GS TMT SPEC SALES – Adyen topline good enough

GS TMT SPEC SALES – Thursday 13th of August Market overnight: US trading was quiet but risk appetite improved after an inline CPI print, with QQQs +75bps and the SOX +2.5% as investors re-risked into the AI/semis complex. Momentum was strong, with 12-month winners up +4% and losers down -1.5%, helped by better AI infrastructure results from CoreWeave (+20%), Lumentum (+13%), Supermicro (+19%) and Nebius, while software and internet lagged around -1%. The main debate is whether the strength in neoclouds and AI infrastructure points to a healthier demand backdrop for NVIDIA, or whether it creates more tension with hyperscalers, which were weak overnight, with MSFT, AMZN and META down around 2%. Key results and trends: The results overnight broadly supportive for AI infrastructure and optical networking, but with high expectations creating more mixed stock reactions. Coherent beat and guided above Street, with management pointing to “exceptional” AI datacentre demand and the shift from copper to optical, but the stock fell after a big pre-results rally. Cisco delivered a strong beat-and-raise and disclosed $4bn of AI hyperscaler orders, but down 4% AH as investors focused on lower gross-margin guidance and whether the $7.5bn AI revenue guide is too conservative. Cerebras also beat and raised FY guidance, but the market wanted a bigger customer announcement and stronger RPO momentum, while Hon Hai remained another positive AI-server datapoint, guiding 2026 revenue above 2025 and AI-server revenue growth of more than 40%.

ADYEN: MIXED but I think good enough (I would be adding) as the positive is that Adyen delivered the topline growth investors needed to see, with Q2 TPV slightly ahead of expectations at +26% and net revenue growth of +22% ex-FX, which sits at the high end of buyside and sell-side expectations and helps de-risk the H2 growth outlook. Importantly, Digital grew +16%, holding up in the area where there had been the most concern around price pressure, while Platforms remained strong at +39%. The debate is likely to centre on the lighter EBITDA margin (strip out the funny and it’s a beat) and weaker FCF, given higher cost per head and the step-up in capex. However, this looks less like a structural issue and more like a temporary pull-forward of 2027 spend into H2 2026, with Adyen securing compute and storage capacity and locking in pricing while supply chains remain tight; management still expects capex to return toward historical levels after 2026. Interesting on the analyst call back mgt dropped in the OpenAI is starting to ramp add that to the Toast deal that helps sentiment…

Prosus / Tencent: Prosus was down over 7% after Tencent’s results, while Tencent itself is down around 4%, as investors looked past broadly resilient operations and focused on the AI investment ramp. The positives were a small revenue beat, gross-margin expansion, strong ads and domestic games growth, and core EBIT up 19% yoy; however, these were offset by much higher capex of CNY52.8bn, around 64% above consensus, plus higher rental prepayments and

a GAAP net income miss. The key debate is whether Tencent’s more aggressive AI spend — focused on Hy model training, WorkBuddy and Weixin Xiaowei rather than near-term cloud monetisation — will create durable franchise value, or simply pressure EPS in the near term. For Prosus, the read-through is negative near-term, but feels to have been full reflected yesterday as Tencent remains its largest asset. Prosus ongoing performance will be tied to where Tencent goes See GIR take this morning on Tencent reiterating their buy OTHER: • Zuckerberg / AI distillation: Mark Zuckerberg is defending distillation , the practice of training a smaller “student” AI model by learning from a larger “teacher” model, arguing that the US should not restrict companies from learning from models they can legitimately observe. The debate has become politically sensitive because US AI companies and the White House have accused Chinese developers of using distillation to copy closed US frontier models without permission, while Meta has reportedly used distillation from open models such as Alibaba’s Qwen. The key distinction…