A Near-Trillion Valuation Built on Compute, Not Just Code
Anthropic’s reported bid to raise at least USD 30 billion (approx. RM138 billion) at a valuation above USD 900 billion (approx. RM4.14 trillion) marks a stark shift in how frontier AI funding is being framed. This is no longer a classic software growth story; it resembles an industrial buildout, where capital is a proxy for securing scarce AI compute infrastructure. Earlier, Anthropic was reportedly valued at USD 380 billion (approx. RM1.75 trillion) after a USD 30 billion (approx. RM138 billion) round, yet investor conversations have already leapt far beyond that level. Such rapid escalation would look reckless in traditional tech, but in frontier AI it reflects a race for GPUs, data centers and power. Demand for Claude will only matter if Anthropic can lock in enough chips and cloud capacity, and the funding narrative increasingly doubles as a supply-chain strategy rather than a mere valuation milestone.
Why Anthropic Needed SpaceX to Lift Claude Usage Limits
Anthropic’s compute partnership with SpaceX, via xAI’s Colossus 1 supercomputer, turns an infrastructure deal into an immediate product story. The company explicitly linked the agreement to higher Claude Code usage limits and increased Claude API rate limits effective May 6, framing the extra capacity as a direct benefit for developers and premium subscribers rather than a distant backend upgrade. Anthropic plans to apply this new compute supply to Claude Pro and Claude Max tiers, reinforcing that paying users sit at the front of the queue when capacity is tight. Dedicated access to Colossus 1 matters because it is a defined allocation, not a vague overflow arrangement. That reserved slice can boost training throughput, reduce queue pressure and absorb traffic spikes. In other words, the SpaceX deal is less about branding and more about giving Anthropic predictable headroom to raise Claude usage limits without degrading reliability.
From Cloud Contracts to Supercomputers: The New AI Supply Chain
Anthropic’s recent deals illustrate how AI compute infrastructure is becoming inseparable from capital raising. Amazon has pledged to invest USD 5 billion (approx. RM23 billion) immediately, with the potential for up to USD 20 billion (approx. RM92 billion) more, tying that financing to access to as much as 5 gigawatts of current and future Trainium capacity and long-term AWS commitments reportedly exceeding USD 100 billion (approx. RM460 billion). Google has reportedly committed USD 10 billion (approx. RM46 billion) at a USD 350 billion (approx. RM1.61 trillion) valuation, with as much as USD 30 billion (approx. RM138 billion) more contingent on performance. These aren’t ordinary venture rounds; they are supply contracts embedded in equity deals. Combined with the Colossus 1 allocation, Anthropic is assembling a diversified compute portfolio that spans hyperscale cloud and specialized supercomputers, seeking resilience against the bottlenecks that increasingly define the AI supply chain.
Infrastructure Scarcity Is Reshaping AI Competition
The scramble for AI compute infrastructure is rewriting competitive dynamics among frontier AI companies. OpenAI, Google, Meta and xAI are all vying for high-end accelerators and power-hungry facilities, turning GPUs and data centers into strategic assets as critical as models themselves. Anthropic’s SpaceX deal, combined with enormous cloud-linked financing, shows how Claude must compete not only for users but for chips, energy and long-term capacity reservations. Access to a named system like Colossus 1, with more than 220,000 NVIDIA GPUs including H100 and H200, underscores how specific hardware allocations can translate into higher reliability, faster training cycles and more generous Claude usage limits. Infrastructure scarcity is also pushing AI firms to explore unconventional partnerships beyond traditional cloud providers. In this environment, valuations reflect not just software potential but the extent to which a company has secured durable, scalable compute pipelines for the next generation of frontier models.