US Stock Storage Sector Rallies Across the Board, SanDisk Up by 4.54%

According to people familiar with the matter, London-based AI cloud infrastructure company Nscale is informing potential investors that its total contracted revenue stands at approximately $51 billion, and it plans to hold its initial public offering (IPO) in the U.S. as early as September 2026. In its presentations to potential investors, Nscale disclosed its growth trajectory: its revenue surged to over $100 million in the second quarter of 2026, a sharp rise from around $37 million in the first quarter and roughly $33 million for the full year of 2025. Nscale had previously operated on a relatively small revenue scale; the size of its contracted contracts disclosed ahead of this IPO, along with its accelerated quarterly revenue growth, will serve as key benchmarks for the market to assess its valuation.

Uniswap co-founder Hayden Adams has published a post detailing an auto-compounding liquidity technology design he contributed to pools.trade, describing it as "quite elegant". The core logic of the mechanism is: once a liquidity position is deposited into a smart contract, anyone can withdraw all unclaimed fees from that position, provided they simultaneously increase the size of the liquidity position by 0.2%. As fees accumulate over time, once their value exceeds 0.2% of the liquidity, searchers are naturally incentivized to add 0.2% of liquidity to claim the fees, forming an auto-compounding cycle that requires no external intervention. Adams called the mechanism "super simple and clean", noting it is built on Uniswap’s token vault concept. He further pointed out that the design also works for auto-compounding of regular Uniswap LP positions, so the team has decided to include it in Uniswap’s roadmap. This means Uniswap liquidity providers can look forward to native auto-compounding functionality in the future.

Nvidia is evaluating adjustments to the HBM configuration of its next-generation AI GPU Rubin Ultra, planning to launch a version with lower video memory than originally scheduled to ease production pressure caused by the shortage of high-end HBM supplies. Sources familiar with the matter revealed that Nvidia has tested at least three versions with different memory capacities over the past few weeks, some adopting lower memory specifications. This means that even Nvidia, which holds a dominant position in the GPU market, has had to make compromises between product specifications and supply capabilities. The Rubin Ultra is positioned above the upcoming Rubin series set for mass production, and was originally planned to be equipped with higher-capacity, higher-bandwidth HBM to boost performance for large model training and inference. If the low-memory configuration is ultimately adopted, customers will need to deploy more GPUs when running large AI workloads such as large language models, leading to higher system costs and increased cluster complexity. For cloud vendors including Microsoft, Meta, Amazon, and Google, which are continuously expanding their AI capital expenditures, data center construction costs may rise further. Nvidia’s move also reflects that despite the company holding the strongest bargaining power in the supply chain, HBM supply constraints remain a core bottleneck for AI computing power expansion. High-end HBM production capacities at SK Hynix and Micron have already been tight, and Nvidia’s proactive configuration reduction further confirms that the HBM supply-demand imbalance will be difficult to ease in the short term.

AMD announced it has entered into a definitive acquisition agreement with AI inference chip firm Taalas. Founded in 2023 and headquartered in Toronto, Canada, Taalas focuses on optimizing inference data flow to drastically reduce compute and memory bottlenecks stemming from general-purpose architectures, enabling highly optimized AI inference capabilities. AMD plans to integrate Taalas’ technology into its full-stack AI platform—including AMD Helios rack-scale solutions, AMD Instinct GPUs, AMD EPYC CPUs, and AMD ROCm software—to deliver system-level solutions. Vamsi Boppana, senior vice president of AMD’s AI Group, noted that Taalas’ technology and world-class engineering team will further strengthen AMD’s AI product portfolio via differentiated inference performance and efficiency. As AI expands into more real-time, high-volume application scenarios, the acquisition will help customers deploy inference workloads more efficiently across growing use cases. Ljubisa Bajic, co-founder and CEO of Taalas, said the company was founded with the core mission of building hardware around models, and joining AMD will grant access to the scale, engineering resources, and global influence needed to accelerate innovation.

Leading investment management firm Roundhill’s optical module stock ETF (LYTE) delivered strong performance on its debut, logging $72 million in trading volume and a 1.44% gain, outperforming Roundhill’s DRAM ETF’s first-day trading volume. The ETF tracks 10 stocks and carries an expense ratio of 65 basis points. As AI data centers shift from copper connections to faster, more efficient optical links, optical interconnects have emerged as a critical bottleneck in AI infrastructure development. The AI optical module market alone is projected to grow 57% this year, rising from $16.5 billion to $26 billion. LYTE primarily invests in leading global photonics and optics companies, providing investors with concentrated exposure to the sector. Its top holdings include Lumentum Holdings Inc. (LITE) at 15.42%, Coherent Corp. (COHR) at 15.23%, and Chinese A-share listed firms Source Photonics, CICT, and SkyLight Communication, which account for 14.59%, 14.22%, and 7.9% of the fund respectively.

Tech companies are taking on massive debt to fund AI investments, but even in this smaller segment, investor caution toward such debt is growing. According to people familiar with the matter, two of the three recent commercial mortgage-backed securities (CMBS) deals for data center financing were forced to widen their pricing spreads from initial discussed levels to attract sufficient demand, including CMBS issuances from KKR-backed CyrusOne and Blackstone-backed QTS Realty Trust. Meanwhile, the risk premium for data center-linked CMBS has risen broadly over the past 12 months. UK data center operator Pure Data Centres (Pure DC), backed by Oaktree Capital, scrapped its planned record €1 billion bond issuance in mid-July in favor of bank financing. At the time, the company was marketing the unsecured bond, but signs of weakening demand for AI-related data center debt had emerged—compounded by a sharp drop in CoreWeave’s stock and bond prices following news that Meta was building its own cloud infrastructure. Investors grew cautious about terms, and the company ultimately deemed concurrent bank loan terms more favorable, leading it to cancel the bond sale. This pullback has been dubbed the "Luddite trade" by the market: investors now view AI data centers as risky assets comparable to traditional office and retail properties, worrying about overbuilding, tenant concentration, and asset depreciation from technological obsolescence. In contrast to past enthusiasm for AI-themed debt, investors are tightening pricing standards for data center mortgages, demanding higher risk compensation. This trend creates a delicate tension with the AI capital expenditure boom: tech giants are still ramping up investments, but price signals from the debt market are starting to sound a cautious warning.