Bernstein said in a July 13 report that memory stocks have outperformed semiconductor equipment names by 661 percentage points since June 2025, raising a market concern over whether equipment shares would fall if memory reverses. Looking at a decade of historical data, the firm said the two groups have not been tightly linked over the long run, with correlation staying low and equipment stocks twice posting independent gains during weaker periods for memory. Those divergence windows lasted about two years each, according to the report. Bernstein also pointed to a longer stretch from 2019 to mid-2025, when memory lagged equipment for six years, only catching up in February 2026 after reaching a 36x cumulative gain over 15 years. The report said memory then accelerated and built up excess returns to an extreme level. If mean reversion takes place, equipment names could see relative outperformance. Bernstein cited SK Hynix’s $67 billion expansion plan and said equipment orders represent more certain revenue. It kept Outperform ratings on Applied Materials, Lam Research and ASML, with price targets of $525, $340 and €2,300, respectively.

Morgan Stanley said in a July 10 report that combined capital spending by five hyperscalers — Microsoft, Google, Amazon, Meta and SpaceX — could reach $1.4 trillion by 2028, more than tripling from 2025 levels. The bank also expects available compute capacity to expand from 30GW to 120GW over the same period. Meta was named the bank’s top pick, with its 2027 and 2028 capex forecasts raised to $225 billion and $250 billion. Morgan Stanley highlighted Meta’s API opportunity, saying Muse Spark 1.1 is priced 30% to 86% below peers and that every 100MW of compute could generate about $8 billion in revenue and roughly $1.9 in incremental EPS. The report added that the key question in the compute race is shifting from how much capacity companies can build to how much they can sell. Morgan Stanley maintained overweight ratings on Meta, Amazon and Google.

A TechFlowPost commentary argues that the biggest prize in tokenized equities is not the stock itself, the blockchain, or even the license, but the user-facing distribution point where the buy order is placed. The piece points to three moves that define the current cycle: Robinhood launched its own chain in July 2026 and passed $100 million in total value locked within seven days; Kraken acquired Backed Finance, the issuer behind xStocks, in December 2025; and Telegram pushed tokenized U.S. equities into its messaging ecosystem. The article says this cycle differs from earlier attempts such as Mirror and FTX-era stock tokens because the underlying assets are now described as real rather than synthetic price exposure. It highlights the ERC-8056 standard backed by Robinhood and Superstate for handling stock splits and dividends onchain, as well as Ondo’s work with Broadridge to let token holders vote in shareholder meetings using private keys. TechFlowPost also argues that value is compressing in the lower layers. Public chains and custodians are becoming utility infrastructure, while issuers face fee pressure as more licensed players enter the market. By contrast, the front-end gateway controls listings, user flow, and liquidity. The commentary concludes that as tokenized assets become more standardized, the scarce asset is the interface that captures the user’s tap on the buy button.

Bernstein argued in a July 13 research note that semiconductor equipment stocks do not need a decline in memory names to outperform. The report came after memory stocks had outperformed chip equipment by 661 percentage points since June 2025, a gap that has raised concern over whether equipment makers would fall if memory pulled back. To test that view, Bernstein compared the three largest memory companies — Samsung, SK Hynix, and Micron — with five major equipment makers: Applied Materials, Lam Research, ASML, KLA, and Tokyo Electron. It found the average rolling correlation between memory and equipment was only 0.4 from 2012 to 2018, rising to 0.6 after 2019. Over the same periods, equipment stocks showed a much tighter 0.8 to 0.9 correlation with the Philadelphia Semiconductor Index. The note also highlighted two extended divergence periods: from January 2015 to December 2016, equipment stocks rose 21.9% while memory fell 16.2%; from January 2021 to December 2022, equipment gained 15.3% while memory dropped 34%. Bernstein’s broader point is that if mean reversion sets in after memory’s recent surge, relative returns may shift toward equipment. The firm also cited SK Hynix’s planned 100 trillion won investment in a new fab in Cheongju and said equipment market and EPS estimates could still move higher before 2028.

A July 9 episode of Limitless Podcast, cited and compiled by TechFlowPost, argued that the market may be misreading the AI memory trade. The show said Samsung posted quarterly profit of $58.5 billion, ahead of analyst expectations of $55 billion and above NVIDIA’s $53 billion for the same period. According to the hosts, the bulk of that profit came from memory, especially high-bandwidth memory, or HBM, which has become a critical component for AI chips. The discussion centered on a tightly concentrated supply chain. The hosts said only three companies meaningfully produce HBM: Samsung, SK Hynix, and Micron. They also claimed that every AI inference requires reloading model weights, pushing memory demand far beyond consumer hardware and lifting prices across DRAM, NAND, and HBM. Price increases cited on the show included a 90% jump in Q1, another 50% to 60% in Q2, and a planned 20% increase in Q3. Even so, memory stocks have weakened. Samsung fell 9% on its earnings day, while SK Hynix dropped 15%, despite record profits and still-rising prices. The hosts described that move as a mix of "sell the news" trading and cycle fears. The episode also highlighted SK Hynix’s July 10 Nasdaq ADR listing, which it said raised about $30 billion and was four times oversubscribed.

Securitize’s July 2 listing on the New York Stock Exchange did more than take the tokenization infrastructure firm public. On the same morning, the company also issued its common stock in tokenized form on Solana and Avalanche, with about $270 million in common shares registered on-chain on day one. The article argues that this structure matters far beyond one listing: if equity can be priced continuously, transferred more freely, and administered through specialized software, venture capital’s long-standing one-stop model may start to fragment. Traditionally, a venture term sheet has bundled capital, valuation, signaling, network access, governance rights, and follow-on support into a single package. The piece says tokenized equity could separate those functions. Market-based price discovery may handle part of fundraising and valuation for more mature companies, while cap-table management, exercise tooling, and programmable governance shift to dedicated providers such as Fairmint, Pulley, Magna, and Sablier. It also points to Coinbase’s acquisitions of LiquiFi in July 2025 and Echo in October 2025 as signs of that unbundling. Still, the article does not claim early-stage venture can simply be replaced. It argues that young startups lack the operating history and public information needed for clean market pricing, leaving curation, credibility, and relationship-driven judgment in VC hands. Follow-on investing also remains hard to replicate under current regulatory frameworks, which are aimed at already public companies rather than Series A startups.

A MarsBit analysis argues that the most valuable part of tokenized equities may not be the assets themselves, the blockchain infrastructure, or even licensing. It may be the interface where users hit the buy button. The piece points to three moves in the same direction: Robinhood launched its own chain in July 2026 and topped $100 million in TVL within seven days; Kraken bought Backed Finance, the issuer behind xStocks, in December 2025; and Telegram pushed tokenized U.S. equities into its messaging ecosystem. The article says the current cycle differs from earlier attempts such as Mirror and FTX because the underlying assets are now structured as real securities rather than simple price exposure. It also highlights new standards and partnerships around stock splits, dividends, and shareholder voting, while noting the growth of on-chain RWA markets. At the same time, U.S. regulation remains a hard constraint. The SEC said in January 2026 that tokenization does not change a security’s legal status, leaving many offerings available only outside the United States. Against that backdrop, the report says firms that control user distribution, rather than just issuance or chain infrastructure, may be best positioned to capture the economics of tokenized stocks.

Odaily reported, citing Jin10, that Citi has raised its price target on Western Digital (WDC.O) to $800 from $685. The brief item only disclosed the old and new target prices and did not include the reason for the revision or any additional details. The update was published as a 7x24 news flash.