TMTB Morning Wrap
Good morning. Futures flat across the board. No formal announcement from US/Iran yet as Iranian Foreign Minister ruled out direct talks for now and Trump said he’s ready to let economic pressure build rather than launch fresh strikes. Oil +2% and yields +2-3bps across the curve.
Will be a nice reprieve on the earnings front with just a few companies reporting this week in Tech:
On the macro front this week we get CPI on Wed and PPI on Thurs. Otherwise, fairly slow morning. Optical outperforming on chatter on X from Jukan and others about it being the latest place to park those memory $s. MNDY -10% with a weaker Q3 showing the increasing winner vs. loser bifurcation in app sw. META +2% after releasing a new local model. HPE/NTAP +3-5% as MS upgrades the hw sector. INTC -4% after a $15B proposed stock offering.
Let’s get to it…
TSM: July sales of NT$468bn (+5.6% MoM, +44.7% YoY) tracking towards 15% q/q vs. street at 14% q/q
AAPL: Jefferies Downgrades to Underperform as Higher-End iPhone Roadmap Faces Yield and Memory-Cost Pressure
Jefferies downgrades AAPL to Underperform, saying supply-chain checks indicate the 20th-anniversary all-glass iPhone planned for Sep-27 has been canceled due to low yields, removing an important ASP and margin catalyst. Jefferies cuts FY28/FY29 EPS by 2.1%/3.4% and says rising memory costs could also limit broader DRAM upgrades, with the iPhone 19 Pro Max potentially moving to 16GB only on the highest-end model. Jefferies adds that higher trade-in values may pull forward iPhone 17 demand but could pressure iPhone 18 sales, reinforcing its view that the path to sustained higher iPhone ASPs looks increasingly difficult.
AAPL/Memory: CXMT Says No Thank You to Apple’s Demand For A Price Cut, as Huawei And Xiaomi Hand It Rare Leverage - WCCFFTECH
Chinese memory manufacturer CXMT has refused Apple's demand for a price cut in negotiations, says a report from the Korean press. Apple has widely been reported to be interested in procuring its chips from CXMT and has purportedly lobbied the US government for the permission to purchase the products for its gadgets sold outside the country. However, the report from the Korean publication Digital Daily outlines that demand from Huawei, Xiaomi and others has placed CXMT in a stronger position to negotiate with the world's largest consumer electronics firm.
SPCX/MSFT: SemiAnalysis Sees 10GW SpaceX Build in 2027 Driving Massive Compute Monetization Upside
SemiAnalysis says SpaceX is on track to reach roughly 10GW of compute by end-2027, enabled by unusually fast data-center construction and premium pricing of roughly $30M–$50M per MW annually. SemiAnalysis estimates frontier inference can generate more than $100B of revenue per GW per year, supporting a path to roughly $300B of SpaceX ARR by end-2027 even if only half of incremental capacity is monetized. SemiAnalysis says Microsoft’s renewed capacity push — with >10GW of binding commitments already signed — makes it a natural large-scale customer for SpaceX’s rapidly expanding compute infrastructure. SemiAnalysis argues SpaceX’s 3–5 month deployment timelines could help MSFT relieve its compute bottleneck and monetize OpenAI models at much higher revenue per MW through Azure, Foundry and Copilot, while supporting SpaceX’s path toward ~$300B of ARR by end-2027.
MSFT: Bernstein Raises PT to $660 as Capacity Analysis Pushes Back on the AI Overbuild Bear Case
Bernstein reiterates Outperform and raises its price target to $660, arguing Microsoft is taking a much more measured approach to AI capacity than headline capex and lease obligations imply. Bernstein says data-center square footage and annualized future lease costs are growing slower than Commercial Cloud revenue, while most purchase commitments are near-term and heavily tied to reusable power, cooling and general cloud infrastructure rather than GPUs. Bernstein also notes MSFT can flex facilities between CPU and GPU workloads, limiting downside if AI demand slows, and says the overbuild bear case materially overstates stranded-capacity risk.
CSCO: hearing Clev positive raising #s as Mythos seen as catalyst for broader infrastructure & bookings/orders tracking ahead
INTC: Proposes $15B Stock offering
“Intel intends to use the net proceeds from the offering for general corporate purposes, which may include, but are not limited to, capital expenditures and working capital. The offering is intended to further enable Intel to pursue the growth opportunities ahead while maintaining a strong balance sheet and its commitment to an investment-grade rating.”
NET: Cloudflare files to sell $2.175B of convertible senior notes due 2031
MNDY -9%: Looks light with a mixed print. Q2 revenue of $365M beat by 2.4% / ~$9M, but Q3 guides ~$4M below Street at the midpoint. FY26 revenue was left unchanged, implying a softer Q4, while operating margins came in modestly better.
AI product ARR doubled q/q and drove 17% of net new ARR, while $500K+ ARR customers grew 68% y/y to 114, suggesting the enterprise motion continues to strengthen even as SMB moderates. RPO grew 34% y/y to $937M.
META: Meta CEO Zuckerberg writes AI letter, it will resume releasing open-source AI models “soon” as part of a much larger plan: delivering personal superintelligence to billions of people & releases new edge model
. As intelligence becomes abundant, the most important question will be how we direct it... We believe that delivering superintelligence to everyone is the way to answer this question. This follows the tradition of putting the power of supercomputers and the internet in everyone’s pockets and on our desks. Rather than centralizing superintelligence, we should distribute it widely and give every person the ability to direct it. This has the potential to begin a new era of personal empowerment where individuals can use this powerful new capability to reach their full potential, pursue their interests, and improve their lives and the world more than ever before. To make this more tangible, here are a few specific ways Meta is building personal superintelligence to improve all of our lives: Everyone will have an exceptionally capable personal agent that understands you, your goals, and everything you care about... Everyone will have incredible tools for creation to express your ideas... Everyone will have powerful tools to create new businesses... Everyone will have a personalized tutor and coach with a PhD in every subject... Everyone will benefit from scientific advances and be able to contribute to scientific progress... Everyone will have free or affordable access to these tools... Meta is committed to building with the principles of individual empowerment as the source of prosperity, invention as AI’s purpose, and a balance of power favoring people as the foundation for addressing safety risks.”
At full precision, a 30-billion parameter model would require over 55 GB of memory — far more than any consumer GPU offers. We use quantization techniques to compress the model’s weights to approximately 4-bit precision, shrinking the language model to under 20 GB. This leaves enough headroom for the model’s working memory (its “KV cache”), the perception encoder for image understanding, and the speculative decoding drafter to run simultaneously within a 24 GB or 32 GB envelope. We validated that this compression introduces minimal to no degradation on agentic tasks.
Seems to rival Opus 4.5, which came out Nov ‘25
