Cerebras Files Nasdaq IPO (CBRS)

Apr 18, 2026 | Stocks | Polyminute News | No comments
Cerebras Files Nasdaq IPO (CBRS)

Cerebras has formally filed its S-1 to list on Nasdaq as CBRS after scrapping 2025 IPO plans, revealing 2025 revenue of $510M (up 76% YoY), first full-year profit of $87.9M, and a $24.6B backlog. A landmark OpenAI partnership—$1B loan, warrants, and 750MW+ capacity through 2028—plus an Amazon equity tie-up underscore its pivot from chip sales to high-margin cloud AI services amid intensifying competition with Nvidia, hyperscalers, and CoreWeave.

Cerebras has reignited its IPO process with a Nasdaq S-1 filing under ticker CBRS, marking a material de-risking of its path to public markets after withdrawing paperwork in 2025. The filing discloses transformative 2025 financials: revenue reached $510 million (76% YoY growth from 2024), flipping from a $485 million net loss to an $87.9 million net profit. Backlog stands at $24.6 billion in remaining performance obligations, with 15% scheduled for recognition in 2026-2027—providing unusually high revenue visibility for an AI infrastructure player still in scale-up mode.

Customer concentration has evolved but remains notable: the Mohamed bin Zayed University of Artificial Intelligence (UAE) accounted for 62% of 2025 revenue, while G42 (Microsoft-backed UAE entity) contributed 24%, down sharply from 87% in H1 2024. This reflects deliberate diversification and a strategic pivot from selling discrete Wafer Scale Engine 3 chips to operating them inside Cerebras-owned/operated data centers as a full-stack cloud service. The model mirrors CoreWeave’s approach but leverages Cerebras’ architectural edge—claimed higher speed and lower cost-per-token for inference and query-heavy workloads versus traditional GPUs.

The anchor relationship is with OpenAI: a multi-year deal committing up to 750MW of compute capacity annually from 2026-2028 (valued >$20 billion), with an option for an additional 1.25GW through 2030. OpenAI provided a $1 billion loan at 6% interest (repayable in cash or services) and warrants for up to 33.4 million Class N shares that vest fully only upon 2GW of take-or-pay volume. The contract is material enough that Cerebras explicitly flags it as “a substantial portion of our projected revenues over the next several years,” while retaining termination rights for OpenAI if SLAs are missed. A separate March 2026 deal with Amazon includes cloud service enablement and a $270 million purchase of Class N stock.

Cerebras now competes directly with Amazon, Microsoft, Alphabet, Oracle, and CoreWeave on the cloud side while still positioning its chips as a GPU alternative. It does not yet own its data centers but signals future capex in that direction. Recent private financing (February 2026: $1B round at $23 billion valuation) and a new Morgan Stanley revolving credit facility (up to $850 million post-IPO) further bolster liquidity. With 708 employees and founder/CEO Andrew Feldman’s track record (SeaMicro exit to AMD), the filing positions Cerebras as one of the few pure-play AI infrastructure IPO candidates in a market starved for scaled, profitable growth stories.

01

First-Order Effects

Obvious, immediate impacts
  • Nasdaq CBRS listing provides immediate liquidity to investors and employees at a likely $20B+ valuation, crystallizing paper gains from the February 2026 round.
  • $24.6B RPO and 2025 profitability de-risk near-term revenue, triggering positive pre-IPO sentiment for AI compute pure-plays.
  • OpenAI $1B loan + warrants injects balance-sheet strength and aligns incentives, funding 750MW+ data-center capacity build-out through 2028.
  • Amazon $270M equity commitment validates hyperscaler willingness to co-invest in non-Nvidia alternatives.
  • Customer mix shift reduces G42 reliance while exposing new UAE concentration risk visible to public-market scrutiny.
02

Second-Order Effects

Cross-sector · cross-geography · time-lagged
  • Accelerates capital inflows into wafer-scale and inference-optimized architectures, pressuring Nvidia’s pricing power in latency-sensitive workloads.
  • Heightens power procurement competition; 750MW+ OpenAI commitment signals tightening hyperscale electricity contracts and favors regions with surplus capacity (e.g., UAE, Texas).
  • Spurs secondary financing for AI-cloud operators; CoreWeave, Lambda, and others may face valuation compression or accelerated M&A as Cerebras sets a profitable public benchmark.
  • UAE AI institutions (MBZUAI + G42) gain credibility as credible Western-aligned compute buyers, potentially unlocking more sovereign AI fund flows into non-Chinese suppliers.
  • Retail IPO appetite for AI infrastructure surges; Anthropic/OpenAI IPO pipelines may accelerate, crowding out smaller hardware bets.
03

Alpha Layer — Opportunities

Trades · strategic positioning · business impacts
  • Validates wafer-scale economics as a credible GPU bypass for inference-scale queries—market currently over-indexed on training-only narratives and under-pricing specialized inference margins.
  • Creates a new “compute-as-a-service” oligopoly layer between chipmakers and hyperscalers; OpenAI’s termination rights introduce asymmetric counterparty risk not yet discounted in broader AI supply-chain multiples.
  • Geopolitical tailwind: UAE’s dual-track AI strategy (US chips + sovereign control) becomes a template, potentially rerouting Middle East petrodollars into US-listed AI infra names and away from pure China exposure.
  • Narrative fracture in “Nvidia monopoly” thesis—consensus still views alternatives as niche; Cerebras’ $20B+ locked-in OpenAI revenue proves otherwise and exposes mispricing in NVDA/AMD optionality.
  • Long-term capex optionality: if Cerebras elects to own data centers post-IPO, it evolves into a vertically integrated AI utility play—highest-conviction asymmetric upside versus pure-chip peers trading at 2027 multiples.

// Share Your Analysis