The Infrastructure Premium: Why Public Markets Are Rewarding Compute over Code

The recent reception of the Cerebras Systems IPO represents a fundamental pivot in how public markets value the AI ecosystem. With a proposed valuation reaching $26.6 billion, investors are no longer solely prioritizing software-centric labs or general-purpose SaaS metrics. Instead, the market is signaling a decisive preference for tangible, contract-backed AI infrastructure—a trend that positions companies like Cerebras, Supermicro, and DigitalOcean as the new industrial base of the sovereign AI era.

The Shift from SaaS to Sovereign Infrastructure

For the past two years, the AI investment narrative was dominated by pure-play software models, with valuations tethered to traditional SaaS metrics like ARR and user growth. However, those metrics fail to capture the capital-intensive reality of scaling frontier compute. As detailed in the Cerebras S-1 filing, the company is successfully pivoting from a pure hardware vendor to an infrastructure provider.

This model—what I call "industrial-compute sovereign"—is defined by three pillars: massive upfront capital expenditure (including data center leasing and advanced cooling), long-lived physical installations, and multi-year, contract-backed revenue. Cerebras’s $24.6 billion in remaining performance obligations, anchored by a $20 billion Master Relationship Agreement (MRA) with OpenAI, provides a level of revenue visibility that pure software models simply cannot offer.

Validation via Hyperscale Partnerships

The "infrastructure premium" is not merely speculative; it is being validated by the world's largest cloud operators. The binding term sheet with AWS—the first hyperscaler to deploy Cerebras systems in its own data centers—is a critical endorsement. It signals that Cerebras’s wafer-scale architecture meets the stringent reliability and scale requirements of the cloud incumbents.

This infrastructure-led growth contrasts sharply with the "sovereign AI" integration approach seen at Supermicro (SMCI), which focuses on high-volume rack-level hardware, and the "AI-native cloud" strategy at DigitalOcean (DOCN). While each firm targets a different layer of the stack, all three are benefiting from the same underlying market realization: in the age of AI, compute is the new strategic commodity, and sovereign entities—from national laboratories to defense organizations—are prioritizing control, data locality, and predictable capacity above all else.

The Risk of the Hardware Premium

However, this enthusiasm introduces a new risk profile. Applying high, "growth-oriented" valuation multiples to infrastructure firms creates a potential fragility if asset utilization rates fall below expectations or if capital expenditure requirements continue to climb without corresponding gross margin expansion. While Cerebras reported 76% YoY revenue growth to $510 million in 2025, its business model remains highly concentrated—62% of its revenue came from MBZUAI in 2025.