OpenAI Secures 20 Trillion Won Chip Supply.. US AI Semiconductor Startup Sees Surge in First Earnings Post-IPO

Cerebras Systems Inc. (NASDAQ: CBRS) reported first-quarter fiscal 2026 results on June 23, delivering revenue of $193.4 million, up 94% year-over-year, and setting a new quarterly record for core sales. Despite a modest GAAP net loss, the company issued non-GAAP guidance for continued revenue and margin expansion in both the second quarter and full fiscal 2026.

Cerebras also announced a multiyear agreement with OpenAI valued at over $20 billion—approximately KRW 28 trillion—for 750 megawatts of inference compute, and forged a new global cloud distribution partnership with Amazon Web Services. In the second quarter, the company further strengthened its balance sheet by raising about $6.4 billion through its IPO, securing $1 billion in Series H funding, obtaining a $1 billion working-capital loan from OpenAI, and arranging an $850 million revolving credit facility.

Separately, an Eclipse Ventures–affiliated fund and director Lior Susan amended their SEC filing to reflect the one-for-one conversion of their Class B common shares into Class A shares and an in-kind distribution of those shares to fund investors. Post-conversion, they hold roughly 11.58 million Class A shares—about 5.2% of Cerebras’s total common stock—with no plans for additional share sales or changes to their engagement in company affairs.

Cerebras debuted on Nasdaq on May 14, pricing its IPO at $185 per share and raising between $5.5 billion and $6.4 billion—making it one of the largest semiconductor IPOs of the year. Since the listing, leading brokerages have issued buy ratings on the stock, citing Cerebras’s AI-chip strategy.

However, after forecasting adjusted gross margins of 38%–41% for fiscal 2026—below the 47% achieved in Q1—Cerebras shares slid about 14% in pre-market trading on June 24, underscoring short-term volatility tied to its outlook.

Cerebras is an American AI infrastructure company best known for its wafer-scale engine, which uses an entire silicon wafer as a single chip, and its CS-3 system. The platform delivers massive compute performance optimized for inference tasks in generative AI models. In the competitive global AI-semiconductor market—where demand for inference chips and cloud infrastructure is soaring alongside the proliferation of large language models—long-term supply agreements with major customers like OpenAI and AWS are emerging as critical drivers of growth and capacity expansion.