It's Been All Downhill Since Cerebras Systems May IPO: 180% Gains Await According to This Wall Street Analyst

Cerebras shares are down 62.07% from their $311.07 debut price and fell 43.41% over the past month, after Q2 GAAP revenue of $180.11 million missed the $193.55 million consensus and GAAP EPS of -$2.98 missed estimates. Analysts remain bullish, with a $282.55 average target implying 139.5% upside and UBS’s $330 target implying about 179.7%, supported by $25.4 billion in remaining performance obligations and a $6.74 billion cash balance. The key swing factor is whether OpenAI maintains its reported 750 MW, $20 billion-plus commitment or shifts high-speed inference to NVIDIA; the article’s author is cautiously bullish but flags customer concentration and deep GAAP losses.

Analysis

Cerebras: backlog is not yet proof of durable demand

The market is discounting conversion risk, not merely punishing a volatile AI IPO. The key second-order issue is customer bargaining power: if OpenAI can route premium inference between Cerebras and NVIDIA, Cerebras may win volume without retaining attractive pricing or utilization. That would make the backlog less valuable than its headline suggests. Lock-up eligibility adds near-term supply risk, but actual selling—not eligibility—is what matters.

Days to weeks: Treat the OpenAI sourcing report as unverified, but assume analyst estimates and targets may lag it. The next earnings update should clarify customer concentration, backlog conversion, and core gross margin. A miss against the company’s stated revenue range or gross margin below its guided 38%–40% would challenge the recovery case. Conversely, explicit confirmation of OpenAI deployment and evidence of actual shipments would reduce the overhang.

1–3 months: The AMD-linked inference deployment is a proof point, not yet evidence of scaled economics. Track customer adoption and utilization, not launch language. 6–18 months: AWS availability and broader hyperscaler sales could validate a second source to NVIDIA, but the thesis requires repeatable deployments and progress toward the stated 60%+ core-margin ambition. A large RPO balance does not by itself establish timing, cancellation risk, or profitability.

Contrarian point: the analyst upside spread is not an independent valuation signal; revisions have not incorporated the OpenAI report, and consensus may be stale. Prefer a catalyst-driven relative trade over buying the apparent discount. The thesis is falsified if OpenAI confirms material substitution away from Cerebras, backlog conversion slows, or margins fail to recover as deployments scale.

Key Decisions for Investors

• Avoid chasing CBRS solely on analyst targets. Wait for management to address OpenAI sourcing and for reported customer deployments to substantiate backlog conversion; reassess after the next earnings update.
• If OpenAI confirms meaningful GPU substitution, consider a small, defined-risk relative position: short CBRS against long NVDA, sized for CBRS-specific volatility. Cover if Cerebras confirms retained OpenAI workloads and subsequent results show revenue delivery within guidance with core gross margin at or above 38%.
• Keep AMD as a catalyst watch rather than a standalone trade: the Q4 disaggregated-inference deployment matters only if adoption and economics are disclosed. A launch without utilization or customer evidence is not thesis confirmation.
• Monitor lock-up-related trading volume and insider/early-holder sales. Eligibility alone is not proof of selling; persistent heavy supply alongside estimate cuts would strengthen the bearish case.