Nvidia rival Cerebras discloses US IPO filing as AI boom drives listings

Cerebras Systems has filed for a U.S. IPO again after withdrawing its prior filing in October, marking a renewed path to the public markets for the Nvidia rival. The AI chipmaker reported revenue of $510 million for the year to December 31, up from $290.3 million, and swung to $1.38 per share in profit from a $9.90 per-share loss. The company also said it has cleared a U.S. national security review tied to G42, reducing a key regulatory overhang.

Analysis

Cerebras coming back to market is less a single-company event than a referendum on whether capital markets will finance the next layer of AI infrastructure without requiring Nvidia’s full stack. If investors buy the inference-first, memory-light thesis, the marginal capital dollar shifts toward workload-specific chips and away from the most expensive part of the AI bill of materials: HBM, advanced packaging, and the suppliers that sit on Nvidia’s demand chain. That creates a subtle but real second-order risk for NVDA: even if overall AI spend keeps rising, a successful alternative architecture can dilute its share of incremental wallet growth over the next 12-24 months.

The bigger immediate catalyst is sentiment: a clean IPO reopen in AI hardware can unlock a wave of private-to-public re-ratings across the ecosystem, which usually compresses discount rates for adjacent private names and revives venture exit expectations. But this is also where the setup can get noisy — a reopened IPO window can be bullish for the category while still being bearish for incumbents if the market decides differentiated chips are finally investable at scale. In that scenario, the losers are not just direct competitors; the broader supply chain tied to memory intensity and premium packaging could see valuation pressure as investors re-underwrite growth assumptions.

The main near-term tail risk is execution and customer concentration. A single-anchor customer narrative can support a premium multiple in a frothy tape, but it also means any delay in deployment cadence or procurement pacing can hit the stock hard over a 1-2 quarter horizon. Separately, the prior foreign-investment scrutiny is a reminder that export-control and national-security headlines can reprice these names quickly, especially if capital markets start extrapolating more cross-border AI partnerships.

Consensus seems to be treating this as a simple sign of AI strength; the more interesting view is that the market may be underestimating how quickly a valid second-source narrative can re-rate the competitive map. That doesn’t make Nvidia structurally impaired, but it does mean the market may have to pay more attention to product segmentation and end-market mix than to headline unit growth alone.

Key Decisions for Investors

• Reduce NVDA exposure tactically over the next 1-3 weeks into IPO enthusiasm; use any strength to trim, because the risk/reward has shifted from monopoly premium to share-loss narrative in inference.
• Initiate a short-dated NVDA / long basket of AI infrastructure beneficiaries pair trade only if the IPO prints well and trades up on day 1-5; this captures rotation away from the incumbent without making a pure market bet.
• Buy call spreads on a public HBM or advanced-packaging proxy with 3-6 month tenor if the IPO is well received; the trade benefits if investors continue to fund AI capex but prefer non-Nvidia picks and memory-intensive supply chain names catch the next bid.
• For event-driven accounts, wait 2-4 weeks post-pricing before touching Cerebras-adjacent beta; initial tape likely overweights scarcity value, but lockup/filing cadence and disclosure quality will matter more once the first halo fades.
• If you want to express the contrarian view, sell downside protection on NVDA only after confirming the IPO does not spur a broader AI hardware rotation; otherwise keep hedges intact because the catalyst is symmetric but the narrative skew is now less one-way.