TMTB Morning Wrap
Good morning. QQQs +77bps on a night that felt the opposite of the previous evening: much better Tech earnings print. The standouts came from software with TEAM (+27%) TWLO (+18%) and NET (+15%) all showing accelerating growth. We think what’s good for app sw is ultimately good for the AI semi trade as well, as Semi described in TMTB Slack last night:
Despite better software prints, it’s nice to see no factor whiplash this morning as better breadth continues to dominate the market over the past few days. Semis +2.8% with Software +1.4%. It feels we might be turning a page on the factor whiplash near-term. In macro land, yields dipping 6-9bps across the curve while crude is flat.
We’ll hit the usual first then move onto earnings recaps for NET, TEAM, TWLO, AKAM, MCHP, ABNB, & TTD
Happy Friday! Let’s get to it…
TECH/RESEARCH NEWS
U: HSBC, BofA and Deutsche Bank All Upgrade to Buy as Vector Growth Inflects and Runtime Data Expands the Runway
HSBC, BofA and Deutsche Bank all upgrade U to Buy following a much stronger 2Q, with HSBC raising its PT to $47.10 from $28.50, BofA to $50 from $30 and DB to $50 from $31. HSBC says Vector’s 23% q/q growth and sustained multi-quarter acceleration put earlier adoption concerns to rest, while HSBC believes broader 28-day attribution, better bidding data and future runtime integration support a 52% FY25–27 Grow revenue CAGR versus roughly 38% for APP and justify valuation closer to AppLovin. BofA says the quarter provides the first financial validation that runtime data can materially improve Vector targeting, with only about 30% of Unity 6 games represented today and roughly 75% of campaigns still not using D28 ROAS; BofA therefore expects >20% q/q Vector growth in both 3Q and 4Q and raises CY27 EBITDA to $1.29B from $789M. DB says Vector’s +90% y/y growth reflects a repeatable model-improvement cycle rather than temporary spend, and DB raises FY26/FY27 revenue estimates by 4%/11% and EBITDA by 17%/36%, arguing improving top-line durability plus removal of non-core businesses creates visibility toward roughly 50% annual adjusted EBITDA growth from FY25–28.
GOOGL: SemiAnalysis Says DeepMind Brain Drain Is a Near-Term GCP Bull Case as TPU Sales Drive Explosive Growth
SemiAnalysis argues DeepMind is no longer operating like a true frontier lab after major leadership departures and weakening Gemini model momentum, but says the biggest beneficiary is GCP as compute is increasingly redirected toward external customers. SemiAnalysis estimates more than 20% of TPU shipments from 3Q26–4Q27 are being sold directly to Anthropic and says GCP could reach more than $73B of third-party AI ARR plus roughly $120B of TPU system sales by end-2027. SemiAnalysis estimates TPU system sales could push 2027 GCP growth into the mid-100% range versus Street near 64%, add roughly $3 to Google EPS in 2027, and drive more than $250B of incremental TPU bookings into RPO
GOOGL: Google’s AI shake-up boosts Brin as DeepMind’s Hassabis steps aside
The FT reports Google is shifting control of its AI effort from London back to Silicon Valley, with Demis Hassabis stepping back from day-to-day leadership of DeepMind and Gemini development moving under Koray Kavukcuoglu in Mountain View, alongside a newly re-engaged Sergey Brin. The FT says the reorg reflects a broader shift from DeepMind’s research-first culture toward commercial urgency, as Google tries to close gaps with OpenAI and Anthropic in coding and enterprise AI. Hassabis becomes Alphabet chief scientist/DeepMind chair and will focus more on long-term research and Isomorphic Labs, while the changes have stirred concern over talent retention and DeepMind’s research culture. Alphabet shares fell 5% after the announcements amid investor concern around AI spending and Google’s ability to convert its research leadership into commercial returns.
OpenAI: OAI’s first hardware product revealed as portable doughnut-shaped AI smart speaker priced above $300
SK Hynix: announced in a regulatory filing that they are actively reviewing additional shareholder return measures to enhance shareholder value, and plan to finalize and announce the specific details during the third quarter. Also, planning a $38B expansion of its chipmaking facilities in South Korea to rapidly double its production capacity and ease a global shortage of memory chips
Memory: Morgan Stanley Says Memory Pullback Is a Tactical Re-Entry as LTAs Improve Cycle Durability
Morgan Stanley remains bullish on memory despite expecting pricing momentum to moderate from 4Q26, arguing the recent correction reflects an aging cycle rather than a break in the AI demand thesis. Morgan Stanley says the key structural change is the rapid expansion of multi-year LTAs, which should improve visibility and reduce volatility: Samsung now has 60%–70% of capacity under LTAs, Micron’s 16 deals cover roughly 20% of DRAM and one-third of NAND volume through 2030, and SanDisk has already committed about half of FY27 and two-thirds of FY28 bits. Morgan Stanley says these agreements will not eliminate cyclicality, but they should better anchor pricing and supply discipline while AI capex and agentic workloads continue to support long-term demand.
ORCL: Deutsche Bank Reiterates Buy as Funding Mix Looks Manageable Despite ~$600B Backlog Buildout
Deutsche Bank reiterates Buy and a $300 price target, saying the key investor debate after hosting meetings with mgmt is how Oracle funds the infrastructure required to deliver against roughly $600B of committed customer contracts while preserving investment-grade credit. DB says the company is likely to rely on a mix of ATM equity and debt, with roughly $20B of FY27 funding already contemplated and potentially another ~$40B needed in FY28 under its base case. DB’s sensitivity work suggests even a much heavier equity-funded scenario would create limited long-term downside if Oracle executes on its fundamentals, with FY30 EPS dilution ranging from roughly 1% to 15% depending on financing mix and issuance price.
STX: Citi Reiterates Buy and $1,300 PT as AI Storage Demand Broadens and Nearline Supply Stays Fully Allocated
Citi reiterates Buy and a $1,300 price target after management meetings reinforced an exceptionally strong demand and pricing backdrop, with nearline exabyte capacity essentially fully allocated through CY27 and into CY28. Citi says AI storage demand is broadening beyond training into inference, agentic AI and eventually physical AI, while KV-cache architectures could shift more context storage from expensive GPUs toward lower-cost HDDs. Citi also says disciplined capacity, HAMR mix and contractual pricing ladders should support continued margin expansion, with STX still underserving demand and seeing little evidence of double-ordering or excess inventory.
China AI: Alibaba plans to charge big users of its next open-source AI model, sources say
Reuters reports Alibaba plans to introduce a revenue-sharing requirement for large commercial users of its upcoming Qwen3.8-Max open-source AI model, marking a shift from largely free self-hosted usage. Reuters says the approach mirrors Moonshot’s Kimi K3 licensing model, which requires commercial agreements for service providers generating >$20M in annual sales. Moonshot is reportedly seeking up to a 30% revenue share, while Alibaba’s rate is still being negotiated. The move suggests Chinese AI labs are increasingly using a freemium/open-source model to monetize heavy commercial adoption while still competing aggressively on price with closed U.S. models.
China AI: Bytedance training a ~10T-parameter AI model, roughly 3x larger than Moonshot’s Kimi K3 and approaching the scale of Anthropic’s Mythos - Financial Times
AMD: AMD buys chip startup that hardwires AI models into its silicon
CNBC:
On Thursday, AMD said it’s entered into an agreement to acquire Taalas, a Toronto-based startup that makes chips for inference. Taalas’ accelerators are customized, or hard-wired for a single AI model, rather than being general purpose.
In exchange for that loss of flexibility, Taalas’ technology promises a less-expensive chip that it says can produce output for specific models thousands of times faster than a traditional GPU. An AMD representative declined to provide a purchase price for the transaction. Taalas has raised a total of $219 million in venture funding since its 2023 founding.
EARNINGS RECAPS (NET, TEAM, TWLO, AKAM, MCHP, ABNB, TTD)
NET +15% delivered its strongest print in years with biggest rev beat since Q4’21, revenue/RPO accelerating to 36%/38%, the FY26 guide rising by more than the quarterly beat, and the clearest evidence yet that Workers and agentic traffic are becoming material revenue drivers.
NET came through with a monster print accelerating growth despite expectations for a decel and guide implies flat growth on a much tougher comp if you assume a slightly smaller beat and should stay in the mid 30s for the next couple of quarters. Valuation will always be the gripe, but narrative here one of the best in sw and name usually does well when top line/RPO is accel’ing.