In the weeks leading up to the IPO, Cerebras Systems, which had been focused on sprinting towards its public listing, unexpectedly encountered an “intruder” from a capital giant. According to an anonymous source cited by Bloomberg on May 13, 2026, SoftBank and its subsidiary chip design company Arm quietly approached this prominent U.S. company known for its wafer-scale AI chips, hoping to push for an acquisition or merger deal before its public market debut — but Cerebras clearly rejected the offer. At this moment, the market was already preparing for the IPO: the expected valuation cited by the media was approximately $34 billion, and Cerebras was considered one of the most coveted assets in the new round of AI training and inference competition, while the sudden “raid” on the eve of the IPO brought the already tense capital game to the forefront. Three parties kept silent on the details of the negotiations, leaving two unresolved core questions: why did SoftBank choose this moment to attempt a grab for a top-tier AI chip asset, and why did the spotlight-holding Cerebras prefer to go public independently rather than accept an offer from this super buyer holding Arm.

The Night Before the IPO: SoftBank and Arm's Rapid March

This time, it was not just SoftBank with its checkbook, but also its subsidiary chip design company Arm. As a major shareholder of Arm, SoftBank did not choose to let this company, which is listed on NASDAQ with the code ARM.O, operate alone; instead, it initiated contact in the weeks leading up to Cerebras' IPO in a "joint action" manner — in a time when AI training and inference chips are seen as core assets in the new tech cycle, this collaborative move itself was a statement: SoftBank was not satisfied with merely being Arm's financial backer but wanted to push Arm to the front line in the pursuit of top-tier AI chip assets.

The timing of the action seemed to be a meticulously calculated "rapid march." If SoftBank waited until Cerebras landed on the public market before attempting the acquisition, the valuation anchor would be dictated by secondary market sentiment, compressing the negotiation space, while the potential shareholder structure would quickly become complex, turning any controlling acquisition into a long and public tug-of-war. However, by taking action in the weeks preceding the IPO, SoftBank faced a target with concentrated equity and valuation still in play — even if the numbers on the negotiating table were not cheap, at least the chips for control were still held between a limited number of hands, making this much more straightforward for capital eager to seize a position in AI computing power than the approach of "let's see after the IPO."

Placed in the context of SoftBank's recent trajectory, this sudden approach is not surprising. While it has been making large-scale bets on tech companies through platforms like the Vision Fund, it also pushed for Arm to go public on NASDAQ, trying to amplify its voice on both the chip architecture and capital market ends; now, taking another step along this path means using Arm's shell to capture wafer-scale AI chip assets like Cerebras, creating a more complete chain of "design-investment-computing power." However, as of May 13, 2026, all three parties — SoftBank, Arm, and Cerebras — had remained silent about this contact made public by Bloomberg, and in the information vacuum between speculation and reality, this pre-IPO knock seemed more like an early tactical signal.

The Wafer-Scale Monster Cerebras: From Marginal to Coveted

Before being eyed by SoftBank, Cerebras had already made a significant presence in the AI community with its monster-level design of “a chip made from a whole wafer.” Unlike the traditional path of rolling processes and stacking along the GPU route, Cerebras directly transformed an entire wafer into a super-large-scale AI chip, focusing on scenarios that require extreme computing power and bandwidth, such as large model training and inference — the larger the model and the more parameters, the more difficult it is for traditional clusters to manage communication overhead and power consumption, while wafer-scale architecture can use "physically connected" computing power to tell a differentiated story of reduced latency and increased utilization.

The surge in large models has propelled this story from a “technical experiment” to the center stage of the capital theater. In recent years, AI training and inference chips have become core assets in global competition, with Nvidia firmly occupying the throne of general computing power, but the capital market is also seeking specialized paths that could rewrite the landscape — Cerebras stands at this node: viewed as an important player in AI inference and training, simultaneously preparing for IPO and continuously monitored by tech and investment circles. The expected valuation of about $34 billion, cited by media, not only reflects optimism about growth prospects but also signals that wafer-scale dedicated computing power is gaining its own chips and voice in the power landscape dominated by Nvidia; Cerebras is no longer a marginal technological wonder, but rather a target that any entity seeking to restructure the AI chip chain must seriously consider.

$34 Billion or More? The Calculation Behind Rejecting the Acquisition

According to anonymous sources cited by Bloomberg, SoftBank and its subsidiary Arm initiated acquisition/merger discussions a few weeks before Cerebras entered the public market, but Cerebras chose to politely decline, and the IPO process was not publicly altered as a result. This means that, between being “bought outright” and “letting the market price it slowly,” the management preferred to bet on the latter: retaining the company's independent identity and submitting the valuation to a public market that is highly starved for AI computing power assets, rather than capping the price at the negotiating table.

The acquisition conditions remain unknown externally, making it impossible to directly compare the pros and cons of "merger price" and "listing price," leading to a deduction of Cerebras' strategy from the structural differences of the two paths. Merging into the SoftBank-Arm system would yield a guaranteed cash or equity payout, but governance would inevitably be brought into the strategic landscape of a major shareholder, with product routes, customer strategies, and capital expenditures needing to revolve around the controlling party; choosing IPO, however, retains valuation flexibility on the basis of the media-cited expected valuation of $34 billion — if demand during the pricing phase is robust, the potential revaluation of the market cap on the first day and subsequent trades far exceeds that of a one-time transaction. For Cerebras, holding the story of a wafer-scale AI chip, this is not only a vote of confidence in its long-term growth but also a judgment on current market sentiment: $34 billion is just the starting point and not the endpoint it is willing to write on an acquisition contract.

SoftBank's AI Gamble: The Road After Missing Out on Cerebras

For SoftBank, Cerebras is just the latest attempt on a timeline of “making moves before the turning point.” Acquiring Arm years ago and then pushing for Arm's NASDAQ debut are fundamentally the same script: buying crucial infrastructure with big bucks and then amplifying valuation and influence through the capital market. Now, with the demand for AI training and inference chips skyrocketing due to the large model trend, SoftBank, through Arm, has gained significant access to global mobile and some data center CPU architectures, which is widely interpreted as its attempt to seize bargaining chips for computing power in the AI era. The joint acquisition approach with Arm a few weeks before the Cerebras IPO was an effort to extend “architectural entry” into “the heart of training chip computing power,” transforming from just the blueprinter of AI data centers to a player that both designs and supplies core components.

However, this acquisition remains at the stage of “contact being rejected,” with no formal offer or regulatory review having been initiated, signifying a reality for SoftBank and Arm: core assets of the AI training chip ecosystem may not be willing to relinquish control through private negotiations. The loss of Cerebras as a potential target does not immediately diminish Arm's status in CPU architecture, but it leaves SoftBank with a missing piece in its vision of bargaining power in AI data centers for a “wafer-scale AI chip,” which means it remains more in control of standards and interfaces rather than holding leading training chip capacity. In the view of the capital market, Cerebras' rejection of the acquisition and insistence on entering IPO at the expected valuation of about $34 billion provide a clear path for other AI chip startups: first, establish their worth in the public market through independent listing, then negotiate with semiconductor giants, cloud vendors, or capital players like SoftBank. In an environment where capital expenditure and acquisition waves for global AI computing infrastructure continue, this path raises the psychological bottom line for future acquisition offers and forces players like SoftBank attempting to “package future computing power” to either pay a higher premium later or learn to engage in long-term, diversified alliance transactions with more independent and powerful chip companies.

The IPO Bell Will Toll: The AI Chip Capital War Heats Up

SoftBank and Arm's "raid acquisition" being rejected truly exposes a deeper undercurrent: top-tier AI chip assets have become so scarce that time must be raced between primary and secondary markets, with sellers no longer eager to “sell out” at once but instead betting on the valuation game in the public market. Cerebras' insistence on the IPO is essentially placing itself in the spotlight, mandating that future acquisition offers must reference public pricing rather than private negotiations. Yet at this moment, this performance is still in its opening act: as of May 13, 2026, Cerebras' IPO has not been completed, the pricing range and financing scale lack official clarity, and the numbers regarding the share issue price and total financing remain at the level of unverified rumors from single sources; SoftBank, Arm, and Cerebras have also not publicly spoken about acquisition discussions, and Bloomberg's report has pushed the originally closed game into the public discourse without filling the information vacuum. After the IPO, this company will likely face a new round of equity battles: whether SoftBank and its subsidiary Arm will change course to slowly accumulate shares in the secondary market or attempt acquisition again during a price correction remains rife with uncertainty, and other tech giants hungry for computing power may also join the fray. For this long-term capital war surrounding AI chip assets, what will truly warrant attention next are three key variables: whether Cerebras' final issuance pricing and post-listing trading performance will confirm the currently elevated valuation expectations, whether SoftBank and Arm will choose to strengthen their voice in AI computing power by holding shares through the secondary market or restarting acquisition negotiations, and whether regulators and overall market sentiment will continue to allow capital consolidation in the face of larger AI chip mergers or apply brakes out of concerns over monopoly and systemic risk; these three points will determine whether Cerebras stands at the center of the table as an independent player or becomes a key piece in a larger capital map.

Join our community, let's discuss together and grow stronger!
Official Telegram community: https://t.me/aicoincn
AiCoin Chinese Twitter: https://x.com/AiCoinzh
OKX Benefits Group: https://aicoin.com/link/chat?cid=l61eM4owQ
Binance Benefits Group: https://aicoin.com/link/chat?cid=ynr7d1P6Z

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。