Cerebras Has a $25.4 Billion Backlog, and One OpenAI Agreement Is Behind Much of It

Cerebras reported Q2 adjusted revenue of $209.9M (+103% YoY) and raised its full-year adjusted revenue outlook to $880M–$890M. Remaining performance obligations (backlog) rose to $25.4B at June 30, largely tied to a December OpenAI agreement for 750MW (plus an option for 1.25GW), with Cerebras expecting to recognize only ~22% of the backlog (~$5.6B) in the 24 months ending June 30, 2028. Despite improved execution (adjusted gross margin up ~900 bps YoY) and ~$8.6B cash post-IPO, the article flags valuation risk versus the slow backlog-to-revenue conversion.

Analysis

The real signal is not demand strength; it is how much of the demand is financeable and executable. A backlog this concentrated and this far out on the curve makes CBRS look more like a project-finance story than a normal semis multiple, which means the equity should trade at a discount to headline contract value until conversion starts showing up in cash flow. NVDA is the cleaner beneficiary of the same AI capex wave because it monetizes capacity faster, with less customer concentration and far less balance-sheet burden; any disappointment at CBRS likely redirects spending back toward incumbent GPU and networking stacks rather than killing inference demand outright.

Near term, the stock can stay supported if management keeps translating backlog into live capacity and if the OpenAI ramp appears linear over the next 2-3 quarters. The reversal risk is execution slippage, customer re-scheduling, or a change in OpenAI’s build-versus-buy mix; any one of those would hit the multiple before it hits reported revenue. Over 6-18 months, the key question is whether this becomes a repeatable multi-customer platform or a one-client financing story; if the latter, the market will re-rate it toward a lower-quality infrastructure vendor multiple despite impressive growth. The consensus seems to be over-anchoring on backlog size and underpricing the denominator risk: when the current valuation already capitalizes several years of perfection, the burden of proof is on sustained OpenAI revenue and margin expansion, not on the order book itself.

Key Decisions for Investors

• Relative-value: long NVDA / short CBRS into strength over the next 1-3 months; thesis is that NVDA captures AI spend with lower execution risk and lower customer concentration. Exit if CBRS shows two consecutive quarters of above-plan OpenAI revenue conversion and margin expansion.
• Avoid chasing CBRS outright after the backlog-driven rerate; wait for a pullback or for proof that at least 25%-30% of quarterly revenue is recurring from multiple customers. If that mix appears, reconsider a smaller starter long.
• If options/liquidity allow, buy CBRS downside via 3-6 month put spreads rather than outright shorting; this expresses the tail risk that backlog conversion slips or customer concentration becomes a valuation overhang. The thesis is invalidated if management raises FY guidance again with no increase in working-capital strain.
• Watch item: compare CBRS quarter-to-quarter OpenAI revenue recognition versus data center capex and financing draw. If revenue ramps but cash burn widens, treat the backlog as lower quality and favor short CBRS versus SMH.