On July 16, 2026, the U.S. stock market saw a pre-market decline in the storage sector, dampening the entire tech chain: SK Hynix fell about 6%, SanDisk dropped over 5%, and Western Digital decreased approximately 4%, continuing the pressure from the previous significant correction. According to AiCoin data, during the same time frame, the on-chain address "Caveman" held a long position of approximately $19.39 million in tech semiconductor-related stocks, which has now incurred an unrealized loss of about $7.67 million, further worsening the account withdrawal that had accumulated earlier. While there was collective disappointment in price levels, there remained strong optimism on a logical front—Dan Bin publicly emphasized that the storage sector's adjustment is nearing a phase bottom, and the long-term demand logic of the AI industry remains intact, while JPMorgan has significantly raised the shipment expectations for servers and CPUs from 2026 to 2028, believing that the large number of inference servers needed for companies to implement AI models is the real driving force. Amid the tug of “demand certainty” and “price corrections,” the AI world is also facing a compliance storm: Anthropic publicly accused Zhipu of using Claude and OpenAI models through model distillation, sparking controversy over whether the acquisition of training data violates service terms. On the on-chain side, according to AiCoin data, the address tradeparagon purchased COHR HIP-3 ticker with 500 HYPE tokens, a trade worth approximately $32,900, participating in this emotional amplification through a different technology asset exposure method compared to traditional equities. The intertwining of tech stock corrections, model compliance disputes, and on-chain speculation constitutes the truest tension in the current AI narrative.
Storage Stocks Plunge Pre-Market: Semiconductor Sentiment Turns Sharp
According to AiCoin data, on July 16, 2026, U.S. pre-market trading saw storage concept stocks continue to weaken after a significant previous adjustment, with SK Hynix down about 6%, SanDisk falling over 5%, and Western Digital decreasing approximately 4%, forming a cohesive plummet. These stocks are in key positions within the AI, server, and storage chain, previously highlighted multiple times by institutional reports as "computing power infrastructure" beneficiaries, but now they collectively head downward in the pre-market quotes, adding another blow to the already tense sentiment in the semiconductor tech sector.
For the funds within the market, this is not merely an unexpected volatility of a single stock, but rather reminiscent of an emotional stampede transmitted along the industry chain: continued declines from the storage end imply that the hardware narrative surrounding AI is unlikely to reignite risk appetite in the short term. Before the official opening of pre-market quotes, quick feedback was already seen on derivatives and on-chain, with long positions highly concentrated in addresses associated with Hynix, SanDisk, and Western Digital. Faced with such consecutive plunges, they can only passively endure the account withdrawal. The drop at this moment is not just red numbers on the screen but marks a clear starting point in time for the significant unrealized losses of tech stock longs on-chain that follow.
Caveman's $19.39 Million Heavy Position Under Pressure
According to AiCoin data, the on-chain address "Caveman" has a total long exposure of approximately $19.39 million in tech semiconductor-related stocks, with the storage direction being the main battlefield for losses: Micron Technology, Hynix, and SanDisk collectively contribute approximately $6.45 million in unrealized losses, accounting for about 84% of its current unrealized withdrawal of around $7.67 million. Almost all losses are borne by these three storage and related semiconductor stocks, effectively handing the entire tech long position's lifeline to the pre-market quotes of a single sector.
From an on-chain perspective, the “Caveman” address, due to its scale and concentration, is becoming a prominent anchor point for speculative sentiment in tech stocks. On one hand, heavy investment in the storage sector has rapidly led to unrealized losses, potentially suppressing similar addresses’ short-term risk appetite for tech semiconductor themes; on the other hand, there have not been clear on-chain actions regarding substantial reductions or sector switches, while the divergence in long-term demand logic between the storage sector and AI intensifies. This high concentration exposure may also be seen by some participants as a potential emotional turning point reference rather than just a simple "long failure sample." Without clear information on opening times, leverage multiples, and liquidation strategies, the fate of whether this $19.39 million heavy position evolves into passive risk reduction or counter-cyclical bets on storage and the AI chain remains an open observation variable influencing thematic speculation.
Dan Bin and JPMorgan Stand Firm on AI Demand Expectations
In the same timeframe when the storage sector continues to face pressure pre-market, Dan Bin chose to give a distinctly different judgment at a moment of public low. He publicly stated that the adjustment of this round in the storage sector is "basically nearing a phase bottom." More critically, he believes that the long-term demand logic supporting this sector has not been rewritten by current stock price fluctuations. After a short-term emotional release, assets with stronger performance certainty are expected to repair first. This is a typical top-down allocation framework: First, lock in the long-term demand for the AI computing power chain, then look for mispriced assets amidst cyclical fluctuations, rather than being led by pre-market declines.
JPMorgan's latest report provides hard data support for this logic. The report adjusted upward the shipment expectations for servers, with a projected growth rate of approximately 22% in 2026, further increasing to about 25% in 2027; it also predicts that CPU shipments will rise from approximately 26 million units to about 68 million units by 2028, nearly doubling. The report clearly states that AI inference is becoming the core driving force of server demand, as companies need to continuously purchase a large number of inference servers when implementing models. This three to five-year allocation perspective starkly contrasts with the short-term unrealized losses of on-chain longs. According to AiCoin data, the address “Caveman” has a long position of about $19.39 million in tech semiconductor stocks, currently incurring about $7.67 million in unrealized losses, with Micron Technology, Hynix, and SanDisk contributing approximately $6.45 million in losses. Traditional institutions and fund managers continue to reinforce long-term AI demand expectations when facing valuation corrections, whereas on-chain participants are forced to choose between short-term losses and holding pressures. This divergence in time dimensions is one of the core tensions in the current thematic battle between technology and AI.
Distillation Controversy Ignites Model Compliance
As the on-chain and secondary markets are still engaged in the game of differing AI demand rhythms, a more fundamental conflict erupted on the model side. Anthropic publicly called out Zhipu for its GLM-5.2 model utilizing distillation methods from Claude and OpenAI models. The controversy did not arise from "distillation" itself but rather how the training data relied upon for distillation was obtained— the main skepticism from the outside is whether large-scale scraping or the use of potentially violation-prone data were employed to complete this process. Model distillation, commonly seen in the industry, is essentially using a strong model to teach a weaker model. However, if the "teacher's" output is obtained through unauthorized or terms-violating means, it swiftly transforms from an engineering technique into potential contractual and compliance risks.
Under the shadow of this turmoil, the answer provided by the Claude Code team is to write "no incidents" into the development process itself. The Claude Code Artifact currently supports calling the MCP connectors to read real-time data for development, but its designer Boris Cherny has repeatedly emphasized, "experiences not documented in rules often end up as rework." In the context of model compliance, this reminder is particularly direct: If the data that can be used, how to call third-party capabilities, and how to document and isolate potential high-risk outputs are not clearly written into rule documents, it could lead to liability for an entire training pipeline afterward. Consequently, standardized development and clear usage boundaries are increasingly viewed as important ways to mitigate model compliance risks, as this AI cycle involving capital, computing power, and compliance pulls in different directions. The restraint of "first clarifying the boundaries" has become one of the few uncertainties that can be actively controlled.
On-Chain Speculation Links Tech Stocks and AI Narrative
At the same time when boundaries are being written into rule documents, on-chain funds are also attempting to bring the tech narrative onto the chain. According to AiCoin data, tradeparagon purchased COHR HIP-3 ticker linked to the semiconductor industry with 500 HYPE tokens, a transaction valued at approximately $32,900. This indirect bet on the semiconductor industry through crypto assets, and the "Caveman" directly going long on real-market stocks such as Micron Technology, Hynix, and SanDisk, form two distinctly different exposure models to technology: one follows traditional stock price fluctuations and incurs an unrealized loss of about $7.67 million, while the other tests the derivative scenario's interest in tech assets on-chain with a smaller scale. Currently, there is no public information about tradeparagon's specific speculative motives or subsequent actions, and whether such behavior will evolve into a larger-scale on-chain tech themed trading remains to be cautiously observed. Overall, the pre-market decline of U.S. storage concept stocks, the pressure and unrealized losses on on-chain longs, Dan Bin and JPMorgan's insistence on long-term AI demand expectations, and the compliance disputes over model distillation and data usage boundaries all intertwine into a mutually affecting tech narrative chain on July 16, 2026, where price, position, and rules parallelly reconstruct the profiles of risks and opportunities. Moving forward, the true variables that will determine whether this round of technology + on-chain + AI narrative can stand firm will be whether storage company's performance can meet expectations, whether heavy positions like "Caveman" choose to stop losses or go against the trend by increasing holdings, and the specific implementation rhythm of mainstream AI development tools regarding compliance rules.
Join our community to discuss and grow stronger together!
AiCoin exclusive Hyperliquid benefits: https://app.hyperliquid.xyz/join/AICOIN88
AiCoin exclusive Aster benefits: https://www.asterdex.com/en-US/referral/9C50e2
On-chain Telegram community: https://t.me/AiCoinWhaleData
On-chain community: https://www.aicoin.com/link/chat?cid=N6OVMor5g
AiCoin on-chain Twitter: https://x.com/aicoinwhaledata
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。
