own models to catch up, mirroring how Friendster's capacity constraints let Facebook overtake it.

Alphabet owns ~14% of Anthropic and 6-10% of SpaceX, plus stakes in Stripe and other private companies — hidden balance sheet value worth several hundred billion dollars beyond the operating businesses.

Alphabet owns roughly 14% of Anthropic and 6-10% of SpaceX, plus stakes in Stripe and others — a several-hundred-billion-dollar equity portfolio sitting on the balance sheet that isn't reflected in operating results.

SK Telecom's 2023 $100M investment in Anthropic is now worth over $2B after the latest funding round, and the stock has gone vertical, up 75% year-to-date.

SK Telecom's $100M 2023 investment in Anthropic is now worth over $2B at the latest round valuation, with the stock up 75% YTD — offering public-market access to private AI compounding.

Anthropic's Claude Mythos reportedly found thousands of undetected bugs including zero-days hidden for two decades — implying every currently deployed cybersecurity product failed to detect vulnerabilities that a frontier AI found immediately.

Salesforce, Microsoft, Palo Alto, CrowdStrike and other SaaS incumbents are down 25-50% on AI disruption fears, but they have record earnings, cash flow to fund AI R&D, and entrenched customer relationships.

Apple collects a 30% first-year / 15% ongoing revenue cut on ChatGPT and Claude usage within iOS, while its hardware-centric model qualifies as 'Halo' with negative CapEx change year-over-year.

A BitTensor subnet project reached 80% of Claude 4's coding capability in 45 days for ~$1M in TAO rewards, demonstrating that decentralized crypto-incentivized training could undercut frontier labs that need tens of billions in capital.

Despite releasing a strong new model and committing $21B to CoreWeave infrastructure, Meta lacks the distribution channels and enterprise relationships to justify its AI spending through advertising improvements alone.

Anthropic's new deal to run on Google TPUs and Broadcom chips signals that the dominant AI company of the moment is choosing custom silicon over NVIDIA GPUs, suggesting AI startups may become less reliant on NVIDIA hardware over time.

By providing cloud infrastructure and TPU chips to competitors like Anthropic while owning 14% of the company, Alphabet wins regardless of which AI model dominates — a multi-way bet on the entire AI ecosystem.

With a 14% stake in Anthropic, TPU chip supply deals, and a strategy of hosting rival AI models as their "landlord," Alphabet profits from multiple outcomes — though AI growth may partly backfill search revenue losses rather than drive pure incremental gains.

Lou Whiteman flags that Alphabet's AI revenue is partly "backfilling what they could be losing" if search declines — unlike pure-play AI beneficiaries, Alphabet's AI growth is a structural hedge, not unbridled incremental expansion.

Anthropic's new deal to run on Google's TPU chips rather than NVIDIA GPUs signals that hyperscalers' custom silicon could become a viable alternative, potentially reducing AI startups' reliance on NVIDIA hardware over time.

At 125x trailing revenue with only 20% growth, SpaceX's IPO valuation dwarfs Google's at-IPO pricing by over 10x while growing revenue 10x slower. The 30% retail allocation exists because institutions won't pay this price.

Despite dominant launch economics and Starlink's growth, the IPO prices at 125x trailing revenue on just 20% growth — compared to Google's IPO at 10x revenue with 240% growth. XAI has negligible enterprise AI market share, making the AI narrative hollow.

At 125x trailing revenue with 20% growth, SpaceX's IPO valuation is 10x richer than Google's was at IPO despite 10x slower growth. The 30% retail allocation signals institutional reluctance to pay this price.

AI built by these three mega-IPOs will cannibalize existing software moats, causing tech multiples to converge down toward non-tech PEs — a painful repricing for incumbent software businesses.

The market cannot absorb trillions of dollars of new supply; SpaceX will fare best as the first IPO, and each subsequent offering faces diminishing demand as selling pressure from long-held private shares overwhelms buy-side capital.