Weekly Pre-IPO Stock Secondary Market Update | May 20, 2026

Cerebras Systems' public market debut at $185 per share, closing its first day up 68%, is among the clearest validation signals the AI infrastructure sector has produced in this cycle. The company builds wafer-scale semiconductor chips purpose-built for inference, an architecture analogous to that of Groq, which was acquired by Nvidia in December, and its concentrated exposure to Middle Eastern customers did nothing to dampen market enthusiasm at the open. The performance should reframe how investors think about the inference layer: demand is not a theoretical thesis but a live market event, and public investors who have been locked out of the AI infrastructure buildout are showing they will pay a premium for any credible access point. GroqCloud, which operates an LPU (language processing unit) inference platform and recently grew from 2 million to 3.5 million developers, represents the private analog to what Cerebras just demonstrated publicly. The key distinction for investors is that GroqCloud is not a GPU rental business in the mold of Coreweave; it is an inference-as-a-service model where customers purchase token throughput rather than raw compute hours. That distinction matters for margin profile, customer stickiness, and the size of the addressable market. The Cerebras IPO is the data point that confirms this segment of AI infrastructure has crossed from private speculation into public market consensus.

SpaceX's April 1st confidential S-1 filing with the SEC has generated an outsized volume of market speculation, but the only fact in the public record is the filing itself. (SpaceX filed their S-1 publicly on Wed - May 20 this week. We recorded this on May 19.) No valuation, no offering size, and no IPO date have been disclosed by the company. Polymarket currently prices a 61% probability that SpaceX opens its IPO above a $2.2 trillion valuation, a figure that reflects prediction market sentiment rather than any data released by the company. The most instructive framework for evaluating that number is the hyperscaler analogy: Amazon went public in 2000 as an online bookseller and became one of the world's largest cloud infrastructure providers two and a half decades later, a transition no analyst had modeled at the time of its IPO. The case for SpaceX eventually reaching $10 trillion in market capitalization rests on the same logic, that SpaceX will monetize space in ways that are not yet visible, just as Amazon's infrastructure buildout unlocked an entirely new business model. The company's current asset base in launch capability and orbital infrastructure gives it a structural moat that no competitor can replicate on a comparable timeline. Investors evaluating the IPO should hold those long-range scenarios separately from the near-term offering mechanics, where the absence of official pricing data means any valuation discussion remains speculative until the company communicates directly with the market.

Starlink's addressable market is systematically underestimated by investors who evaluate it through the lens of US broadband competition. Approximately 2.3 billion people globally have no internet access at all, the bulk of them in Africa, parts of South America, and parts of Asia, and that population represents Starlink's primary growth vector rather than a secondary or edge use case. The revenue model most likely to define Starlink's long-term unit economics is not subscriber fees but a transaction layer analogous to Apple's App Store, where Apple collects roughly 30% of all revenue transacted through its platform, including from AI subscription services flowing through iOS. Starlink, as the connectivity backbone for populations previously unreachable by terrestrial networks, is positioned to take a similar percentage of every commercial transaction that flows through its network. The scale of that opportunity is a direct function of the number of people brought online, and 2.3 billion new internet users is not a niche scenario. Starlink's contribution to SpaceX's long-term valuation is therefore not primarily a function of its current subscriber base but of the structural position it is building in global commerce infrastructure.