Anthropic's $2T IPO, Zuck's AI Manifesto, Nvidia's $500B AI Bet, Grok's Comeback

What It Covers

Chamath, Sacks, and investor Gavin Baker analyze Anthropic's projected $2 trillion IPO targeting October 2024, with annualized revenue between $100–120 billion after 10x year-over-year growth for three consecutive years. They also examine Zuckerberg's 6,500-word open-source AI manifesto, NVIDIA's $500 billion GPU financing initiative with Goldman and BlackRock, and Grok 4.6's emergence as a credible frontier model competitor.

Key Questions Answered

• •Anthropic Revenue Trajectory: Anthropic's annualized run rate ends 2024 at $100–120 billion after three consecutive years of 10x growth. The panel consensus projects 2025 exit ARR at $400–500 billion — roughly 4x growth — constrained primarily by physical infrastructure limits (energy, turbines, data center construction) rather than demand. At a $2 trillion IPO valuation, that represents 16–20x forward sales, a fraction of Palantir's and SpaceX's comparable multiples at listing.
• •Frontier vs. Open-Source Pricing Power: Anthropic and OpenAI maintain pricing premiums by staying six months ahead of open-source competitors. That lead is their entire business model. If regulatory approval requirements resembling FAA or FDA timelines (five-plus years per model) were imposed, that six-month advantage would evaporate instantly, commoditizing their tokens and collapsing revenue. Investors should monitor any regulatory framework proposals as an existential risk signal for frontier lab valuations.
• •NVIDIA's GPU Securitization Model: NVIDIA is partnering with Goldman Sachs, BlackRock, KKR, Apollo, and Blackstone to create asset-backed financing for GPU clusters — treating compute like airplane financing, where the asset itself (not just the borrower's creditworthiness) secures the loan. NVIDIA provides residual value guarantees on GPU rental rates, lowering financing costs. In return, NVIDIA receives revenue-share above a floor threshold, potentially making it a capital-light cloud business generating royalty-like income.
• •Compute Lifespan Extends Investment Thesis: CoreWeave reports renting Ampere-generation GPUs (released 2020) at economically profitable rates through 2029 — a nine-year asset life. This extended lifespan directly supports the securitization model, since lenders need confidence that GPU compute retains rental value long enough to repay loans. Investors evaluating AI infrastructure should use nine-year depreciation schedules rather than the three-to-four-year assumptions common in earlier hyperscaler debates.
• •Anthropic IPO as Market Pace Car: Once Anthropic files its S-1 and begins reporting quarterly earnings, it becomes the primary demand signal for the entire AI supply chain. The revenue flow runs: Anthropic pays SpaceX roughly $50 billion per gigawatt for compute; SpaceX pays NVIDIA $30 billion for chips; NVIDIA pays TSMC, Micron, and SK Hynix downstream. Any revenue deceleration in Anthropic's quarterly results will propagate through every layer of that chain, making their earnings the single most watched data point in tech.
• •Centralized vs. Decentralized AI Framing: Zuckerberg's manifesto reframes the AI safety debate from "open vs. closed" to "centralized vs. decentralized." The effective altruist position holds that AI is too dangerous to distribute; Zuckerberg, Musk, and Jensen argue it is too dangerous to centralize. Historical precedent consistently favors decentralization. Practically, any attempt to cartelize frontier AI among two or three US companies while restricting open-source development cedes the open-source layer entirely to China, undermining US competitiveness regardless of safety intent.
• •Grok 4.6 Breaks Frontier Duopoly: Databricks benchmark data places Grok 4.6 above GPT-4.5 on capability-per-dollar metrics, ending the Anthropic/OpenAI two-player frontier dynamic. The turnaround took six months and resulted from two specific moves: acquiring Cursor and its engineering team, and installing SpaceX operational leadership at xAI. A larger Grok 4.7 model is weeks away. Elon simultaneously holds a 90-day cancellable compute contract with Anthropic, giving him optionality to redirect capacity to xAI if Grok's demand warrants it.