Nvidia day

It’s the same story, different day. Trump yesterday first called off an attack on Iran, then said he would resume strikes if they can’t agree on a deal. NATO said it could help secure shipping routes around the Strait of Hormuz if disruptions persist into early July. Europeans reportedly eased some sanctions on Russia as energy security concerns remain high.

Alas, geopolitical uncertainties continue. Traffic through the Strait of Hormuz remains at a near standstill, world oil inventories continue tightening, and oil prices keep rising. The latter fuels global inflation expectations and pushes global yields higher on rising bets that central banks may have to fight price pressures despite the worsening economic outlook. The OECD revised its economic outlook lower, factoring in the prolonged Middle East conflict and its global repercussions.

In this context, the US 30-year yield has climbed back to levels last seen before the subprime crisis, while the US 10-year yield approached the 4.70% mark as the Japanese 10-year yield soared toward 2.80% — levels last seen before 2000 — and consolidated near that level this morning.

The gap between US and Japanese yields is approaching an uncomfortable zone — around 190bp for the 10-year spread — increasing the risk of a carry unwind that could send another shockwave across global risk markets. I believe that a move toward the 2.90–3% range in the 10-year JGB yield could trigger a broader reverse carry trade, pulling USDJPY lower along with global equities.

On the FX front, the USDJPY is softer this morning, but not because of the narrowing yield differential — which makes US dollar returns less attractive — but because Japanese officials again warned they could intervene if the pair rises further. This morning, USDJPY is trading between its 50-DMA on the downside and 159.10 on the upside. Any move higher could trigger another intervention and another short-term relief rally for the yen, without necessarily leading to a sustainable bearish shift in USDJPY.

On the equities front, the S&P 500 fell for a third straight day on the back of rising yields, the Nikkei is down more than 1.6% at the time of writing, and the Kospi is retreating nearly 2% as investors’ focus shifts from shiny earnings to geopolitical turmoil.

It’s in this chaotic market — where earnings optimism has started to lose momentum — that Nvidia will announce its Q1 results today, after the bell.

All eyes on Nvidia!

Expectations are, of course, sky high. The company is expected to report around $79bn in revenue — roughly 15% higher than last quarter and nearly 80% above the same quarter last year.

Margins are also expected to remain exceptionally strong, around 75%, confirming that Nvidia still enjoys enormous pricing power despite the massive Blackwell ramp and rising competition.

But Nvidia’s earnings no longer carry the same existential weight they did at the very beginning of the AI craze. Back then, markets were obsessed with training AI models. GPUs became essential because they are incredibly efficient at handling thousands of calculations simultaneously — exactly what AI training requires. Imagine trying to get from point A to B by simultaneously testing millions of possible paths through C, D, F, X, Y or Z. GPUs are built for that kind of parallel processing power. CPUs, on the other hand, are designed for sequential computations.

As such, once the models are trained with GPUs, the focus increasingly shifts toward inference — running the trained model — where TPUs and CPUs can also play a major role, while memory chips are needed to store and process information efficiently.

That’s why GPUs say more about the raw power and evolution of AI models, while CPUs and memory chips increasingly say more about real-world AI adoption and scaling. This growing importance of CPUs and memory infrastructure is also why traditional CPU and memory chip makers have taken over part of the AI narrative — and why Nvidia is developing its own CPU technologies within its next-generation Vera Rubin platform.

Investors will therefore closely watch whether the company can maintain strong margins while scaling production and preparing the transition toward the next-generation Vera Rubin platform — designed for the next phase of AI focused on massive-scale inference, reasoning and AI “factories.”

And the competition for running models efficiently at lower cost is fierce. Besides traditional chipmakers like AMD and Intel, Nvidia’s biggest clients — Big Tech companies like Amazon, Google and Meta — are all working on their own in-house chips to build the most energy- and cost-efficient alternatives to Nvidia’s ultra-powerful premium products.

Nvidia closed near $220 per share yesterday, around $16 below the all-time high reached last week. Options markets are currently pricing an implied move of roughly 6% to 8% in either direction following the earnings announcement. But the direction is difficult to predict. Expectations are now so elevated that even a solid beat may not trigger the euphoric rallies seen during the early AI boom.

Regardless of the strength of Nvidia’s results, three risks are piling up as Big Tech companies continue to trade at near-perfect valuations:

1. Prospects of tighter financial conditions linked to the Middle East conflict, which could slow AI adoption and scale back revenue expectations
2. Capacity constraints due to supply chain disruptions, which could limit production, increase costs and weigh on profit expectations
3. Rising competition, which could eat into margins.

And that’s why even great results from Nvidia may not prevent profit-taking that would feel healthy at current valuations.