The AI Valuation Cycle

Canonical Synthesis

Author: terry-tang | Last updated: 2026-04-19

The AI valuation cycle of 2023-2026 is best understood as the interaction of three distinct but mutually reinforcing forces: Nvidia's supplier economics establishing the financial scale of the buildout; genuine and rapidly growing revenue at frontier AI labs providing a revenue anchor for valuations; and a concentration of sovereign and strategic capital creating demand pressure that has consistently pushed valuations ahead of conventional revenue multiples.

Nvidia is the cycle's foundation. The $130.5 billion in FY2025 revenue — at 70%+ gross margins — represents the financial substrate on which everything else rests. Every training run, every inference cluster, every AI product serving millions of users runs on hardware that predominantly flows through Nvidia. The $11 billion Blackwell revenue in a single quarter, described as the fastest product ramp in company history, confirmed that the buildout was accelerating rather than plateauing. Nvidia's position as the essential supplier to both the labs and the hyperscalers gives it extraordinary pricing power, and its strategic investments in Thinking Machines, xAI, OpenAI, and Anthropic extend that position into equity stakes in its largest customers.

OpenAI's revenue trajectory is the cycle's most important validation signal. Growing from roughly $3 billion ARR at the start of 2025 to $20 billion by year-end — and reaching $2 billion monthly by March 2026 — represents one of the fastest enterprise revenue ramps in software history. This revenue growth is real and verifiable; it reflects genuine adoption by consumers (900 million weekly active users) and enterprises (8 of Fortune 10 for Anthropic; comparable enterprise depth for OpenAI). Valuation multiples of 35-40x forward revenue are ambitious, but they are anchored in a revenue base that has consistently exceeded forecasts.

Anthropic's trajectory from $1 billion to $14 billion ARR in fifteen months, with Claude Code alone generating $2.5 billion ARR, demonstrates that the revenue story is not limited to OpenAI. Coding tools — the clearest single category where AI has produced measurable productivity gains — appear to be the primary driver. This is significant because it suggests the revenue growth is grounded in genuine enterprise productivity rather than speculative consumer adoption.

The Thinking Machines $2 billion seed round at $12 billion valuation represents the cycle's most extreme expression of talent premium. A company five months old, with no product and no revenue, achieved a valuation comparable to large established software companies entirely on the basis of founding team credentials. The capital environment of 2025 made this possible; the question is whether the valuation will be validated by commercial outcomes or will represent a speculative outlier.

Sovereign and strategic capital has been the demand accelerant that has pushed valuations beyond what pure commercial analysis would support. GIC, MGX, QIA, PIF, and other sovereign vehicles have invested across multiple competing labs simultaneously — a hedging strategy that maximizes influence regardless of which lab wins. Nvidia, Microsoft, and Amazon have invested alongside their commercial relationships, creating equity stakes that align financial and strategic interests. This concentration of non-return-maximizing capital (sovereigns seeking access and influence, strategics seeking distribution and IP) creates demand at valuations that pure financial investors might not support.

The Arc

2023-2024: The Foundation Layer. ChatGPT demonstrated consumer AI product-market fit. Nvidia's revenue began its explosive growth. OpenAI and Anthropic raised billions on capability promise rather than revenue proof.

Early 2025: Revenue Anchoring. OpenAI reached $10 billion ARR in June, providing the first genuine revenue anchor for frontier AI valuations. Nvidia FY2025 results confirmed the infrastructure buildout was accelerating. The SoftBank round established $300 billion as the valuation ceiling — quickly superseded.

Mid-Late 2025: The Escalation. Anthropic's Series F at $183 billion nearly tripled its March valuation. Thinking Machines raised a $2 billion seed at $12 billion. The PBC restructuring completed. Microsoft's exclusivity ended. The valuation cycle entered a vertical phase.

2026: The Summit (So Far). xAI at $230 billion. Anthropic at $380 billion. OpenAI at $852 billion. Anthropic's $14 billion ARR and OpenAI's $24 billion annualized revenue provide revenue anchoring, but the multiples imply continued rapid growth as a base case. Amazon's AGI-contingent stake embeds AI industry milestones directly into investment terms.

Open Questions

• Do AGI-contingent investment structures — Amazon's $35 billion stake contingent on IPO or AGI — create perverse incentives to declare AGI prematurely, or accelerate IPO timelines regardless of market readiness?
• If revenue growth in 2026-2027 disappoints relative to the growth assumptions embedded in current valuations (35-40x forward revenue), how severe and rapid would the valuation correction be? The labs have no liquid public market — corrections would first appear in secondary markets and then in future funding rounds.
• Are current valuations primarily reflecting compute economics (Nvidia's margins flowing upstream into lab valuations) or capability/product narratives? The answer matters for understanding whether a slowdown in the infrastructure buildout would deflate AI lab valuations independently of revenue.
• What happens to sovereign and strategic capital flows if geopolitical tensions escalate around AI? GIC and MGX holding equity across OpenAI, Anthropic, and xAI creates complex interests that may not remain stable under AI governance negotiations between the US, Singapore, UAE, and other states.
• Is the Thinking Machines $12 billion seed a signal that talent concentration creates durable venture value, or is it a speculative outlier that will be difficult to justify commercially? The answer will become clear when Thinking Machines either raises a Series A or struggles to grow.
• Can retail investors, now holding $3 billion of OpenAI pre-IPO equity, achieve liquidity without an IPO? The answer shapes the political and commercial pressure for OpenAI to go public, which in turn shapes the timeline for Amazon's $35 billion contingent commitment.