A trader spent just $120 to buy 19.1 million units of the meme coin Niulai, reaping an 822-fold return.

Harvard University’s endowment fund’s latest 13F filing shows that as of June 30, its holdings of BlackRock’s spot Bitcoin ETF (IBIT) remained at 3.0446 million shares, valued at approximately $101.4 million — matching the exact number held at the end of the first quarter, ending two consecutive quarters of reductions. Harvard had cut its IBIT position by 21% in Q4 2025, followed by a 43% reduction in Q1 2026. Currently, IBIT accounts for roughly 2.4% of Harvard’s disclosed $4.26 billion U.S. equity portfolio. Its gold-related product holdings are worth more: as of the end of Q2, its combined positions in the iShares Gold Trust (IAU) and SPDR Gold Trust (GLD) total approximately $171.2 million. Harvard previously liquidated its BlackRock spot Ethereum ETF holdings worth $86.8 million, and added no new Ethereum-related positions in the second quarter. Harvard’s largest holding is SpaceX, with 12,935,100 shares valued at $2.21 billion, making it the largest single stock position disclosed in the filing. The total U.S. equity assets Harvard disclosed in this filing are approximately $4.3 billion, meaning SpaceX’s stake accounts for 52% of that total. Among other institutions, Abu Dhabi sovereign wealth fund Mubadala and the Abu Dhabi Investment Council each maintained their IBIT positions unchanged at 14.7219 million shares and 8.2187 million shares respectively, for a combined value of approximately $764 million. JPMorgan Chase increased its IBIT holdings from roughly 8.3 million shares to 10.4 million shares; Morgan Stanley cut its IBIT position from 17.3 million shares to approximately 16.5 million shares, a 4.5% reduction.

According to Bloomberg, bond investors are monitoring roughly $700 billion in potential off-balance-sheet guarantee obligations at major AI firms. As financing for AI chips expands, these so-called "residual value support" arrangements could grow further. Following NVIDIA’s announcement this week of a $500 billion financing partnership, it may also provide tens of billions of dollars in residual value support for related debt transactions. Such financing typically uses special purpose vehicles (SPVs) to borrow funds for chip purchases, repaying debt with cash flow from user contracts. If clients stop payments, the assets are re-leased or sold to settle the debt; if a shortfall remains, the guarantor covers the gap. NVIDIA CEO Jensen Huang stated the company can provide residual value support for up to 25% of a given project, tailored to individual deals. CreditSights analysts note this is effectively a "put option sold" by NVIDIA: costs are low during the AI boom, but if the sector suddenly enters a severe downturn, clients default, and hardware values drop, the importance of these guarantees will rise sharply. Rating agencies have also begun treating some of these arrangements as debt-like contingent liabilities. Earlier, Meta used a similar structure for roughly $27 billion and $13 billion in data center debt financing; Broadcom provided most of the residual value support for a $35 billion AI chip financing tied to Anthropic. Moody’s warned that a surge in such transactions in the near term could limit Broadcom’s financial flexibility and pressure its credit profile.

According to Barron's, one of the factors driving the recent rebound in storage stocks is SanDisk (SNDK)’s strong long-term outlook for the NAND market. KeyBanc forecasts DRAM prices will rise 15% to 20% quarter-over-quarter in Q3 2026, with another 15% gain in Q4. For NAND, prices are expected to jump 30% to 40% QoQ in Q3, followed by a 15% increase in Q4. Calculated via simple compound interest, DRAM’s cumulative rise from Q2 levels will reach roughly 32% to 38% by year-end, while NAND’s cumulative increase will hit 50% to 61%. Although storage prices have already surged sharply, the rally is continuing beyond market expectations, making whether DRAM and NAND price hikes will persist into Q4 a key variable for the storage sector recently. On the valuation side, UBS analyst Timothy Arcuri set a $1,625 price target for Micron Technology (MU). Given current supply shortages keeping near-term earnings elevated, UBS did not use 2027-2028 earnings; instead, it based its valuation on 2029 expected EPS, applying a roughly 11x price-to-earnings ratio. UBS believes that even if the storage industry normalizes or enters a downturn cycle by 2029, Micron’s long-term profitability is likely to remain higher than in past cycles. UBS also noted that a rising proportion of long-term supply agreements (LTAs) could reduce the earnings volatility the storage industry has historically experienced due to overexpansion, oversupply, and price declines. FactSet data shows that Wall Street analysts’ average price target for Micron is $1,549, above its August 14 closing price of $970.20.

Citrini analyst Jukan said in a research note that NVIDIA’s $500 billion financing guarantee deal with Wall Street will help prevent an impact on GPU sales when hyperscale cloud providers face tightening cash flow and can no longer purchase GPUs entirely via advance payments. He noted that the arrangement will also accelerate the expansion of the GPU market’s Total Addressable Market (TAM), and is not a form of "circular financing". A notable shift is that Wall Street is increasingly regarding GPUs as collateral that can be monetized to unlock value.

According to Bloomberg, S&P 500 constituents posted a 31% year-over-year profit increase in the second quarter, exceeding the prior consensus estimate of 23%—the strongest growth for the index since Bloomberg Intelligence began tracking data in 1992, excluding recovery periods following major recessions. Over 90% of S&P 500 members have reported their earnings, with the first half’s overall profit performance on track to be the strongest since 2021. Analysts attribute the earnings beat to two factors: U.S. economic resilience and margin expansion driven by artificial intelligence (AI). The net profit margin of S&P 500 constituents has risen from a level that struggled to break 14% to nearly 16%. Mark Hackett, chief market strategist at Nationwide, noted that AI, which was largely a cost center for years, has reached an inflection point this year and is now becoming a profit driver. As profit growth outpaces the index’s price gains, the S&P 500’s 12-month forward price-to-earnings (P/E) ratio has fallen from around 26 at the start of the year to just under 22. Wall Street strategists have also steadily raised their forecasts: the average year-end target for the S&P 500 has climbed to 7,894 points, leaving roughly 1% upside from the index’s record high set this week; the full-year profit growth forecast has been lifted from 15% at the start of the year to 27%. This round of profit growth is no longer limited to large tech firms. As of August 12, roughly three-quarters of the approximately 1,500 U.S. public companies that have reported earnings beat both earnings per share (EPS) and revenue estimates. The healthcare sector was the only S&P 500 group to post a profit contraction in the second quarter.

According to Reuters, NVIDIA is in talks to invest up to $3 billion in SB Energy, a subsidiary of SoftBank Group. SB Energy is developing power and data center infrastructure for a large data center project planned by OpenAI in Ohio, U.S. This potential investment is part of the financing arrangements between NVIDIA, OpenAI, and SB Energy for the Ohio data center project, with the three parties currently discussing providing around $100 billion in credit support for the initiative. NVIDIA is considering investing $1.5 billion when the project agreement is signed, and another $1.5 billion when SB Energy holds its planned IPO. SB Energy could go public as early as next month, and plans to raise at least $5 billion via its IPO. Founded in 2019, the company is currently constructing multiple large-scale power and data center projects to meet infrastructure demands driven by the growth in AI workloads. Additionally, NVIDIA has adjusted its support plan for OpenAI’s Ohio data center project, with the initial guarantee expected to be less than $120 billion, a significant drop from the previously discussed $250 billion.