"NVIDIA Corporation challenger" Cerebras (CBRS.US) made a strong debut, and Mujie quickly positioned herself.

According to the latest transaction disclosure information, "Wood Sister" Cathy Wood's ARK Innovation ETF ARK Next Generation Internet ETF (ARKK) and ARK Next Generation Internet ETF (ARKW) purchased 105,616 shares of newly listed AI chip manufacturer Cerebras Systems (CBRS.US) on Thursday. Based on Wednesday's closing price, the combined position is valued at approximately $32.8 million. Silicon Valley AI chip company Cerebras Systems officially debuted on the NASDAQ on May 14, staging a long-awaited technology IPO frenzy. The company's IPO was priced at $185 per share, significantly higher than the initial expected range of $115 to $125. In the week before listing, due to overwhelming institutional demand - with subscription orders exceeding the number of shares available for issuance by 20 times - Cerebras raised the offering price twice and expanded the offering size. The stock price surged to $350 at the opening, reaching a high of $386 during trading, a 108% increase, triggering NASDAQ volatility halts. It ultimately closed at $311.07, a 68.2% increase from the IPO price. Based on the closing price, Cerebras has a market value of approximately $67 billion; including options and warrants, the fully diluted valuation is around $86 billion. The company sold 30 million shares, raising $5.55 billion, and if underwriters exercise the 4.5 million over-allotment option, the total financing could reach $6.38 billion. This is the largest IPO globally since 2019 and the largest IPO in the US tech industry since Uber Technologies, Inc.'s listing. The IPO attracted subscription orders that were 20 times the number of shares offered. Wafer-level chips: Disruptive technology or expensive art? This is not an isolated position. On the same day, ARK reduced holdings in Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR (TSM) by 41,540 shares (approximately $16.6 million), AMD by 11,510 shares (approximately $5.13 million), and semiconductor testing equipment manufacturer Teradyne by 22,576 shares (approximately $8.2 million). Some analysts describe this reshuffling as a "generational hardware switch" - taking profits from mature semiconductor giants and shifting focus to emerging computing infrastructure challenging the existing landscape with wafer-level chip (WSE) technology. For "disruptive innovation" advocate Cathy Wood, this transaction is a typical projection of her investment logic for 2026 - she previously stated that a new large investment cycle combined with the trend towards deregulation will be significantly beneficial, with core themes including AI, Siasun Robot & Automation, computing power, and new energy. The heavy involvement in Cerebras on its IPO day demonstrates her strong bet on the long-term prospects of AI infrastructure. Cerebras has attracted attention primarily due to its unique technological approach. The company's flagship Wafer-Scale Engine (WSE) technology produces single AI processors directly from a 12-inch wafer, instead of cutting multiple individual chips from the wafer like traditional manufacturers. Its third-generation product, WSE-3, uses a 5nm process, with a chip area of 46,225 square millimeters, integrating 900,000 computing cores and 4 trillion transistors, achieving an AI computing power of 125 petaflops. Compared to traditional GPU clusters, WSE-3's area is 58 times that of NVIDIA Corporation's B200, with on-chip memory bandwidth 2,625 times that of the latter, and inference speed reportedly 10 to 20 times faster than mainstream GPU solutions. CEO Feldman stated that the real turning point came in the first half of 2025: At that time, AI models had become smart enough to have real practical value. Last year, our business experienced explosive growth. In his view, as AI transitions from the training era to the inference era, the WSE architecture, which can generate tokens at lower costs and faster speeds, will present a structural opportunity. However, technological advantages do not guarantee commercial success. Some analysts point out that once there is a defect in the wafer-level chip process, the entire wafer must be discarded, affecting production flexibility and yield. Despite Cerebras claiming to have a self-developed fault-tolerant architecture to bypass defective areas, the stability of large-scale production remains to be verified. Additionally, NVIDIA Corporation maintains an 86% market share in the AI accelerator market in 2025, making it difficult to shake its position in the short term. The financial reality of Cerebras: Accounting magic behind profitability Nevertheless, behind this capital feast, Cerebras' financial data hides concerns overshadowed by market exuberance. In its prospectus, Cerebras reported revenue of $510 million in 2025, a 76% year-on-year growth, and successfully turned losses into profits - with a net profit of $238 million, earnings per share of $1.38, a significant improvement from a loss of $9.90 per share the previous year. Looking at these numbers, it appears to be a high-growth and profitable AI chip company on paper. However, a closer look at the accounting reveals a more complex story. In 2025, Cerebras recorded GAAP net profit of $237.8 million, but $363.3 million of this came from a one-time non-cash accounting adjustment - stemming from a paper gain resulting from the termination of a forward contract liability related to the UAE AI company G42. Excluding this item and adding back stock-based compensation costs, the company's non-GAAP net loss in 2025 was $75.7 million, even widening from $21.8 million in losses in 2024. This means that Cerebras' "profitability" is essentially a result of accounting treatment, rather than the realization of operational profits. In an IPO where valuation has soared to nearly $67 billion, investors should be cautious of these financial details. Shadow of G42 and reliance on OpenAI: Double dilemma of customer concentration Another core risk faced by Cerebras is the extreme concentration of its customer base. In 2024, over 85% of Cerebras' revenue came from the UAE AI company G42. This high dependence directly triggered a national security review by the Committee on Foreign Investment in the United States (CFIUS), leading Cerebras to withdraw its IPO application in 2024. The updated prospectus shows that G42's revenue share has dropped to 24% last year, but the Mohammed bin Zayed University of Artificial Intelligence in the UAE still contributed 62% of revenue. The combined revenue from the two Middle Eastern entities still accounted for 86% of the company's 2025 revenue. CEO Andrew Feldman responded to this by saying, "There are indeed some whale-sized customers in the market, a feature of this market." However, this is more of a business model issue than a market feature - a company targeting NVIDIA Corporation should be selling chips to "everyone," not just a few sovereign-related customers. To address this concern, Cerebras shifted focus to American customers in the past year. In January 2026, the company signed a multi-year cooperation agreement with OpenAI worth over $20 billion, deploying 750 megawatts of high-speed AI inference power for OpenAI, along with a $1 billion loan and options arrangement. In March, AWS announced the deployment of Cerebras CS-3 chips in its data centers, opening up to developers through Amazon Bedrock. However, the shift from G42 to OpenAI essentially replaces one dependence with another. According to Morgan Stanley's forecast, the sales of Cerebras in the next two years are expected to be primarily from OpenAI. The market can always argue that "big customers are premium customers," until the day when those big customers stop placing orders. Valuation debate: Is a 100x price-to-sales ratio reasonable? With a market value of $67 billion, Cerebras has a price-to-sales ratio of 130 times its 2025 revenue of $510 million. For a company still incurring losses at the non-GAAP level, is this valuation justified? There is a significant difference of opinion in the market. A senior research analyst at Renaissance Capital pointed out that the valuation remains reasonable when calculated based on the $185 IPO price and 2028 revenue and EBITDA estimates, but at the current trading price, the long-term valuation is already on the high side. Another analyst bluntly stated, "The market is filled with fantasies about AI, but enthusiasm is often not a rational pricing tool." However, Cerebras has a remaining performance obligation (RPO) of $24.6 billion, including the multi-year contract with OpenAI worth over $20 billion, providing support for its high-growth expectations. If these contracts are successfully delivered and generate income, the current valuation is not entirely without fundamental support. But the prerequisite is that the demand for AI inference does indeed explode as Feldman predicted, and Cerebras' wafer-level chip can sustain its technological advantage in this race for computing power.