Despite recent concerns around a softening macro, AI spending fatigue and Nvidia’s Blackwell slight delay, the second part of the year and early 2025 are actually shaping up well for the Tech industry.
Oracle’s just released FQ1 numbers were highly encouraging and suggested that AI training in data centers remains buoyant, with the company claiming $3 billion AI bookings in the quarter and its OCI unit (cloud infrastructure) growth accelerating to 49% year-on-year from 44% in the previous quarter.
Given that Oracle’s FQ1 encompasses July and August, these numbers give confidence about continued AI computing efforts on the major cloud platforms (Microsoft Azure, Amazon AWS and Google Cloud) in Q3, with the risk clearly on the upside.
On the infrastructure capex front, Oracle announced a doubling of its FY25 capex to $15 billion. As Oracle’s capex has been only $2.3 billion in FQ1, this guidance hints at a massive ramp in coming quarters, an obvious positive for AI chips and networking equipment.
Talking about the outlook for semiconductors, latest data from Taiwan point to continued strength with semiconductor companies’ August revenue up 25% year-on-year and up 33% at TSMC as AI remains a powerful driver. With strong shipments expected for the next generation of iPhone, hyperscalers (e.g. Oracle) keeping raising their capex intentions and Nvidia Blackwell chips entering mass production in early 2025, the momentum should remain sustained going forward.
Another powerful catalyst could be the release of the AI-enabled iPhone 16. As we said in a previous report, we are not impressed for now by Apple Intelligence whose main features are an enhanced Siri (adding text input, email summarization…) and Visual Intelligence that allows users to point their iPhone at anything and “learn about everything they see”.
That said, it’s highly likely that the introduction of AI in the Apple ecosystem will give a major boost to the roll-out and usage of AI apps across various platforms in the next couple of years, regardless of the immediate adoption of these features by iPhone users. We can indeed expect most smartphones, tablets, PCs, cars and other connected devices to come equipped at some point with inference chips to process AI tasks on-device and app/software makers and traditional companies (industrials, autos, financials…) to boost spending on the training of their own models.
In conclusion, we are in the very early stages of AI apps proliferation and user adoption, meaning that capex from hyperscalers should remain sustained in the near future. Hence, we believe that the brutal derating of the last few weeks is a buying opportunity, with companies such as TSMC, Nvidia and Broadcom now trading at 16x 2025 EPS, 22x and 24x, respectively.